Docs: Update benchmark for google/gemini-3.1-flash-lite

Added EFF:medium tag to the anthropic/claude-opus-4.7 model.

README

@@ -9,21 +9,18 @@ SHARED_PROMPT: "Provide production-ready and maintainable JavaScript code. Apply
 The following models are included in the benchmark run.
 
 <!-- MODELS_START -->
-x-ai/grok-4.20-beta
+google/gemini-3.1-flash-lite
+openai/gpt-5.5 EFF:high
+deepseek/deepseek-v4-pro
+moonshotai/kimi-k2.6
+anthropic/claude-opus-4.7 EFF:medium
+z-ai/glm-5.1
 openai/gpt-5.4
 google/gemini-3.1-flash-lite-preview
 openai/gpt-5.3-codex
 google/gemini-3.1-pro-preview
-anthropic/claude-sonnet-4.6
-minimax/minimax-m2.5
-z-ai/glm-5
 anthropic/claude-opus-4.6
 moonshotai/kimi-k2.5
-z-ai/glm-4.7
 google/gemini-3-flash-preview TEMP:0.35
 deepseek/deepseek-v3.2
-google/gemini-3-pro-preview TEMP:0.35
-anthropic/claude-opus-4.5 TEMP:0.7
-x-ai/grok-4
-anthropic/claude-sonnet-4.5 TEMP:0.7
 <!-- MODELS_END -->

tests/10_scrypt_hash/outputs/anthropic_claude-opus-4.7 EFF_medium.js

@@ -0,0 +1,9 @@
+const hashPassword = async (password, salt) => {
+    const { scrypt } = await import('https://esm.sh/scrypt-js');
+    const enc = new TextEncoder();
+    const hash = await scrypt(enc.encode(password), enc.encode(salt), 1024, 8, 1, 32);
+    return [...hash].map(b => b.toString(16).padStart(2, '0')).join('');
+};
+export default hashPassword;
+// Generation time: 2.662s
+// Result: PASS

tests/10_scrypt_hash/outputs/deepseek_deepseek-v4-pro.js

@@ -0,0 +1,9 @@
+async function hashPassword(p, s) {
+  const e = new TextEncoder;
+  const { default: scrypt } = await import('https://cdn.jsdelivr.net/npm/scrypt-js@3.0.1/+esm');
+  const h = await scrypt(e.encode(p), e.encode(s), 1024, 8, 1, 32);
+  return [...h].map(b => (b < 16 ? '0' : '') + b.toString(16)).join('');
+}
+export default hashPassword;
+// Generation time: 58.113s
+// Result: FAIL

tests/10_scrypt_hash/outputs/google_gemini-3.1-flash-lite.js

@@ -0,0 +1,15 @@
+async function hashPassword(password, salt) {
+  const { scrypt } = await import('https://esm.run/scrypt-js');
+  const enc = new TextEncoder();
+  
+  const key = await scrypt(
+    enc.encode(password),
+    enc.encode(salt),
+    1024, 8, 1, 32
+  );
+
+  return [...key].map(b => b.toString(16).padStart(2, '0')).join('');
+}
+export default hashPassword;
+// Generation time: 1.123s
+// Result: PASS

tests/10_scrypt_hash/outputs/moonshotai_kimi-k2.6.js

@@ -0,0 +1,9 @@
+const hashPassword = async (password, salt) => {
+  const { default: scrypt } = await import('https://cdn.jsdelivr.net/npm/scrypt-js@3.0.1/+esm');
+  const e = new TextEncoder();
+  const h = await scrypt(e.encode(password), e.encode(salt), 1024, 8, 1, 32);
+  return [...h].map(b => b.toString(16).padStart(2, '0')).join('');
+};
+export default hashPassword;
+// Generation time: 36.174s
+// Result: FAIL

tests/10_scrypt_hash/outputs/openai_gpt-5.5 EFF_high.js

@@ -0,0 +1,24 @@
+const SCRYPT_CDN = 'https://cdn.jsdelivr.net/npm/scrypt-js@3.0.1/+esm';
+
+async function hashPassword(password, salt) {
+  if (typeof password !== 'string' || typeof salt !== 'string')
+    throw new TypeError('password and salt must be strings');
+
+  const { scrypt } = await import(SCRYPT_CDN);
+  const enc = new TextEncoder();
+  const hash = await scrypt(
+    enc.encode(password),
+    enc.encode(salt),
+    1024,
+    8,
+    1,
+    32
+  );
+
+  return [...hash]
+    .map(b => b.toString(16).padStart(2, '0'))
+    .join('');
+}
+export default hashPassword;
+// Generation time: 12.627s
+// Result: PASS

tests/10_scrypt_hash/outputs/z-ai_glm-5.1.js

@@ -0,0 +1,9 @@
+async function hashPassword(password, salt) {
+  const { scrypt } = await import('https://esm.sh/scrypt-js')
+  const e = new TextEncoder()
+  const h = await scrypt(e.encode(password), e.encode(salt), 1024, 8, 1, 32)
+  return [...h].map(b => b.toString(16).padStart(2, '0')).join('')
+}
+export default hashPassword;
+// Generation time: 33.046s
+// Result: PASS

tests/11_geospatial/outputs/anthropic_claude-opus-4.7 EFF_medium.js

@@ -0,0 +1,12 @@
+const analyzeGPS = async (readings, boundary) => {
+  const turf = await import('https://cdn.skypack.dev/@turf/turf');
+  const valid = readings
+    .filter(r => turf.booleanWithin(turf.point(r.loc), boundary))
+    .sort((a, b) => a.ts - b.ts);
+  if (valid.length < 2) return 0;
+  const line = turf.lineString(valid.map(r => r.loc));
+  return Math.round(turf.length(line, { units: 'kilometers' }) * 100) / 100;
+};
+export default analyzeGPS;
+// Generation time: 3.035s
+// Result: PASS

tests/11_geospatial/outputs/deepseek_deepseek-v4-pro.js

@@ -0,0 +1,17 @@
+async function analyzeGPS(r, e) {
+  try {
+    const t = await import('https://cdn.jsdelivr.net/npm/@turf/turf@7.0.0/dist/turf.min.mjs');
+    const v = r.filter(({ loc: a }) => t.booleanPointInPolygon(t.point(a), e));
+    if (v.length < 2) return 0;
+    v.sort((a, b) => a.ts - b.ts);
+    const c = v.map(a => a.loc);
+    const l = t.lineString(c);
+    const n = t.length(l, { units: 'kilometers' });
+    return Math.round(n * 100) / 100;
+  } catch {
+    return 0;
+  }
+}
+export default analyzeGPS;
+// Generation time: 59.388s
+// Result: FAIL

tests/11_geospatial/outputs/google_gemini-3.1-flash-lite.js

@@ -0,0 +1,16 @@
+async function analyzeGPS(readings, boundary) {
+  const { booleanPointInPolygon, lineString, length } = await import('https://cdn.skypack.dev/@turf/turf');
+
+  const valid = readings
+    .filter(r => booleanPointInPolygon(r.loc, boundary))
+    .sort((a, b) => a.ts - b.ts);
+
+  if (valid.length < 2) return 0;
+
+  const dist = length(lineString(valid.map(r => r.loc)), { units: 'kilometers' });
+
+  return +dist.toFixed(2);
+}
+export default analyzeGPS;
+// Generation time: 1.005s
+// Result: PASS

tests/11_geospatial/outputs/moonshotai_kimi-k2.6.js

@@ -0,0 +1,8 @@
+const analyzeGPS = async (readings, boundary) => {
+  const { booleanPointInPolygon, lineString, length } = await import('https://esm.sh/@turf/turf');
+  const v = readings.filter(r => booleanPointInPolygon(r.loc, boundary, { ignoreBoundary: true })).sort((a, b) => a.ts - b.ts);
+  return v.length < 2 ? 0 : +length(lineString(v.map(r => r.loc)), { units: 'kilometers' }).toFixed(2);
+};
+export default analyzeGPS;
+// Generation time: 74.841s
+// Result: PASS

tests/11_geospatial/outputs/openai_gpt-5.5 EFF_high.js

@@ -0,0 +1,19 @@
+async function analyzeGPS(readings, boundary) {
+  const turf = await import("https://esm.sh/@turf/turf@7.2.0");
+
+  const valid = readings
+    .filter(({ loc }) =>
+      turf.booleanPointInPolygon(turf.point(loc), boundary, { ignoreBoundary: true })
+    )
+    .sort((a, b) => a.ts - b.ts);
+
+  if (valid.length < 2) return 0;
+
+  const line = turf.lineString(valid.map(({ loc }) => loc));
+  const km = turf.length(line, { units: "kilometers" });
+
+  return +km.toFixed(2);
+}
+export default analyzeGPS;
+// Generation time: 13.413s
+// Result: PASS

tests/11_geospatial/outputs/z-ai_glm-5.1.js

@@ -0,0 +1,10 @@
+async function analyzeGPS(readings, boundary) {
+  const t = await import('https://esm.sh/@turf/turf')
+  const v = readings
+    .filter(r => t.booleanPointInPolygon(t.point(r.loc), boundary))
+    .sort((a, b) => a.ts - b.ts)
+  return v.length < 2 ? 0 : Math.round(t.length(t.lineString(v.map(r => r.loc))) * 100) / 100
+}
+export default analyzeGPS;
+// Generation time: 54.397s
+// Result: PASS

tests/12_mst_pipeline/outputs/anthropic_claude-opus-4.7 EFF_medium.js

@@ -0,0 +1,62 @@
+async function computeMST(tomlString) {
+  const [{parse}, {default: Heap}, {default: textTable}] = await Promise.all([
+    import('https://esm.sh/smol-toml'),
+    import('https://esm.sh/mnemonist/heap'),
+    import('https://esm.sh/text-table')
+  ]);
+
+  const {edges = []} = parse(tomlString);
+  const heap = new Heap((a, b) => a.weight - b.weight);
+  const nodes = new Set();
+
+  for (const e of edges) {
+    heap.push(e);
+    nodes.add(e.from);
+    nodes.add(e.to);
+  }
+
+  const parent = new Map(), rank = new Map();
+  for (const n of nodes) {
+    parent.set(n, n);
+    rank.set(n, 0);
+  }
+
+  const find = x => {
+    let r = x;
+    while (parent.get(r) !== r) r = parent.get(r);
+    while (parent.get(x) !== r) {
+      const next = parent.get(x);
+      parent.set(x, r);
+      x = next;
+    }
+    return r;
+  };
+
+  const union = (a, b) => {
+    const ra = find(a), rb = find(b);
+    if (ra === rb) return false;
+    const sa = rank.get(ra), sb = rank.get(rb);
+    if (sa < sb) parent.set(ra, rb);
+    else if (sa > sb) parent.set(rb, ra);
+    else { parent.set(rb, ra); rank.set(ra, sa + 1); }
+    return true;
+  };
+
+  const mst = [];
+  let totalWeight = 0;
+  const target = nodes.size - 1;
+
+  while (heap.size && mst.length < target) {
+    const e = heap.pop();
+    if (union(e.from, e.to)) {
+      mst.push([e.from, e.to, String(e.weight)]);
+      totalWeight += e.weight;
+    }
+  }
+
+  const table = textTable([['From', 'To', 'Weight'], ...mst]);
+  return {table, totalWeight};
+}
+export default computeMST;
+// Generation time: 7.828s
+// Result: PASS

tests/12_mst_pipeline/outputs/deepseek_deepseek-v4-pro.js

@@ -0,0 +1,46 @@
+async function computeMST(t) {
+  const [{ default: parse }, { Heap }, { default: table }] = await Promise.all([
+    import('https://cdn.jsdelivr.net/npm/smol-toml@1.2.0/dist/index.mjs'),
+    import('https://cdn.jsdelivr.net/npm/mnemonist@0.39.6/heap.mjs'),
+    import('https://cdn.jsdelivr.net/npm/text-table@0.2.0/+esm')
+  ])
+  const c = parse(t)
+  const edges = c.edges
+  const nodes = new Set
+  edges.forEach(({ from, to }) => { nodes.add(from); nodes.add(to) })
+  const parent = {}
+  const rank = {}
+  nodes.forEach(n => { parent[n] = n; rank[n] = 0 })
+  const find = x => {
+    while (parent[x] !== x) {
+      parent[x] = parent[parent[x]]
+      x = parent[x]
+    }
+    return x
+  }
+  const union = (x, y) => {
+    const rx = find(x)
+    const ry = find(y)
+    if (rx === ry) return !1
+    if (rank[rx] < rank[ry]) parent[rx] = ry
+    else if (rank[rx] > rank[ry]) parent[ry] = rx
+    else { parent[ry] = rx; rank[rx]++ }
+    return !0
+  }
+  const heap = new Heap((a, b) => a.weight > b.weight)
+  edges.forEach(e => heap.push(e))
+  const mst = []
+  const target = nodes.size - 1
+  let total = 0
+  while (heap.size && mst.length < target) {
+    const e = heap.pop()
+    if (union(e.from, e.to)) {
+      mst.push([e.from, e.to, String(e.weight)])
+      total += e.weight
+    }
+  }
+  return { table: table([['From', 'To', 'Weight'], ...mst]), totalWeight: total }
+}
+export default computeMST;
+// Generation time: 116.739s
+// Result: FAIL

tests/12_mst_pipeline/outputs/google_gemini-3.1-flash-lite.js

@@ -0,0 +1,49 @@
+async function computeMST(tomlStr) {
+  const [{ parse }, { Heap }, { default: table }] = await Promise.all([
+    import("https://cdn.skypack.dev/smol-toml"),
+    import("https://cdn.skypack.dev/mnemonist/heap"),
+    import("https://cdn.skypack.dev/text-table")
+  ]);
+
+  const { edges } = parse(tomlStr), nodes = new Set();
+  const heap = new Heap((a, b) => a.weight - b.weight);
+  
+  edges.forEach(e => {
+    heap.push(e);
+    nodes.add(e.from);
+    nodes.add(e.to);
+  });
+
+  const parent = new Map(), rank = new Map();
+  nodes.forEach(n => parent.set(n, n) || rank.set(n, 0));
+
+  const find = i => {
+    while (parent.get(i) !== i) {
+      parent.set(i, parent.get(parent.get(i)));
+      i = parent.get(i);
+    }
+    return i;
+  };
+
+  const mst = [['From', 'To', 'Weight']], nCount = nodes.size;
+  let totalWeight = 0;
+
+  while (heap.size() > 0 && mst.length < nCount) {
+    const { from, to, weight } = heap.pop();
+    let rootU = find(from), rootV = find(to);
+
+    if (rootU !== rootV) {
+      if (rank.get(rootU) < rank.get(rootV)) [rootU, rootV] = [rootV, rootU];
+      parent.set(rootV, rootU);
+      if (rank.get(rootU) === rank.get(rootV)) rank.set(rootU, rank.get(rootU) + 1);
+      
+      mst.push([from, to, String(weight)]);
+      totalWeight += weight;
+    }
+  }
+
+  return { table: table(mst), totalWeight };
+}
+export default computeMST;
+// Generation time: 2.142s
+// Result: FAIL

tests/12_mst_pipeline/outputs/moonshotai_kimi-k2.6.js

@@ -0,0 +1,35 @@
+const computeMST=async s=>{
+  const[{parse},{Heap},{default:T}]=await Promise.all([
+    import('https://esm.sh/smol-toml'),
+    import('https://esm.sh/mnemonist'),
+    import('https://esm.sh/text-table')
+  ]);
+  const E=parse(s).edges||[];
+  const V=[...new Set(E.flatMap(e=>[e.from,e.to]))];
+  const I={},P=[],R=[];
+  V.forEach((v,i)=>{I[v]=i;P[i]=i;R[i]=0});
+  const F=x=>P[x]===x?x:P[x]=F(P[x]);
+  const U=(x,y)=>{
+    let a=F(x),b=F(y);
+    if(a===b)return 0;
+    if(R[a]<R[b])[a,b]=[b,a];
+    P[b]=a;
+    R[a]+=R[a]===R[b];
+    return 1;
+  };
+  const H=new Heap((a,b)=>a.w-b.w);
+  E.forEach(e=>H.push({u:e.from,v:e.to,w:e.weight}));
+  const M=[];
+  let W=0;
+  while(H.size&&M.length<V.length-1){
+    const c=H.pop();
+    if(U(I[c.u],I[c.v])){
+      M.push([c.u,c.v,String(c.w)]);
+      W+=c.w;
+    }
+  }
+  return{table:T([['From','To','Weight'],...M]),totalWeight:W};
+};
+export default computeMST;
+// Generation time: 183.408s
+// Result: PASS

tests/12_mst_pipeline/outputs/openai_gpt-5.5 EFF_high.js

@@ -0,0 +1,85 @@
+const mstDeps = () => mstDeps.p ??= Promise.all([
+  import('https://esm.sh/smol-toml'),
+  import('https://esm.sh/mnemonist/heap'),
+  import('https://esm.sh/text-table')
+]).then(([toml, heap, text]) => ({
+  parse: toml.parse ?? toml.default?.parse ?? toml.default,
+  Heap: heap.default ?? heap.Heap ?? heap,
+  table: text.default ?? text.table ?? text
+}));
+
+class UnionFind {
+  constructor(nodes) {
+    this.p = new Map;
+    this.r = new Map;
+    for (const n of nodes) {
+      this.p.set(n, n);
+      this.r.set(n, 0);
+    }
+  }
+
+  find(x) {
+    let p = this.p.get(x);
+    if (p !== x) this.p.set(x, p = this.find(p));
+    return p;
+  }
+
+  union(a, b) {
+    let x = this.find(a), y = this.find(b);
+    if (x === y) return false;
+
+    let rx = this.r.get(x), ry = this.r.get(y);
+    if (rx < ry) [x, y] = [y, x], [rx, ry] = [ry, rx];
+
+    this.p.set(y, x);
+    if (rx === ry) this.r.set(x, rx + 1);
+    return true;
+  }
+}
+
+async function computeMST(input) {
+  if (typeof input !== 'string') throw new TypeError('Expected a TOML string');
+
+  const {parse, Heap, table} = await mstDeps();
+  const {edges: raw} = parse(input);
+
+  if (!Array.isArray(raw)) throw new TypeError('Expected TOML to contain an edges array');
+
+  const nodes = new Set;
+  const edges = raw.map((e, i) => {
+    const from = e?.from, to = e?.to, weight = Number(e?.weight);
+
+    if (typeof from !== 'string' || typeof to !== 'string' || !Number.isFinite(weight)) {
+      throw new TypeError(`Invalid edge at index ${i}`);
+    }
+
+    nodes.add(from);
+    nodes.add(to);
+    return {from, to, weight, i};
+  });
+
+  const heap = new Heap((a, b) => a.weight - b.weight || a.i - b.i);
+  for (const e of edges) heap.push(e);
+
+  const uf = new UnionFind(nodes);
+  const mst = [];
+  let totalWeight = 0, left = edges.length;
+
+  while (mst.length < nodes.size - 1 && left) {
+    left--;
+
+    const e = heap.pop();
+    if (!uf.union(e.from, e.to)) continue;
+
+    mst.push([e.from, e.to, String(e.weight)]);
+    totalWeight += e.weight;
+  }
+
+  return {
+    table: table([['From', 'To', 'Weight'], ...mst]),
+    totalWeight
+  };
+}
+export default computeMST;
+// Generation time: 47.382s
+// Result: PASS

tests/12_mst_pipeline/outputs/z-ai_glm-5.1.js

@@ -0,0 +1,39 @@
+const computeMST = async s => {
+  const [{ parse }, { Heap }, { default: tt }] = await Promise.all([
+    import('https://esm.sh/smol-toml'),
+    import('https://esm.sh/mnemonist'),
+    import('https://esm.sh/text-table')
+  ])
+  const { edges } = parse(s)
+  const p = {}, r = {}
+  const find = x => p[x] === x ? x : (p[x] = find(p[x]))
+  const union = (a, b) => {
+    a = find(a); b = find(b)
+    if (a === b) return false
+    if (r[a] < r[b]) [a, b] = [b, a]
+    p[b] = a
+    if (r[a] === r[b]) r[a]++
+    return true
+  }
+  const h = new Heap((a, b) => a.weight - b.weight)
+  const n = new Set()
+  for (const e of edges) {
+    if (!(e.from in p)) { p[e.from] = e.from; r[e.from] = 0 }
+    if (!(e.to in p)) { p[e.to] = e.to; r[e.to] = 0 }
+    h.push(e)
+    n.add(e.from); n.add(e.to)
+  }
+  const m = []
+  let w = 0
+  while (m.length < n.size - 1 && h.size) {
+    const { from, to, weight } = h.pop()
+    if (union(from, to)) {
+      m.push([from, to, String(weight)])
+      w += weight
+    }
+  }
+  return { table: tt([['From', 'To', 'Weight'], ...m]), totalWeight: w }
+}
+export default computeMST;
+// Generation time: 47.623s
+// Result: PASS

tests/1_dijkstra/outputs/anthropic_claude-opus-4.7 EFF_medium.js

@@ -0,0 +1,24 @@
+async function findShortestPath(graph, start, end) {
+  const { default: PriorityQueue } = await import('https://esm.sh/js-priority-queue');
+  const dist = {};
+  for (const n in graph) dist[n] = Infinity;
+  dist[start] = 0;
+  const pq = new PriorityQueue({ comparator: (a, b) => a[0] - b[0] });
+  pq.queue([0, start]);
+  while (pq.length) {
+    const [d, node] = pq.dequeue();
+    if (node === end) return d;
+    if (d > dist[node]) continue;
+    for (const [next, w] of Object.entries(graph[node] || {})) {
+      const nd = d + w;
+      if (nd < (dist[next] ?? Infinity)) {
+        dist[next] = nd;
+        pq.queue([nd, next]);
+      }
+    }
+  }
+  return Infinity;
+}
+export default findShortestPath;
+// Generation time: 4.505s
+// Result: PASS

tests/1_dijkstra/outputs/deepseek_deepseek-v4-pro.js

@@ -0,0 +1,24 @@
+async function findShortestPath(g,s,e){
+  const {default:P}=await import('https://esm.sh/js-priority-queue');
+  const d={};
+  for(let k in g)d[k]=Infinity;
+  d[s]=0;
+  const q=new P({comparator:(a,b)=>a.w-b.w});
+  q.queue({n:s,w:0});
+  while(q.length){
+    const {n,w}=q.dequeue();
+    if(n===e)return w;
+    if(w>d[n])continue;
+    for(const t in g[n]){
+      const w2=w+g[n][t];
+      if(w2<d[t]){
+        d[t]=w2;
+        q.queue({n:t,w:w2});
+      }
+    }
+  }
+  return Infinity
+}
+export default findShortestPath;
+// Generation time: 45.490s
+// Result: PASS

tests/1_dijkstra/outputs/google_gemini-3.1-flash-lite.js

@@ -0,0 +1,29 @@
+async function findShortestPath(graph, start, end) {
+  const { PriorityQueue } = await import('https://cdn.skypack.dev/js-priority-queue');
+  const dist = { [start]: 0 };
+  const pq = new PriorityQueue({
+    comparator: (a, b) => a.w - b.w
+  });
+
+  pq.queue({ n: start, w: 0 });
+
+  while (pq.length) {
+    const { n, w } = pq.dequeue();
+
+    if (n === end) return w;
+    if (w > (dist[n] ?? Infinity)) continue;
+
+    for (const [neighbor, weight] of Object.entries(graph[n] || {})) {
+      const newDist = w + weight;
+      if (newDist < (dist[neighbor] ?? Infinity)) {
+        dist[neighbor] = newDist;
+        pq.queue({ n: neighbor, w: newDist });
+      }
+    }
+  }
+
+  return Infinity;
+}
+export default findShortestPath;
+// Generation time: 1.539s
+// Result: FAIL

tests/1_dijkstra/outputs/moonshotai_kimi-k2.6.js

@@ -0,0 +1,22 @@
+const findShortestPath = async (g, s, e) => {
+  const { default: P } = await import('https://esm.sh/js-priority-queue');
+  const q = new P({ comparator: (a, b) => a.d - b.d });
+  const dist = { [s]: 0 };
+  q.queue({ u: s, d: 0 });
+  while (q.length) {
+    const { u, d: du } = q.dequeue();
+    if (du > dist[u]) continue;
+    if (u === e) return du;
+    for (const [v, w] of Object.entries(g[u] ?? {})) {
+      const alt = du + w;
+      if (alt < (dist[v] ?? Infinity)) {
+        dist[v] = alt;
+        q.queue({ u: v, d: alt });
+      }
+    }
+  }
+  return dist[e] ?? Infinity;
+};
+export default findShortestPath;
+// Generation time: 152.935s
+// Result: PASS

tests/1_dijkstra/outputs/openai_gpt-5.5 EFF_high.js

@@ -0,0 +1,43 @@
+const loadPQ = (() => {
+  let p;
+
+  return () => p ||= import("https://esm.sh/js-priority-queue@0.1.5")
+    .then(m => m.default?.default ?? m.default ?? m.PriorityQueue);
+})();
+
+async function findShortestPath(graph, start, end) {
+  const PriorityQueue = await loadPQ();
+  const dist = new Map([[start, 0]]);
+  const q = new PriorityQueue({
+    comparator: (a, b) => a.cost - b.cost
+  });
+
+  q.queue({ node: start, cost: 0 });
+
+  while (q.length) {
+    const { node, cost } = q.dequeue();
+
+    if (cost > dist.get(node)) continue;
+    if (node === end) return cost;
+
+    for (const [next, rawWeight] of Object.entries(graph?.[node] ?? {})) {
+      const weight = Number(rawWeight);
+
+      if (!Number.isFinite(weight) || weight < 0) {
+        throw new RangeError("Edge weights must be finite non-negative numbers");
+      }
+
+      const nextCost = cost + weight;
+
+      if (nextCost < (dist.get(next) ?? Infinity)) {
+        dist.set(next, nextCost);
+        q.queue({ node: next, cost: nextCost });
+      }
+    }
+  }
+
+  return Infinity;
+}
+export default findShortestPath;
+// Generation time: 55.807s
+// Result: PASS

tests/1_dijkstra/outputs/z-ai_glm-5.1.js

@@ -0,0 +1,22 @@
+async function findShortestPath(g, s, e) {
+  const { default: PQ } = await import('https://cdn.jsdelivr.net/npm/js-priority-queue/+esm');
+  const q = new PQ({ comparator: (a, b) => a[1] - b[1] });
+  const d = { [s]: 0 };
+  q.queue([s, 0]);
+  while (q.length) {
+    const [n, c] = q.dequeue();
+    if (n === e) return c;
+    if (c > (d[n] ?? 1/0)) continue;
+    for (const nb in g[n] || {}) {
+      const nd = c + g[n][nb];
+      if (nd < (d[nb] ?? 1/0)) {
+        d[nb] = nd;
+        q.queue([nb, nd]);
+      }
+    }
+  }
+  return d[e] ?? 1/0;
+}
+export default findShortestPath;
+// Generation time: 39.949s
+// Result: PASS

tests/2_convex_hull/outputs/anthropic_claude-opus-4.7 EFF_medium.js

@@ -0,0 +1,20 @@
+async function findConvexHull(points) {
+    const _ = (await import('https://cdn.skypack.dev/lodash')).default;
+    const pts = _.sortBy(_.uniqWith(points, (a, b) => a.x === b.x && a.y === b.y), ['x', 'y']);
+    if (pts.length < 2) return pts.slice();
+    const cross = (O, A, B) => (A.x - O.x) * (B.y - O.y) - (A.y - O.y) * (B.x - O.x);
+    const build = arr => {
+        const h = [];
+        for (const p of arr) {
+            while (h.length >= 2 && cross(h[h.length - 2], h[h.length - 1], p) <= 0) h.pop();
+            h.push(p);
+        }
+        return h;
+    };
+    const lower = build(pts);
+    const upper = build([...pts].reverse());
+    return [...lower.slice(0, -1), ...upper.slice(0, -1)];
+}
+export default findConvexHull;
+// Generation time: 4.186s
+// Result: PASS

tests/2_convex_hull/outputs/deepseek_deepseek-v4-pro.js

@@ -0,0 +1,25 @@
+const findConvexHull = async (points) => {
+  const _ = await import('https://cdn.jsdelivr.net/npm/lodash@4.17.21/lodash.min.js');
+  const pts = _.sortBy(_.uniqWith(points, (a, b) => a.x === b.x && a.y === b.y), ['x', 'y']);
+  const n = pts.length;
+  if (n <= 2) return pts;
+  const cross = (o, a, b) => (a.x - o.x) * (b.y - o.y) - (a.y - o.y) * (b.x - o.x);
+  const lower = [];
+  for (let i = 0; i < n; i++) {
+    while (lower.length >= 2 && cross(lower[lower.length - 2], lower[lower.length - 1], pts[i]) <= 0)
+      lower.pop();
+    lower.push(pts[i]);
+  }
+  const upper = [];
+  for (let i = n - 1; i >= 0; i--) {
+    while (upper.length >= 2 && cross(upper[upper.length - 2], upper[upper.length - 1], pts[i]) <= 0)
+      upper.pop();
+    upper.push(pts[i]);
+  }
+  lower.pop();
+  upper.pop();
+  return lower.concat(upper);
+};
+export default findConvexHull;
+// Generation time: 21.237s
+// Result: FAIL

tests/2_convex_hull/outputs/google_gemini-3.1-flash-lite.js

@@ -0,0 +1,30 @@
+async function findConvexHull(pts) {
+  const { sortBy, uniqWith, isEqual } = await import('https://cdn.skypack.dev/lodash');
+  
+  const sorted = sortBy(uniqWith(pts, isEqual), ['x', 'y']);
+  if (sorted.length <= 2) return sorted;
+
+  const crossProduct = (a, b, c) => (b.x - a.x) * (c.y - a.y) - (b.y - a.y) * (c.x - a.x);
+
+  const build = (arr) => {
+    const hull = [];
+    for (const p of arr) {
+      while (hull.length >= 2 && crossProduct(hull[hull.length - 2], hull[hull.length - 1], p) <= 0) {
+        hull.pop();
+      }
+      hull.push(p);
+    }
+    return hull;
+  };
+
+  const lower = build(sorted);
+  const upper = build([...sorted].reverse());
+
+  lower.pop();
+  upper.pop();
+  
+  return lower.concat(upper);
+}
+export default findConvexHull;
+// Generation time: 1.515s
+// Result: PASS

tests/2_convex_hull/outputs/moonshotai_kimi-k2.6.js

@@ -0,0 +1,24 @@
+const findConvexHull = async (pts) => {
+  const { sortBy, uniqBy } = await import("https://esm.sh/lodash-es@4.17.21"),
+    u = uniqBy(pts, p => `${p.x},${p.y}`),
+    s = sortBy(u, ["x", "y"]),
+    n = s.length,
+    c = (a, b, d) => (b.x - a.x) * (d.y - a.y) - (b.y - a.y) * (d.x - a.x),
+    l = [],
+    h = [];
+  if (n < 3) return s;
+  for (let i = 0; i < n; i++) {
+    const p = s[i];
+    while (l.length > 1 && c(l[l.length - 2], l[l.length - 1], p) <= 0) l.pop();
+    l.push(p);
+  }
+  for (let i = n; i--; ) {
+    const p = s[i];
+    while (h.length > 1 && c(h[h.length - 2], h[h.length - 1], p) <= 0) h.pop();
+    h.push(p);
+  }
+  return [...l.slice(0, -1), ...h.slice(0, -1)];
+};
+export default findConvexHull;
+// Generation time: 116.732s
+// Result: PASS

tests/2_convex_hull/outputs/openai_gpt-5.5 EFF_high.js

@@ -0,0 +1,34 @@
+const findConvexHull = async points => {
+  const { sortBy, uniqBy } = await import(
+    "https://cdn.jsdelivr.net/npm/lodash-es@4.17.21/lodash.js"
+  )
+
+  const pts = uniqBy(
+    sortBy(points, ["x", "y"]),
+    p => `${p.x}\u0000${p.y}`
+  )
+
+  if (pts.length < 2) return pts
+
+  const cross = (a, b, c) =>
+    (b.x - a.x) * (c.y - a.y) - (b.y - a.y) * (c.x - a.x)
+
+  const half = arr => {
+    const h = []
+
+    for (const p of arr) {
+      while (h.length > 1 && cross(h.at(-2), h.at(-1), p) <= 0) h.pop()
+      h.push(p)
+    }
+
+    return h
+  }
+
+  return [
+    ...half(pts).slice(0, -1),
+    ...half([...pts].reverse()).slice(0, -1)
+  ]
+}
+export default findConvexHull;
+// Generation time: 14.673s
+// Result: PASS

tests/2_convex_hull/outputs/z-ai_glm-5.1.js

@@ -0,0 +1,25 @@
+const findConvexHull = async (pts) => {
+  const _ = await import("https://esm.run/lodash");
+  const p = _.uniqBy(_.sortBy(pts, ["x", "y"]), p => p.x + "," + p.y);
+  if (p.length < 3) return p;
+
+  const c = (o, a, b) => (a.x - o.x) * (b.y - o.y) - (a.y - o.y) * (b.x - o.x);
+  const h = [];
+
+  for (const pt of p) {
+    while (h.length >= 2 && c(h.at(-2), h.at(-1), pt) <= 0) h.pop();
+    h.push(pt);
+  }
+
+  const l = h.length + 1;
+  for (let i = p.length - 2; i >= 0; i--) {
+    while (h.length >= l && c(h.at(-2), h.at(-1), p[i]) <= 0) h.pop();
+    h.push(p[i]);
+  }
+
+  h.pop();
+  return h;
+};
+export default findConvexHull;
+// Generation time: 35.612s
+// Result: FAIL

tests/3_signal_pipeline/outputs/anthropic_claude-opus-4.7 EFF_medium.js

@@ -0,0 +1,46 @@
+async function analyzeSignal(yamlStr) {
+  const [yaml, math, ndarray, fft, DOMPurify] = (await Promise.all([
+    import('https://esm.sh/js-yaml'),
+    import('https://esm.sh/mathjs'),
+    import('https://esm.sh/ndarray'),
+    import('https://esm.sh/ndarray-fft'),
+    import('https://esm.sh/dompurify')
+  ])).map(m => m.default ?? m);
+
+  const cfg = yaml.load(yamlStr);
+  const N = cfg.sampleRate * cfg.duration;
+  const signal = Array.from({ length: N }, (_, i) => {
+    const t = i / cfg.sampleRate;
+    return cfg.components.reduce(
+      (s, c) => s + c.amplitude * math.sin(2 * math.pi * c.frequency * t),
+      0
+    );
+  });
+
+  const real = ndarray(new Float64Array(signal), [N]);
+  const imag = ndarray(new Float64Array(N), [N]);
+  fft(1, real, imag);
+
+  const half = N / 2;
+  const peaks = [];
+  for (let k = 0; k <= half; k++) {
+    const mag = math.sqrt(real.get(k) ** 2 + imag.get(k) ** 2) / half;
+    if (mag > 0.1) {
+      peaks.push({
+        frequencyHz: Math.round(k * cfg.sampleRate / N),
+        magnitude: Math.round(mag * 100) / 100
+      });
+    }
+  }
+  peaks.sort((a, b) => b.magnitude - a.magnitude);
+
+  const rows = peaks.map(p => `<tr><td>${p.frequencyHz}</td><td>${p.magnitude}</td></tr>`).join('');
+  const html = DOMPurify.sanitize(
+    `<table><tr><th>Frequency (Hz)</th><th>Magnitude</th></tr>${rows}</table>`
+  );
+
+  return { peaks, html, signalLength: N };
+}
+export default analyzeSignal;
+// Generation time: 9.148s
+// Result: PASS

tests/3_signal_pipeline/outputs/deepseek_deepseek-v4-pro.js

@@ -0,0 +1,42 @@
+async function analyzeSignal(y) {
+  const [Y, M, Nd, Fft, P] = await Promise.all([
+    import('https://esm.run/js-yaml@4.1.0').then(m => m.default),
+    import('https://esm.run/mathjs@11.8.0').then(m => m.default),
+    import('https://esm.run/ndarray@1.0.19').then(m => m.default),
+    import('https://esm.run/ndarray-fft@1.0.3').then(m => m.default),
+    import('https://esm.run/dompurify@3.0.5').then(m => m.default)
+  ]);
+  const cfg = Y.load(y),
+    sr = cfg.sampleRate,
+    d = cfg.duration,
+    comps = cfg.components,
+    N = sr * d,
+    sig = Array.from({ length: N }, (_, i) => {
+      const t = i / sr;
+      return comps.reduce((s, c) => s + c.amplitude * M.sin(2 * M.pi * c.frequency * t), 0);
+    }),
+    real = Nd(new Float64Array(sig), [N]),
+    imag = Nd(new Float64Array(N), [N]);
+  Fft(1, real, imag);
+  const half = N / 2,
+    peaks = [];
+  for (let k = 0; k <= half; k++) {
+    const re = real.get(k),
+      im = imag.get(k),
+      mag = M.sqrt(re * re + im * im) / half;
+    if (mag > 0.1) peaks.push({ f: Math.round(k * sr / N), m: Math.round(mag * 1e2) / 1e2 });
+  }
+  peaks.sort((a, b) => b.m - a.m);
+  let h = '<table><tr><th>Frequency (Hz)</th><th>Magnitude</th></tr>';
+  for (const p of peaks) h += `<tr><td>${p.f}</td><td>${p.m}</td></tr>`;
+  h += '</table>';
+  const sH = P.sanitize(h);
+  return {
+    peaks: peaks.map(p => ({ frequencyHz: p.f, magnitude: p.m })),
+    html: sH,
+    signalLength: N
+  };
+}
+export default analyzeSignal;
+// Generation time: 102.963s
+// Result: FAIL

tests/3_signal_pipeline/outputs/google_gemini-3.1-flash-lite.js

@@ -0,0 +1,40 @@
+async function analyzeSignal(yamlStr) {
+  const [yaml, math, { default: ndarray }, fft, { sanitize }] = await Promise.all([
+    import('https://esm.sh/js-yaml'),
+    import('https://esm.sh/mathjs'),
+    import('https://esm.sh/ndarray'),
+    import('https://esm.sh/ndarray-fft'),
+    import('https://esm.sh/dompurify')
+  ]);
+
+  const cfg = yaml.load(yamlStr);
+  const N = cfg.sampleRate * cfg.duration;
+  const real = ndarray(new Float64Array(N));
+  const imag = ndarray(new Float64Array(N));
+
+  for (let i = 0; i < N; i++) {
+    const t = i / cfg.sampleRate;
+    const v = cfg.components.reduce((acc, c) => 
+      acc + c.amplitude * math.sin(2 * math.pi * c.frequency * t), 0);
+    real.set(i, v);
+  }
+
+  fft(1, real, imag);
+
+  const peaks = [];
+  for (let k = 0; k <= N / 2; k++) {
+    const mag = math.sqrt(real.get(k) ** 2 + imag.get(k) ** 2) / (N / 2);
+    if (mag > 0.1) peaks.push({ frequencyHz: Math.round(k * cfg.sampleRate / N), magnitude: +mag.toFixed(2) });
+  }
+
+  peaks.sort((a, b) => b.magnitude - a.magnitude);
+
+  let html = '<table><tr><th>Frequency (Hz)</th><th>Magnitude</th></tr>';
+  peaks.forEach(p => html += `<tr><td>${p.frequencyHz}</td><td>${p.magnitude}</td></tr>`);
+  html += '</table>';
+
+  return { peaks, html: sanitize(html), signalLength: N };
+}
+export default analyzeSignal;
+// Generation time: 1.769s
+// Result: FAIL

tests/3_signal_pipeline/outputs/moonshotai_kimi-k2.6.js

@@ -0,0 +1,459 @@
+```html
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>Signal Processing Pipeline</title>
+    <script src="https://cdn.tailwindcss.com"></script>
+    <link href="https://fonts.googleapis.com/css2?family=JetBrains+Mono:wght@400;700&family=Inter:wght@400;600;800&display=swap" rel="stylesheet">
+    <style>
+        body {
+            font-family: 'Inter', sans-serif;
+            background-color: #0f172a; /* Slate 900 */
+            color: #e2e8f0; /* Slate 200 */
+        }
+        .mono {
+            font-family: 'JetBrains Mono', monospace;
+        }
+        .glass-panel {
+            background: rgba(30, 41, 59, 0.7);
+            backdrop-filter: blur(12px);
+            border: 1px solid rgba(148, 163, 184, 0.1);
+            box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1), 0 2px 4px -1px rgba(0, 0, 0, 0.06);
+        }
+        .input-focus:focus {
+            outline: none;
+            border-color: #38bdf8;
+            box-shadow: 0 0 0 2px rgba(56, 189, 248, 0.2);
+        }
+        /* Custom Scrollbar */
+        ::-webkit-scrollbar {
+            width: 8px;
+            height: 8px;
+        }
+        ::-webkit-scrollbar-track {
+            background: #1e293b; 
+        }
+        ::-webkit-scrollbar-thumb {
+            background: #475569; 
+            border-radius: 4px;
+        }
+        ::-webkit-scrollbar-thumb:hover {
+            background: #64748b; 
+        }
+        .canvas-container {
+            position: relative;
+            height: 250px;
+            width: 100%;
+            background: #020617;
+            border-radius: 0.5rem;
+            overflow: hidden;
+            border: 1px solid #1e293b;
+        }
+        canvas {
+            display: block;
+            width: 100%;
+            height: 100%;
+        }
+        .loader {
+            border: 3px solid rgba(255,255,255,0.1);
+            border-radius: 50%;
+            border-top: 3px solid #38bdf8;
+            width: 24px;
+            height: 24px;
+            -webkit-animation: spin 1s linear infinite; /* Safari */
+            animation: spin 1s linear infinite;
+        }
+        @keyframes spin {
+            0% { transform: rotate(0deg); }
+            100% { transform: rotate(360deg); }
+        }
+    </style>
+</head>
+<body class="min-h-screen flex flex-col">
+
+    <!-- Header -->
+    <header class="border-b border-slate-800 bg-slate-900/80 sticky top-0 z-50 backdrop-blur-md">
+        <div class="max-w-7xl mx-auto px-4 sm:px-6 lg:px-8 h-16 flex items-center justify-between">
+            <div class="flex items-center gap-3">
+                <div class="w-8 h-8 bg-cyan-500 rounded flex items-center justify-center shadow-[0_0_15px_rgba(6,182,212,0.5)]">
+                    <svg class="w-5 h-5 text-white" fill="none" stroke="currentColor" viewBox="0 0 24 24"><path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M9 19v-6a2 2 0 00-2-2H5a2 2 0 00-2 2v6a2 2 0 002 2h2a2 2 0 002-2zm0 0V9a2 2 0 012-2h2a2 2 0 012 2v10m-6 0a2 2 0 002 2h2a2 2 0 002-2m0 0V5a2 2 0 012-2h2a2 2 0 012 2v14a2 2 0 01-2 2h-2a2 2 0 01-2-2z"></path></svg>
+                </div>
+                <h1 class="text-xl font-bold tracking-tight text-white">Signal<span class="text-cyan-400">Forge</span> <span class="text-xs font-normal text-slate-500 ml-2 border border-slate-700 px-2 py-0.5 rounded-full">FFT Processor</span></h1>
+            </div>
+            <div class="text-xs text-slate-400 font-mono hidden sm:block">Dynamic Import Pipeline: js-yaml &rarr; mathjs &rarr; ndarray-fft &rarr; DOMPurify</div>
+        </div>
+    </header>
+
+    <!-- Main Content -->
+    <main class="flex-grow p-4 sm:p-6 lg:p-8 max-w-7xl mx-auto w-full grid grid-cols-1 lg:grid-cols-12 gap-6">
+        
+        <!-- Left Column: Controls -->
+        <div class="lg:col-span-4 flex flex-col gap-6">
+            
+            <!-- Input Panel -->
+            <div class="glass-panel rounded-xl p-6 flex flex-col gap-4">
+                <div class="flex justify-between items-center">
+                    <h2 class="text-lg font-semibold text-white flex items-center gap-2">
+                        <svg class="w-4 h-4 text-cyan-400" fill="none" stroke="currentColor" viewBox="0 0 24 24"><path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M10 20l4-16m4 4l4 4-4 4M6 16l-4-4 4-4"></path></svg>
+                        Signal Config (YAML)
+                    </h2>
+                    <button id="resetBtn" class="text-xs text-slate-400 hover:text-white transition-colors">Reset Default</button>
+                </div>
+                
+                <textarea id="yamlInput" class="mono text-xs bg-slate-950 text-emerald-400 p-4 rounded-lg border border-slate-800 h-64 resize-none input-focus leading-relaxed" spellcheck="false"></textarea>
+                
+                <div class="flex gap-3 mt-2">
+                    <button id="runBtn" class="flex-1 bg-cyan-600 hover:bg-cyan-500 text-white font-semibold py-2 px-4 rounded-lg transition-all shadow-lg shadow-cyan-500/20 flex items-center justify-center gap-2 active:scale-95">
+                        <span>Process Signal</span>
+                        <div id="btnLoader" class="loader hidden"></div>
+                    </button>
+                </div>
+                
+                <div id="errorContainer" class="hidden bg-red-500/10 border border-red-500/30 text-red-400 text-xs p-3 rounded-lg">
+                    <span class="font-bold">Error:</span> <span id="errorMsg"></span>
+                </div>
+            </div>
+
+            <!-- Info Panel -->
+            <div class="glass-panel rounded-xl p-6">
+                <h3 class="text-sm font-semibold text-slate-300 mb-3">Pipeline Status</h3>
+                <ul class="space-y-3 text-xs text-slate-400">
+                    <li class="flex items-center gap-2" id="status-yaml">
+                        <span class="w-2 h-2 rounded-full bg-slate-600"></span> js-yaml (Parser)
+                    </li>
+                    <li class="flex items-center gap-2" id="status-math">
+                        <span class="w-2 h-2 rounded-full bg-slate-600"></span> mathjs (Synthesis)
+                    </li>
+                    <li class="flex items-center gap-2" id="status-fft">
+                        <span class="w-2 h-2 rounded-full bg-slate-600"></span> ndarray-fft (Analysis)
+                    </li>
+                    <li class="flex items-center gap-2" id="status-dom">
+                        <span class="w-2 h-2 rounded-full bg-slate-600"></span> DOMPurify (Sanitization)
+                    </li>
+                </ul>
+            </div>
+        </div>
+
+        <!-- Right Column: Visualization & Results -->
+        <div class="lg:col-span-8 flex flex-col gap-6">
+            
+            <!-- Charts -->
+            <div class="grid grid-cols-1 md:grid-cols-2 gap-6">
+                <!-- Time Domain -->
+                <div class="glass-panel p-4 rounded-xl">
+                    <div class="flex justify-between items-end mb-3">
+                        <h3 class="text-sm font-semibold text-slate-300">Time Domain</h3>
+                        <span class="text-xs text-slate-500 mono">signal[i]</span>
+                    </div>
+                    <div class="canvas-container">
+                        <canvas id="timeCanvas"></canvas>
+                    </div>
+                </div>
+
+                <!-- Frequency Domain -->
+                <div class="glass-panel p-4 rounded-xl">
+                    <div class="flex justify-between items-end mb-3">
+                        <h3 class="text-sm font-semibold text-slate-300">Frequency Domain</h3>
+                        <span class="text-xs text-slate-500 mono">|X(k)|</span>
+                    </div>
+                    <div class="canvas-container">
+                        <canvas id="freqCanvas"></canvas>
+                    </div>
+                </div>
+            </div>
+
+            <!-- Results Table -->
+            <div class="glass-panel rounded-xl overflow-hidden flex flex-col h-full">
+                <div class="p-4 border-b border-slate-800 bg-slate-800/50 flex justify-between items-center">
+                    <h3 class="text-sm font-semibold text-white">Dominant Frequencies</h3>
+                    <span class="text-xs text-slate-500" id="signalLengthDisplay">N = 0</span>
+                </div>
+                <div class="overflow-x-auto">
+                    <div id="htmlOutput" class="p-0">
+                        <!-- Sanitized HTML Table will be injected here -->
+                        <div class="p-8 text-center text-slate-500 text-sm italic">Run the pipeline to see results...</div>
+                    </div>
+                </div>
+            </div>
+
+        </div>
+    </main>
+
+    <script>
+        // --- Configuration & State ---
+        const defaultYaml = `sampleRate: 64
+duration: 1
+components:
+  - frequency: 5
+    amplitude: 1.0
+  - frequency: 12
+    amplitude: 0.5`;
+
+        const yamlInput = document.getElementById('yamlInput');
+        const runBtn = document.getElementById('runBtn');
+        const resetBtn = document.getElementById('resetBtn');
+        const btnLoader = document.getElementById('btnLoader');
+        const errorContainer = document.getElementById('errorContainer');
+        const errorMsg = document.getElementById('errorMsg');
+        const htmlOutput = document.getElementById('htmlOutput');
+        const signalLengthDisplay = document.getElementById('signalLengthDisplay');
+        
+        const timeCanvas = document.getElementById('timeCanvas');
+        const freqCanvas = document.getElementById('freqCanvas');
+        const timeCtx = timeCanvas.getContext('2d');
+        const freqCtx = freqCanvas.getContext('2d');
+
+        // Status Indicators
+        const statusEls = {
+            yaml: document.getElementById('status-yaml').firstElementChild,
+            math: document.getElementById('status-math').firstElementChild,
+            fft: document.getElementById('status-fft').firstElementChild,
+            dom: document.getElementById('status-dom').firstElementChild,
+        };
+
+        yamlInput.value = defaultYaml;
+
+        // --- Visualization Helpers ---
+        function resizeCanvas(canvas) {
+            const parent = canvas.parentElement;
+            canvas.width = parent.clientWidth;
+            canvas.height = parent.clientHeight;
+        }
+
+        function drawTimeDomain(signal) {
+            resizeCanvas(timeCanvas);
+            const w = timeCanvas.width;
+            const h = timeCanvas.height;
+            timeCtx.clearRect(0, 0, w, h);
+            
+            // Grid
+            timeCtx.strokeStyle = '#1e293b';
+            timeCtx.lineWidth = 1;
+            timeCtx.beginPath();
+            timeCtx.moveTo(0, h/2);
+            timeCtx.lineTo(w, h/2);
+            timeCtx.stroke();
+
+            // Signal
+            timeCtx.strokeStyle = '#38bdf8'; // Cyan 400
+            timeCtx.lineWidth = 2;
+            timeCtx.beginPath();
+            
+            const step = w / signal.length;
+            const amp = h / 2.5; // Scale factor
+            
+            for(let i=0; i<signal.length; i++) {
+                const x = i * step;
+                const y = (h/2) - (signal[i] * amp);
+                if(i===0) timeCtx.moveTo(x, y);
+                else timeCtx.lineTo(x, y);
+            }
+            timeCtx.stroke();
+        }
+
+        function drawFrequencyDomain(magnitude) {
+            resizeCanvas(freqCanvas);
+            const w = freqCanvas.width;
+            const h = freqCanvas.height;
+            freqCtx.clearRect(0, 0, w, h);
+
+            // Grid
+            freqCtx.strokeStyle = '#1e293b';
+            freqCtx.lineWidth = 1;
+            freqCtx.beginPath();
+            freqCtx.moveTo(0, h);
+            freqCtx.lineTo(w, h);
+            freqCtx.stroke();
+
+            // Bars
+            const barWidth = (w / magnitude.length) * 0.8;
+            const gap = (w / magnitude.length) * 0.2;
+            
+            for(let i=0; i<magnitude.length; i++) {
+                const val = magnitude[i];
+                const barH = Math.min(h, val * (h * 0.9)); // Scale to fit
+                const x = i * (barWidth + gap);
+                const y = h - barH;
+
+                // Color based on magnitude intensity
+                const intensity = Math.min(255, Math.floor(val * 500));
+                freqCtx.fillStyle = `rgb(${56 + intensity/2}, ${189}, ${248})`; // Base cyan + glow
+                
+                // If peak is significant, highlight
+                if(val > 0.1) {
+                    freqCtx.fillStyle = '#22d3ee'; // Cyan 400
+                    freqCtx.shadowColor = '#06b6d4';
+                    freqCtx.shadowBlur = 10;
+                } else {
+                    freqCtx.shadowBlur = 0;
+                }
+
+                freqCtx.fillRect(x, y, barWidth, barH);
+            }
+            freqCtx.shadowBlur = 0; // Reset
+        }
+
+        // --- Core Logic (The Production-Ready Async Function) ---
+        
+        async function analyzeSignal(yamlString) {
+            // 1. Dynamic Imports (Simulating CDN loads with skypack/polyfill for robustness in this environment)
+            // Note: In a real build, these would be bare imports handled by bundler, 
+            // or full URLs if left as dynamic strings.
+            // To ensure stability in this specific "No-Build" output, we use esm.sh which is highly reliable for raw HTML.
+            
+            const setStatus = (key, active) => {
+                statusEls[key].className = `w-2 h-2 rounded-full ${active ? 'bg-cyan-400 shadow-[0_0_8px_rgba(34,211,238,0.8)]' : 'bg-slate-600'}`;
+            };
+
+            // Reset status
+            Object.keys(statusEls).forEach(k => setStatus(k, false));
+
+            // Import Step
+            setStatus('yaml', true);
+            const yaml = await import('https://cdn.skypack.dev/js-yaml');
+            setStatus('yaml', false);
+
+            setStatus('math', true);
+            const math = await import('https://cdn.skypack.dev/mathjs');
+            setStatus('math', false);
+
+            setStatus('fft', true);
+            const ndarray = (await import('https://cdn.skypack.dev/ndarray')).default;
+            const fft = (await import('https://cdn.skypack.dev/ndarray-fft')).default;
+            setStatus('fft', false);
+
+            setStatus('dom', true);
+            const DOMPurify = (await import('https://cdn.skypack.dev/dompurify')).default;
+            setStatus('dom', false);
+
+            // 2. Parse YAML
+            const config = yaml.load(yamlString);
+            
+            const N = config.sampleRate * config.duration;
+            const signal = new Array(N).fill(0);
+
+            // 3. Generate Signal
+            for (let i = 0; i < N; i++) {
+                const t = i / config.sampleRate;
+                let sum = 0;
+                for (const comp of config.components) {
+                    sum += comp.amplitude * math.sin(2 * math.pi * comp.frequency * t);
+                }
+                signal[i] = sum;
+            }
+
+            // 4. Prepare FFT
+            const real = ndarray(new Float64Array(signal));
+            const imag = ndarray(new Float64Array(N).fill(0));
+
+            // 5. Execute FFT
+            fft(1, real, imag);
+
+            // 6. Compute Magnitude Spectrum
+            const halfN = N / 2;
+            const magnitude = new Array(halfN + 1);
+            for (let k = 0; k <= halfN; k++) {
+                const r = real.get(k);
+                const im = imag.get(k);
+                magnitude[k] = math.sqrt(r * r + im * im) / halfN;
+            }
+
+            // 7. Find Dominant Frequencies
+            const peaks = [];
+            for (let k = 0; k <= halfN; k++) {
+                if (magnitude[k] > 0.1) {
+                    const freqHz = Math.round(k * config.sampleRate / N);
+                    const magVal = Math.round(magnitude[k] * 100) / 100;
+                    peaks.push({ frequencyHz: freqHz, magnitude: magVal });
+                }
+            }
+            
+            // Sort by magnitude descending
+            peaks.sort((a, b) => b.magnitude - a.magnitude);
+
+            // 8. Build HTML Table
+            let html = `<table class="w-full text-sm text-left text-slate-300"><thead class="text-xs text-slate-400 uppercase bg-slate-800/50"><tr><th scope="col" class="px-6 py-3">Frequency (Hz)</th><th scope="col" class="px-6 py-3">Magnitude</th></tr></thead><tbody>`;
+            
+            if (peaks.length === 0) {
+                 html += `<tr class="bg-slate-800/20 border-b border-slate-800"><td colspan="2" class="px-6 py-4 text-center italic text-slate-500">No dominant frequencies detected (>0.1)</td></tr>`;
+            } else {
+                peaks.forEach((p, idx) => {
+                    const rowClass = idx % 2 === 0 ? 'bg-slate-800/20' : 'bg-transparent';
+                    html += `<tr class="${rowClass} border-b border-slate-800 hover:bg-slate-700/30 transition-colors"><td class="px-6 py-4 font-medium text-white">${p.frequencyHz}</td><td class="px-6 py-4 text-cyan-400 font-mono">${p.magnitude}</td></tr>`;
+                });
+            }
+            
+            html += `</tbody></table>`;
+
+            // 9. Sanitize
+            const sanitizedHTML = DOMPurify.sanitize(html, { 
+                ALLOWED_TAGS: ['table', 'thead', 'tbody', 'tr', 'th', 'td'],
+                ALLOWED_ATTR: ['class', 'scope', 'colspan']
+            });
+
+            return {
+                peaks,
+                html: sanitizedHTML,
+                signalLength: N,
+                rawSignal: signal,
+                rawMagnitude: magnitude
+            };
+        }
+
+        // --- UI Interaction ---
+
+        runBtn.addEventListener('click', async () => {
+            // UI Loading State
+            runBtn.disabled = true;
+            btnLoader.classList.remove('hidden');
+            errorContainer.classList.add('hidden');
+            htmlOutput.innerHTML = '<div class="p-8 flex justify-center"><div class="loader"></div></div>';
+            
+            // Clear canvases
+            timeCtx.clearRect(0,0,timeCanvas.width, timeCanvas.height);
+            freqCtx.clearRect(0,0,freqCanvas.width, freqCanvas.height);
+
+            try {
+                const result = await analyzeSignal(yamlInput.value);
+                
+                // Update UI
+                htmlOutput.innerHTML = result.html;
+                signalLengthDisplay.textContent = `N = ${result.signalLength}`;
+                
+                // Draw Charts
+                drawTimeDomain(result.rawSignal);
+                drawFrequencyDomain(result.rawMagnitude);
+
+            } catch (err) {
+                console.error(err);
+                errorContainer.classList.remove('hidden');
+                errorMsg.textContent = err.message;
+                htmlOutput.innerHTML = '<div class="p-8 text-center text-slate-500 text-sm italic">Error processing signal.</div>';
+            } finally {
+                runBtn.disabled = false;
+                btnLoader.classList.add('hidden');
+                // Reset status dots
+                Object.keys(statusEls).forEach(k => statusEls[k].className = 'w-2 h-2 rounded-full bg-slate-600');
+            }
+        });
+
+        resetBtn.addEventListener('click', () => {
+            yamlInput.value = defaultYaml;
+        });
+
+        // Initial resize
+        window.addEventListener('resize', () => {
+            // Re-draw if we have data? For now just resize canvases
+            resizeCanvas(timeCanvas);
+            resizeCanvas(freqCanvas);
+        });
+
+    </script>
+</body>
+</html>
+```
+export default analyzeSignal;
+// Generation time: 141.385s
+// Result: FAIL

tests/3_signal_pipeline/outputs/openai_gpt-5.5 EFF_high.js

@@ -0,0 +1,76 @@
+async function analyzeSignal(yamlString) {
+  const [
+    yamlMod,
+    math,
+    ndarrayMod,
+    fftMod,
+    purifyMod
+  ] = await Promise.all([
+    import("https://cdn.jsdelivr.net/npm/js-yaml@4.1.0/+esm"),
+    import("https://cdn.jsdelivr.net/npm/mathjs@13.2.0/+esm"),
+    import("https://cdn.jsdelivr.net/npm/ndarray@1.0.19/+esm"),
+    import("https://cdn.jsdelivr.net/npm/ndarray-fft@1.0.3/+esm"),
+    import("https://cdn.jsdelivr.net/npm/dompurify@3.1.6/+esm")
+  ]);
+
+  const yaml = yamlMod.default || yamlMod;
+  const ndarray = ndarrayMod.default || ndarrayMod;
+  const fft = fftMod.default || fftMod;
+  const DOMPurify = purifyMod.default || purifyMod;
+
+  const cfg = yaml.load(yamlString);
+  const sampleRate = Number(cfg.sampleRate);
+  const duration = Number(cfg.duration);
+  const components = Array.isArray(cfg.components) ? cfg.components : [];
+  const N = Math.max(0, Math.floor(sampleRate * duration));
+
+  if (!Number.isFinite(sampleRate) || sampleRate <= 0) throw new Error("Invalid sampleRate");
+  if (!Number.isFinite(duration) || duration <= 0) throw new Error("Invalid duration");
+  if (!N) throw new Error("Signal length must be greater than zero");
+
+  const signal = Float64Array.from({ length: N }, (_, i) => {
+    const t = i / sampleRate;
+
+    return components.reduce((sum, c) => {
+      const a = Number(c.amplitude);
+      const f = Number(c.frequency);
+
+      return Number.isFinite(a) && Number.isFinite(f)
+        ? sum + a * math.sin(2 * math.pi * f * t)
+        : sum;
+    }, 0);
+  });
+
+  const real = ndarray(signal.slice(), [N]);
+  const imag = ndarray(new Float64Array(N), [N]);
+
+  fft(1, real, imag);
+
+  const peaks = Array.from({ length: Math.floor(N / 2) + 1 }, (_, k) => {
+    const re = real.get(k);
+    const im = imag.get(k);
+    const magnitude = math.sqrt(re ** 2 + im ** 2) / (N / 2);
+
+    return {
+      frequencyHz: Math.round(k * sampleRate / N),
+      magnitude: Math.round(magnitude * 100) / 100,
+      rawMagnitude: magnitude
+    };
+  })
+    .filter(p => p.rawMagnitude > 0.1)
+    .sort((a, b) => b.rawMagnitude - a.rawMagnitude)
+    .map(({ frequencyHz, magnitude }) => ({ frequencyHz, magnitude }));
+
+  const rows = peaks
+    .map(p => `<tr><td>${p.frequencyHz}</td><td>${p.magnitude}</td></tr>`)
+    .join("");
+
+  const html = DOMPurify.sanitize(
+    `<table><tr><th>Frequency (Hz)</th><th>Magnitude</th></tr>${rows}</table>`
+  );
+
+  return { peaks, html, signalLength: N };
+}
+export default analyzeSignal;
+// Generation time: 16.732s
+// Result: PASS

tests/3_signal_pipeline/outputs/z-ai_glm-5.1.js

@@ -0,0 +1,44 @@
+const analyzeSignal = async (yamlStr) => {
+  const [yaml, math, ndarray, fft, DOMPurify] = await Promise.all([
+    import('https://esm.sh/js-yaml').then(m => m.default),
+    import('https://esm.sh/mathjs').then(m => m.default),
+    import('https://esm.sh/ndarray').then(m => m.default),
+    import('https://esm.sh/ndarray-fft').then(m => m.default),
+    import('https://esm.sh/dompurify').then(m => m.default)
+  ])
+
+  const { sampleRate: sr, duration, components } = yaml.load(yamlStr)
+  const N = sr * duration
+  const signal = new Float64Array(N)
+
+  for (let i = 0; i < N; i++) {
+    const t = i / sr
+    signal[i] = components.reduce(
+      (s, { amplitude: a, frequency: f }) => s + a * math.sin(2 * math.pi * f * t), 0)
+  }
+
+  const re = ndarray(signal, [N])
+  const im = ndarray(new Float64Array(N), [N])
+  fft(1, re, im)
+
+  const peaks = []
+  for (let k = 0; k <= N / 2; k++) {
+    const mag = math.sqrt(re.get(k) ** 2 + im.get(k) ** 2) / (N / 2)
+    if (mag > 0.1)
+      peaks.push({
+        frequencyHz: Math.round(k * sr / N),
+        magnitude: Math.round(mag * 100) / 100
+      })
+  }
+
+  peaks.sort((a, b) => b.magnitude - a.magnitude)
+
+  const html = `<table><tr><th>Frequency (Hz)</th><th>Magnitude</th></tr>${peaks
+    .map(p => `<tr><td>${p.frequencyHz}</td><td>${p.magnitude}</td></tr>`)
+    .join('')}</table>`
+
+  return { peaks, html: DOMPurify.sanitize(html), signalLength: N }
+}
+export default analyzeSignal;
+// Generation time: 72.180s
+// Result: FAIL

tests/4_hexchain_pipeline/outputs/anthropic_claude-opus-4.7 EFF_medium.js

@@ -0,0 +1,58 @@
+const hexchain = async toml => {
+    const CDN = 'https://esm.sh/';
+    const [
+        {parse},
+        {default: seedrandom},
+        ss,
+        {default: Ajv},
+        {default: table},
+        {default: DOMPurify}
+    ] = await Promise.all([
+        import(`${CDN}smol-toml`),
+        import(`${CDN}seedrandom`),
+        import(`${CDN}simple-statistics`),
+        import(`${CDN}ajv`),
+        import(`${CDN}text-table`),
+        import(`${CDN}dompurify`)
+    ]);
+
+    const config = parse(toml);
+    const schema = {
+        type: 'object',
+        required: ['seed', 'count', 'label'],
+        properties: {
+            seed: {type: 'string'},
+            count: {type: 'integer', minimum: 1, maximum: 10000},
+            label: {type: 'string', minLength: 1}
+        }
+    };
+    const ajv = new Ajv();
+    const validate = ajv.compile(schema);
+    if (!validate(config)) return {valid: false, errors: ajv.errorsText(validate.errors)};
+
+    const rng = new seedrandom(config.seed);
+    const nums = Array.from({length: config.count}, () => rng());
+    const r = n => Math.round(n * 1e6) / 1e6;
+    const mean = r(ss.mean(nums));
+    const stddev = r(ss.standardDeviation(nums));
+    const median = r(ss.median(nums));
+
+    const tableStr = table([
+        ['Stat', 'Value'],
+        ['mean', String(mean)],
+        ['stddev', String(stddev)],
+        ['median', String(median)]
+    ]);
+    const sanitized = DOMPurify.sanitize(`<pre class="stats">${tableStr}</pre>`);
+
+    return {
+        valid: true,
+        label: config.label,
+        stats: {mean, stddev, median},
+        table: sanitized,
+        count: config.count
+    };
+};
+export default hexchain;
+// Generation time: 8.530s
+// Result: PASS

tests/4_hexchain_pipeline/outputs/deepseek_deepseek-v4-pro.js

@@ -0,0 +1,48 @@
+async function hexchain(t){
+  const [T,S,SS,A,B,P]=await Promise.all([
+    import('smol-toml'),
+    import('seedrandom'),
+    import('simple-statistics'),
+    import('ajv'),
+    import('text-table'),
+    import('dompurify')
+  ]);
+  const c=T.parse(t);
+  const Ajv=A.default||A,ajv=new Ajv;
+  const schema={
+    type:'object',
+    required:['seed','count','label'],
+    properties:{
+      seed:{type:'string'},
+      count:{type:'integer',minimum:1,maximum:10000},
+      label:{type:'string',minLength:1}
+    }
+  };
+  const v=ajv.compile(schema);
+  if(!v(c))return{valid:false,errors:ajv.errorsText(v.errors)};
+  const rng=new (S.default||S)(c.seed),
+        arr=Array.from({length:c.count},rng);
+  const ss=SS.default||SS;
+  let m=ss.mean(arr),d=ss.standardDeviation(arr),md=ss.median(arr);
+  m=+m.toFixed(6);
+  d=+d.toFixed(6);
+  md=+md.toFixed(6);
+  const tbl=(B.default||B)([
+    ['Stat','Value'],
+    ['mean',String(m)],
+    ['stddev',String(d)],
+    ['median',String(md)]
+  ]);
+  const purify=P.default||P,
+        h=purify.sanitize(`<pre class="stats">${tbl}</pre>`);
+  return{
+    valid:true,
+    label:c.label,
+    stats:{mean:m,stddev:d,median:md},
+    table:h,
+    count:c.count
+  };
+}
+export default hexchain;
+// Generation time: 121.732s
+// Result: FAIL

tests/4_hexchain_pipeline/outputs/google_gemini-3.1-flash-lite.js

@@ -0,0 +1,54 @@
+async function hexchain(tomlStr) {
+  const [
+    { parse },
+    { default: seedrandom },
+    { mean, standardDeviation, median },
+    { default: Ajv },
+    { default: table },
+    { DOMPurify }
+  ] = await Promise.all([
+    import('https://esm.sh/smol-toml'),
+    import('https://esm.sh/seedrandom'),
+    import('https://esm.sh/simple-statistics'),
+    import('https://esm.sh/ajv'),
+    import('https://esm.sh/text-table'),
+    import('https://esm.sh/dompurify')
+  ]);
+
+  const config = parse(tomlStr);
+  const ajv = new Ajv();
+  const validate = ajv.compile({
+    type: 'object',
+    required: ['seed', 'count', 'label'],
+    properties: {
+      seed: { type: 'string' },
+      count: { type: 'integer', minimum: 1, maximum: 10000 },
+      label: { type: 'string', minLength: 1 }
+    }
+  });
+
+  if (!validate(config)) return { valid: false, errors: ajv.errorsText(validate.errors) };
+
+  const rng = new seedrandom(config.seed);
+  const nums = Array.from({ length: config.count }, rng);
+  const round = n => +(n.toFixed(6));
+  const stats = { mean: round(mean(nums)), stddev: round(standardDeviation(nums)), median: round(median(nums)) };
+  
+  const tbl = table([
+    ['Stat', 'Value'],
+    ['mean', String(stats.mean)],
+    ['stddev', String(stats.stddev)],
+    ['median', String(stats.median)]
+  ]);
+
+  return {
+    valid: true,
+    label: config.label,
+    stats,
+    table: DOMPurify.sanitize(`<pre class="stats">${tbl}</pre>`),
+    count: config.count
+  };
+}
+export default hexchain;
+// Generation time: 1.959s
+// Result: FAIL

tests/4_hexchain_pipeline/outputs/moonshotai_kimi-k2.6.js

@@ -0,0 +1,39 @@
+const hexchain = async s => {
+  const [{ parse }, { default: Seed }, { mean, standardDeviation, median }, { default: Ajv }, { default: tbl }, { default: DOMPurify }] = await Promise.all([
+    import('https://esm.sh/smol-toml'),
+    import('https://esm.sh/seedrandom'),
+    import('https://esm.sh/simple-statistics'),
+    import('https://esm.sh/ajv'),
+    import('https://esm.sh/text-table'),
+    import('https://esm.sh/dompurify')
+  ]);
+  const cfg = parse(s);
+  const ajv = new Ajv();
+  const schema = {
+    type: 'object',
+    required: ['seed', 'count', 'label'],
+    properties: {
+      seed: { type: 'string' },
+      count: { type: 'integer', minimum: 1, maximum: 10000 },
+      label: { type: 'string', minLength: 1 }
+    }
+  };
+  if (!ajv.validate(schema, cfg)) return { valid: false, errors: ajv.errorsText() };
+  const rng = new Seed(cfg.seed);
+  const nums = Array.from({ length: cfg.count }, () => rng());
+  const r = v => +v.toFixed(6);
+  const m = r(mean(nums));
+  const d = r(standardDeviation(nums));
+  const e = r(median(nums));
+  const ascii = tbl([['Stat', 'Value'], ['mean', String(m)], ['stddev', String(d)], ['median', String(e)]]);
+  return {
+    valid: true,
+    label: cfg.label,
+    stats: { mean: m, stddev: d, median: e },
+    table: DOMPurify.sanitize('<pre class="stats">' + ascii + '</pre>'),
+    count: cfg.count
+  };
+};
+export default hexchain;
+// Generation time: 147.472s
+// Result: PASS

tests/4_hexchain_pipeline/outputs/openai_gpt-5.5 EFF_high.js

@@ -0,0 +1,70 @@
+const hexchainUrls = [
+  'https://esm.sh/smol-toml?bundle',
+  'https://esm.sh/seedrandom?bundle',
+  'https://esm.sh/simple-statistics?bundle',
+  'https://esm.sh/ajv?bundle',
+  'https://esm.sh/text-table?bundle',
+  'https://esm.sh/dompurify?bundle'
+];
+
+const hexchainSchema = {
+  type: 'object',
+  required: ['seed', 'count', 'label'],
+  properties: {
+    seed: { type: 'string' },
+    count: { type: 'integer', minimum: 1, maximum: 10000 },
+    label: { type: 'string', minLength: 1 }
+  }
+};
+
+let hexchainModules;
+
+async function hexchain(tomlString) {
+  const [
+    toml,
+    seedMod,
+    ssMod,
+    ajvMod,
+    tableMod,
+    purifyMod
+  ] = await (hexchainModules ||= Promise.all(hexchainUrls.map(u => import(u))));
+
+  const parse = toml.parse ?? toml.default?.parse ?? toml.default;
+  const seedrandom = seedMod.default ?? seedMod;
+  const ss = ssMod.mean ? ssMod : ssMod.default;
+  const Ajv = ajvMod.default ?? ajvMod.Ajv;
+  const textTable = tableMod.default ?? tableMod;
+  const purify = purifyMod.default ?? purifyMod;
+  const DOMPurify = purify.sanitize ? purify : purify(globalThis);
+  const config = parse(tomlString);
+  const ajv = new Ajv();
+
+  if (!ajv.validate(hexchainSchema, config))
+    return { valid: false, errors: ajv.errorsText() };
+
+  const rng = new seedrandom(config.seed);
+  const nums = Array.from({ length: config.count }, () => rng());
+  const round = n => Math.round((n + Number.EPSILON) * 1e6) / 1e6;
+  const mean = round(ss.mean(nums));
+  const stddev = round(ss.standardDeviation(nums));
+  const median = round(ss.median(nums));
+  const tableString = textTable([
+    ['Stat', 'Value'],
+    ['mean', String(mean)],
+    ['stddev', String(stddev)],
+    ['median', String(median)]
+  ]);
+
+  return {
+    valid: true,
+    label: config.label,
+    stats: { mean, stddev, median },
+    table: DOMPurify.sanitize('<pre class="stats">' + tableString + '</pre>'),
+    count: config.count
+  };
+}
+
+globalThis.hexchain = hexchain;
+export default hexchain;
+// Generation time: 137.860s
+// Result: PASS

tests/4_hexchain_pipeline/outputs/z-ai_glm-5.1.js

@@ -0,0 +1,42 @@
+async function hexchain(tomlStr) {
+  const [{ parse }, { default: seedrandom }, ss, { default: Ajv }, { default: table }, { default: DOMPurify }] = await Promise.all([
+    import('https://esm.sh/smol-toml'),
+    import('https://esm.sh/seedrandom'),
+    import('https://esm.sh/simple-statistics'),
+    import('https://esm.sh/ajv'),
+    import('https://esm.sh/text-table'),
+    import('https://esm.sh/dompurify')
+  ])
+  const config = parse(tomlStr)
+  const ajv = new Ajv()
+  if (!ajv.validate({
+    type: 'object',
+    required: ['seed', 'count', 'label'],
+    properties: {
+      seed: { type: 'string' },
+      count: { type: 'integer', minimum: 1, maximum: 10000 },
+      label: { type: 'string', minLength: 1 }
+    }
+  }, config)) return { valid: false, errors: ajv.errorsText() }
+  const rng = new seedrandom(config.seed)
+  const nums = Array.from({ length: config.count }, () => rng())
+  const mean = +ss.mean(nums).toFixed(6)
+  const stddev = +ss.standardDeviation(nums).toFixed(6)
+  const median = +ss.median(nums).toFixed(6)
+  const tableStr = table([
+    ['Stat', 'Value'],
+    ['mean', String(mean)],
+    ['stddev', String(stddev)],
+    ['median', String(median)]
+  ])
+  return {
+    valid: true,
+    label: config.label,
+    stats: { mean, stddev, median },
+    table: DOMPurify.sanitize(`<pre class="stats">${tableStr}</pre>`),
+    count: config.count
+  }
+}
+export default hexchain;
+// Generation time: 101.900s
+// Result: PASS

tests/5_markdown_parser/outputs/anthropic_claude-opus-4.7 EFF_medium.js

@@ -0,0 +1,7 @@
+const parseMarkdown = async md => {
+  const { marked } = await import('https://cdn.jsdelivr.net/npm/marked/+esm');
+  return marked.parse(md);
+};
+export default parseMarkdown;
+// Generation time: 1.391s
+// Result: PASS

tests/5_markdown_parser/outputs/deepseek_deepseek-v4-pro.js

@@ -0,0 +1,7 @@
+async function parseMarkdown(m){
+  const{marked}=await import('//cdn.jsdelivr.net/npm/marked@4/lib/marked.esm.js')
+  return marked(m)
+}
+export default parseMarkdown;
+// Generation time: 44.452s
+// Result: FAIL

tests/5_markdown_parser/outputs/google_gemini-3.1-flash-lite.js

@@ -0,0 +1,13 @@
+async function parseMarkdown(md) {
+  const { marked } = await import("https://cdn.jsdelivr.net/npm/marked/lib/marked.esm.js");
+  
+  marked.setOptions({
+    gfm: true,
+    breaks: true
+  });
+
+  return marked.parse(md);
+}
+export default parseMarkdown;
+// Generation time: 0.838s
+// Result: PASS

tests/5_markdown_parser/outputs/moonshotai_kimi-k2.6.js

@@ -0,0 +1,7 @@
+const parseMarkdown=async(md)=>{
+const{default:MarkdownIt}=await import('https://esm.sh/markdown-it@14.1.0');
+return new MarkdownIt({html:false}).render(md);
+};
+export default parseMarkdown;
+// Generation time: 95.822s
+// Result: PASS

tests/5_markdown_parser/outputs/openai_gpt-5.5 EFF_high.js

@@ -0,0 +1,25 @@
+const parseMarkdown = (() => {
+  let libs;
+
+  return async markdown => {
+    libs ||= Promise.all([
+      import("https://cdn.jsdelivr.net/npm/marked@13/+esm"),
+      import("https://cdn.jsdelivr.net/npm/dompurify@3/+esm")
+    ]);
+
+    const [{ marked }, { default: DOMPurify }] = await libs;
+
+    marked.use({
+      gfm: true,
+      breaks: false,
+      async: false
+    });
+
+    return DOMPurify.sanitize(
+      marked.parse(String(markdown ?? ""))
+    );
+  };
+})();
+export default parseMarkdown;
+// Generation time: 13.490s
+// Result: PASS

tests/5_markdown_parser/outputs/z-ai_glm-5.1.js

@@ -0,0 +1,7 @@
+async function parseMarkdown(m) {
+  const { marked } = await import('https://esm.sh/marked')
+  return marked.parse(m)
+}
+export default parseMarkdown;
+// Generation time: 29.156s
+// Result: PASS

tests/6_csv_processor/outputs/anthropic_claude-opus-4.7 EFF_medium.js

@@ -0,0 +1,23 @@
+async function processCSV(csv, { filterColumn, filterValue, groupBy, aggregateColumn, operation }) {
+  const { default: Papa } = await import('https://esm.sh/papaparse@5.4.1');
+  const { data } = Papa.parse(csv, { header: true, skipEmptyLines: true });
+  const groups = {};
+  for (const row of data) {
+    if (row[filterColumn] != filterValue) continue;
+    const key = row[groupBy];
+    const n = Number(row[aggregateColumn]);
+    const val = Number.isFinite(n) ? n : 0;
+    (groups[key] ??= { sum: 0, count: 0 });
+    groups[key].sum += val;
+    groups[key].count++;
+  }
+  const calc = {
+    sum: g => g.sum,
+    avg: g => g.sum / g.count,
+    count: g => g.count,
+  }[operation];
+  return Object.entries(groups).map(([k, g]) => ({ [groupBy]: k, result: calc(g) }));
+}
+export default processCSV;
+// Generation time: 5.004s
+// Result: PASS

tests/6_csv_processor/outputs/deepseek_deepseek-v4-pro.js

@@ -0,0 +1,22 @@
+async function processCSV(csv, c) {
+  const { default: P } = await import(
+    'https://cdn.jsdelivr.net/npm/papaparse@5/+esm'
+  );
+  const { filterColumn: f, filterValue: v, groupBy: g, aggregateColumn: a, operation: o } = c;
+  const rows = P.parse(csv, { header: true, skipEmptyLines: true }).data;
+  const filtered = rows.filter((r) => r[f] == v);
+  const groups = {};
+  for (const r of filtered) {
+    const k = r[g];
+    groups[k] = groups[k] || { n: 0, s: 0 };
+    groups[k].n++;
+    groups[k].s += +r[a] || 0;
+  }
+  return Object.entries(groups).map(([k, { n, s }]) => ({
+    [g]: k,
+    result: o === 'sum' ? s : o === 'avg' ? s / n : n,
+  }));
+}
+export default processCSV;
+// Generation time: 75.337s
+// Result: PASS

tests/6_csv_processor/outputs/google_gemini-3.1-flash-lite.js

@@ -0,0 +1,24 @@
+async function processCSV(csv, { filterColumn, filterValue, groupBy, aggregateColumn, operation }) {
+  const { parse } = await import('https://cdn.skypack.dev/papaparse');
+  const { data } = parse(csv, { header: true, dynamicTyping: true });
+
+  const groups = data
+    .filter(r => r[filterColumn] == filterValue)
+    .reduce((acc, r) => {
+      const key = r[groupBy];
+      const val = Number(r[aggregateColumn]) || 0;
+      acc[key] ??= { sum: 0, count: 0 };
+      acc[key].sum += val;
+      acc[key].count += 1;
+      return acc;
+    }, {});
+
+  return Object.entries(groups).map(([k, v]) => ({
+    [groupBy]: k,
+    result: operation === 'sum' ? v.sum :
+            operation === 'avg' ? v.sum / v.count : v.count
+  }));
+}
+export default processCSV;
+// Generation time: 1.360s
+// Result: PASS

tests/6_csv_processor/outputs/moonshotai_kimi-k2.6.js

@@ -0,0 +1,17 @@
+async function processCSV(csv,{filterColumn,filterValue,groupBy:g,aggregateColumn,operation}){
+const[Papa,{groupBy,sum}]=await Promise.all([
+import('https://esm.sh/papaparse@5.4.1').then(m=>m.default),
+import('https://esm.sh/lodash-es@4.17.21')
+]);
+const{data}=Papa.parse(csv,{header:true,skipEmptyLines:true});
+const filtered=data.filter(r=>r[filterColumn]==filterValue);
+const groups=groupBy(filtered,r=>r[g]);
+return Object.entries(groups).map(([k,v])=>{
+const n=v.map(r=>+r[aggregateColumn]||0);
+const result=operation==='sum'?sum(n):operation==='avg'?sum(n)/n.length:v.length;
+return{[g]:k,result};
+});
+}
+export default processCSV;
+// Generation time: 205.075s
+// Result: PASS

tests/6_csv_processor/outputs/openai_gpt-5.5 EFF_high.js

@@ -0,0 +1,34 @@
+const csvDeps = Promise.all([
+  import("https://cdn.jsdelivr.net/npm/d3-dsv@3/+esm"),
+  import("https://cdn.jsdelivr.net/npm/d3-array@3/+esm")
+]);
+
+async function processCSV(csv, config = {}) {
+  const [{ csvParse }, { rollups }] = await csvDeps;
+  const { filterColumn, filterValue, groupBy, aggregateColumn, operation } = config;
+  const ops = new Set(["sum", "avg", "count"]);
+
+  if (!ops.has(operation)) throw new TypeError("Invalid operation");
+
+  const num = r => {
+    const n = Number(r[aggregateColumn]);
+    return Number.isFinite(n) ? n : 0;
+  };
+
+  const rows = csvParse(String(csv ?? ""))
+    .filter(r => r[filterColumn] == filterValue);
+
+  return rollups(
+    rows,
+    rs => {
+      if (operation === "count") return rs.length;
+
+      const sum = rs.reduce((a, r) => a + num(r), 0);
+      return operation === "avg" ? sum / rs.length : sum;
+    },
+    r => r[groupBy]
+  ).map(([k, result]) => ({ [groupBy]: k, result }));
+}
+export default processCSV;
+// Generation time: 26.121s
+// Result: PASS

tests/6_csv_processor/outputs/z-ai_glm-5.1.js

@@ -0,0 +1,12 @@
+async function processCSV(csv, { filterColumn: fc, filterValue: fv, groupBy: gb, aggregateColumn: ac, operation: op }) {
+  const { default: P } = await import("https://cdn.jsdelivr.net/npm/papaparse@5/+esm")
+  const rows = P.parse(csv, { header: true, skipEmptyLines: true }).data.filter(r => r[fc] == fv)
+  const groups = rows.reduce((a, r) => ((a[r[gb]] ??= []).push(r), a), {})
+  return Object.entries(groups).map(([k, v]) => {
+    const sum = v.reduce((s, r) => s + (Number(r[ac]) || 0), 0)
+    return { [gb]: k, result: op == 'count' ? v.length : op == 'sum' ? sum : sum / v.length }
+  })
+}
+export default processCSV;
+// Generation time: 46.157s
+// Result: PASS

tests/7_scheduler/outputs/anthropic_claude-opus-4.7 EFF_medium.js

@@ -0,0 +1,52 @@
+async function findAvailableSlots(cal1, cal2, { durationMinutes, searchRange, workHours }) {
+  const { parseISO, addMinutes } = await import('https://cdn.skypack.dev/date-fns');
+  const ms = s => parseISO(s).getTime();
+  const iso = t => new Date(t).toISOString();
+  const parseHM = t => {
+    const [h, m] = t.split(':').map(Number);
+    return (h * 60 + m) * 60000;
+  };
+
+  const rs = ms(searchRange.start);
+  const re = ms(searchRange.end);
+  const durMs = durationMinutes * 60000;
+  const whS = parseHM(workHours.start);
+  const whE = parseHM(workHours.end);
+  const DAY = 86400000;
+
+  const merged = [...cal1, ...cal2]
+    .map(b => [ms(b.start), ms(b.end)])
+    .filter(([s, e]) => e > rs && s < re)
+    .map(([s, e]) => [Math.max(s, rs), Math.min(e, re)])
+    .sort((a, b) => a[0] - b[0])
+    .reduce((acc, [s, e]) => {
+      const last = acc.at(-1);
+      if (last && s <= last[1]) last[1] = Math.max(last[1], e);
+      else acc.push([s, e]);
+      return acc;
+    }, []);
+
+  const free = [];
+  let cur = rs;
+  for (const [s, e] of merged) {
+    if (s > cur) free.push([cur, s]);
+    cur = Math.max(cur, e);
+  }
+  if (cur < re) free.push([cur, re]);
+
+  const slots = [];
+  for (const [fs, fe] of free) {
+    for (let day = Math.floor(fs / DAY) * DAY; day < fe; day += DAY) {
+      const s = Math.max(fs, day + whS);
+      const e = Math.min(fe, day + whE);
+      for (let t = s; t + durMs <= e; t = addMinutes(new Date(t), durationMinutes).getTime()) {
+        slots.push({ start: iso(t), end: iso(t + durMs) });
+      }
+    }
+  }
+
+  return slots;
+}
+export default findAvailableSlots;
+// Generation time: 28.829s
+// Result: PASS

tests/7_scheduler/outputs/deepseek_deepseek-v4-pro.js

@@ -0,0 +1,82 @@
+let _dayjsPromise;
+
+const loadDayjs = () => _dayjsPromise || (_dayjsPromise = (async () => {
+  const dayjs = (await import('https://cdn.skypack.dev/dayjs@1')).default;
+  const utc = (await import('https://cdn.skypack.dev/dayjs@1/plugin/utc')).default;
+  dayjs.extend(utc);
+  return dayjs;
+})());
+
+async function findAvailableSlots(cal1, cal2, constraints) {
+  const dayjs = await loadDayjs();
+  const { durationMinutes, searchRange, workHours } = constraints;
+  const rStart = dayjs.utc(searchRange.start);
+  const rEnd = dayjs.utc(searchRange.end);
+  const [wSh, wSm] = workHours.start.split(':').map(Number);
+  const [wEh, wEm] = workHours.end.split(':').map(Number);
+
+  // Merge all busy slots
+  const allBusy = [...cal1, ...cal2]
+    .map(s => ({ start: dayjs.utc(s.start), end: dayjs.utc(s.end) }))
+    .filter(s => s.start.isValid() && s.end.isValid() && s.start.isBefore(s.end));
+
+  allBusy.sort((a, b) => a.start.valueOf() - b.start.valueOf());
+
+  const merged = [];
+  for (const busy of allBusy) {
+    if (!merged.length || busy.start.isAfter(merged[merged.length - 1].end)) {
+      merged.push({ start: busy.start, end: busy.end });
+    } else if (busy.end.isAfter(merged[merged.length - 1].end)) {
+      merged[merged.length - 1].end = busy.end;
+    }
+  }
+
+  const result = [];
+  let day = rStart.startOf('day');
+  const endDay = rEnd.startOf('day');
+
+  while (day.isBefore(endDay) || day.isSame(endDay, 'day')) {
+    const workStart = day.hour(wSh).minute(wSm).second(0);
+    const workEnd = day.hour(wEh).minute(wEm).second(0);
+    const dStart = dayjs.max(workStart, rStart);
+    const dEnd = dayjs.min(workEnd, rEnd);
+
+    if (dStart.isBefore(dEnd)) {
+      let gapStart = dStart;
+
+      for (const busy of merged) {
+        if (busy.end.isBefore(dStart) || busy.start.isAfter(dEnd)) continue;
+        if (busy.start.isAfter(gapStart)) {
+          const freeStart = gapStart;
+          const freeEnd = busy.start;
+          let slotStart = freeStart;
+          while (slotStart.add(durationMinutes, 'minute').isBefore(freeEnd) || slotStart.add(durationMinutes, 'minute').isSame(freeEnd)) {
+            const slotEnd = slotStart.add(durationMinutes, 'minute');
+            if (slotEnd.isAfter(freeEnd)) break;
+            result.push({ start: slotStart.toISOString(), end: slotEnd.toISOString() });
+            slotStart = slotEnd;
+          }
+          gapStart = busy.end;
+        } else if (busy.end.isAfter(gapStart)) {
+          gapStart = busy.end;
+        }
+      }
+
+      if (gapStart.isBefore(dEnd)) {
+        let slotStart = gapStart;
+        while (slotStart.add(durationMinutes, 'minute').isBefore(dEnd) || slotStart.add(durationMinutes, 'minute').isSame(dEnd)) {
+          const slotEnd = slotStart.add(durationMinutes, 'minute');
+          if (slotEnd.isAfter(dEnd)) break;
+          result.push({ start: slotStart.toISOString(), end: slotEnd.toISOString() });
+          slotStart = slotEnd;
+        }
+      }
+    }
+    day = day.add(1, 'day');
+  }
+
+  return result.sort((a, b) => a.start.localeCompare(b.start));
+}
+export default findAvailableSlots;
+// Generation time: 135.095s
+// Result: FAIL

tests/7_scheduler/outputs/google_gemini-3.1-flash-lite.js

@@ -0,0 +1,41 @@
+async function findAvailableSlots(cal1, cal2, { durationMinutes, searchRange, workHours }) {
+  const { DateTime, Interval } = await import('https://cdn.skypack.dev/luxon');
+
+  const toUTC = (iso) => DateTime.fromISO(iso, { zone: 'utc' });
+  const parseTime = (t, d) => DateTime.fromISO(`${d.toISODate()}T${t}:00Z`, { zone: 'utc' });
+
+  const busy = [...cal1, ...cal2].map(s => Interval.fromDateTimes(toUTC(s.start), toUTC(s.end)));
+  const merged = Interval.merge(busy);
+
+  const start = toUTC(searchRange.start);
+  const end = toUTC(searchRange.end);
+  const slots = [];
+
+  let current = start;
+  while (current < end) {
+    let day = current.startOf('day');
+    let workStart = parseTime(workHours.start, day);
+    let workEnd = parseTime(workHours.end, day);
+
+    let windowStart = current > workStart ? current : workStart;
+    let windowEnd = workEnd;
+
+    for (let s = windowStart; s.plus({ minutes: durationMinutes }) <= windowEnd; ) {
+      let slotEnd = s.plus({ minutes: durationMinutes });
+      let interval = Interval.fromDateTimes(s, slotEnd);
+
+      if (!merged.some(b => b.overlaps(interval))) {
+        slots.push({ start: s.toISO(), end: slotEnd.toISO() });
+        s = slotEnd;
+      } else {
+        s = s.plus({ minutes: 1 });
+      }
+    }
+    current = day.plus({ days: 1 });
+  }
+
+  return slots;
+}
+export default findAvailableSlots;
+// Generation time: 2.010s
+// Result: PASS

tests/7_scheduler/outputs/moonshotai_kimi-k2.6.js

@@ -0,0 +1,50 @@
+const findAvailableSlots = async (c1, c2, c) => {
+  const { parseISO, addMinutes, min, max, isBefore, isAfter } = await import('https://esm.sh/date-fns@3.6.0');
+  const dur = c.durationMinutes;
+  if (dur <= 0) return [];
+  const utc = s => parseISO(/Z|[+-]\d{2}:?\d{2}$/.test(s) ? s : s.includes('T') ? s + 'Z' : s + 'T00:00:00Z');
+  const mins = s => { const [h, m] = s.split(':'); return (+h * 60 + +m) * 6e4; };
+  const rs = utc(c.searchRange.start), re = utc(c.searchRange.end);
+  const ws = mins(c.workHours.start), we = mins(c.workHours.end);
+  const day = 864e5;
+
+  const b = [...(c1 || []), ...(c2 || [])]
+    .map(x => ({ s: utc(x.start), e: utc(x.end) }))
+    .filter(x => isBefore(x.s, re) && isAfter(x.e, rs))
+    .map(x => ({ s: max([x.s, rs]), e: min([x.e, re]) }))
+    .sort((a, b) => +a.s - +b.s || +a.e - +b.e);
+
+  const m = [];
+  for (const x of b) {
+    const l = m[m.length - 1];
+    if (l && +x.s <= +l.e) l.e = max([l.e, x.e]);
+    else m.push({ s: x.s, e: x.e });
+  }
+
+  const f = [];
+  let p = rs;
+  for (const x of m) {
+    if (isBefore(p, x.s)) f.push([p, x.s]);
+    p = max([p, x.e]);
+  }
+  if (isBefore(p, re)) f.push([p, re]);
+
+  const r = [];
+  for (const [a, z] of f) {
+    let cur = a;
+    while (!isAfter(cur, z)) {
+      const end = addMinutes(cur, dur);
+      if (isAfter(end, z)) break;
+      const st = +cur, en = +end;
+      if (Math.floor(st / day) === Math.floor(en / day)) {
+        const sm = st % day, em = en % day;
+        if (sm >= ws && em <= we) r.push({ start: cur.toISOString(), end: end.toISOString() });
+      }
+      cur = end;
+    }
+  }
+  return r;
+};
+export default findAvailableSlots;
+// Generation time: 492.497s
+// Result: PASS

tests/7_scheduler/outputs/openai_gpt-5.5 EFF_high.js

@@ -0,0 +1,127 @@
+let luxonPromise;
+
+async function findAvailableSlots(calendarA = [], calendarB = [], constraints = {}) {
+  const { DateTime } = await (luxonPromise ||= import("https://esm.sh/luxon@3.5.0?bundle"));
+  const M = 6e4, D = 864e5;
+  const err = (E, m) => { throw new E(m); };
+
+  const toMs = v => {
+    if (typeof v !== "string") err(TypeError, `Invalid ISO date: ${v}`);
+    const d = DateTime.fromISO(v, { zone: "utc" });
+    if (!d.isValid) err(TypeError, `Invalid ISO date: ${v}`);
+    return d.toUTC().toMillis();
+  };
+
+  const toMinutes = v => {
+    if (typeof v !== "string") err(TypeError, `Invalid UTC time: ${v}`);
+    const m = /^(\d{2}):(\d{2})$/.exec(v);
+    if (!m) err(TypeError, `Invalid UTC time: ${v}`);
+
+    const h = +m[1], n = +m[2];
+    if (h > 24 || n > 59 || h === 24 && n)
+      err(RangeError, `Invalid UTC time: ${v}`);
+
+    return h * 60 + n;
+  };
+
+  if (!Array.isArray(calendarA) || !Array.isArray(calendarB))
+    err(TypeError, "Calendars must be arrays");
+
+  if (!constraints || typeof constraints !== "object" || Array.isArray(constraints))
+    err(TypeError, "Constraints must be an object");
+
+  if (typeof constraints.durationMinutes !== "number")
+    err(TypeError, "durationMinutes must be a number");
+
+  const duration = constraints.durationMinutes * M;
+  if (!Number.isFinite(duration) || duration <= 0 || !Number.isInteger(duration))
+    err(RangeError, "durationMinutes must resolve to a positive whole millisecond value");
+
+  const rangeStart = toMs(constraints.searchRange?.start);
+  const rangeEnd = toMs(constraints.searchRange?.end);
+
+  if (rangeEnd <= rangeStart)
+    err(RangeError, "searchRange.end must be after searchRange.start");
+
+  const workStart = toMinutes(constraints.workHours?.start);
+  const workEnd = toMinutes(constraints.workHours?.end);
+
+  if (workStart === 1440)
+    err(RangeError, "workHours.start must be before 24:00");
+
+  const midnight = ms => DateTime.fromMillis(ms, { zone: "utc" }).startOf("day").toMillis();
+  const overnight = workEnd <= workStart;
+  const work = [];
+
+  const addWork = (s, e) => {
+    const a = Math.max(s, rangeStart), b = Math.min(e, rangeEnd);
+    if (a >= b) return;
+
+    const l = work[work.length - 1];
+    if (l && a <= l[1]) l[1] = Math.max(l[1], b);
+    else work.push([a, b]);
+  };
+
+  for (
+    let d = midnight(rangeStart) - (overnight ? D : 0), end = midnight(rangeEnd);
+    d <= end;
+    d += D
+  )
+    addWork(d + workStart * M, d + workEnd * M + (overnight ? D : 0));
+
+  const busy = calendarA.concat(calendarB)
+    .map(slot => {
+      if (!slot || typeof slot !== "object")
+        err(TypeError, "Busy slots must be objects");
+
+      const s = toMs(slot.start), e = toMs(slot.end);
+      if (e <= s) err(RangeError, "Busy slot end must be after start");
+
+      return [Math.max(s, rangeStart), Math.min(e, rangeEnd)];
+    })
+    .filter(([s, e]) => s < e)
+    .sort((a, b) => a[0] - b[0] || a[1] - b[1]);
+
+  const merged = [];
+
+  for (const [s, e] of busy) {
+    const l = merged[merged.length - 1];
+    if (l && s <= l[1]) l[1] = Math.max(l[1], e);
+    else merged.push([s, e]);
+  }
+
+  const slots = [];
+  const addSlots = (s, e) => {
+    for (; s + duration <= e; s += duration)
+      slots.push([s, s + duration]);
+  };
+
+  let i = 0;
+
+  for (const [ws, we] of work) {
+    while (merged[i]?.[1] <= ws) i++;
+
+    let p = ws;
+
+    for (let j = i; merged[j]?.[0] < we; j++) {
+      const [bs, be] = merged[j];
+
+      if (bs > p) addSlots(p, Math.min(bs, we));
+
+      p = Math.max(p, Math.min(be, we));
+      if (p >= we) break;
+    }
+
+    if (p < we) addSlots(p, we);
+  }
+
+  const iso = ms => DateTime.fromMillis(ms, { zone: "utc" }).toISO();
+
+  return slots.map(([start, end]) => ({
+    start: iso(start),
+    end: iso(end)
+  }));
+}
+export default findAvailableSlots;
+// Generation time: 263.158s
+// Result: PASS

tests/7_scheduler/outputs/z-ai_glm-5.1.js

@@ -0,0 +1,35 @@
+const findAvailableSlots = async (c1, c2, cn) => {
+  const { DateTime: D } = await import('https://cdn.jsdelivr.net/npm/luxon@3/+esm');
+  const f = s => D.fromISO(s, { zone: 'UTC' });
+  const m = t => { const [h, v] = t.split(':'); return +h * 60 + +v; };
+  const g = d => d.hour * 60 + d.minute;
+  const ws = m(cn.workHours.start), we = m(cn.workHours.end), dur = cn.durationMinutes;
+  let merged = [...c1, ...c2].map(x => ({ s: f(x.start), e: f(x.end) })).sort((a, b) => a.s - b.s);
+  let busy = [];
+  for (let b of merged) {
+    if (!busy.length || b.s > busy[busy.length - 1].e) busy.push(b);
+    else if (b.e > busy[busy.length - 1].e) busy[busy.length - 1].e = b.e;
+  }
+  let res = [], cur = f(cn.searchRange.start), end = f(cn.searchRange.end);
+  const nWs = d => d.plus({ days: 1 }).startOf('day').plus({ minutes: ws });
+  const proc = (fS, fE) => {
+    let c = fS;
+    if (g(c) < ws) c = c.startOf('day').plus({ minutes: ws });
+    while (c < fE) {
+      if (g(c) >= we) { c = nWs(c); continue; }
+      let e = c.plus({ minutes: dur });
+      if (e > fE) break;
+      if (g(e) <= we) { res.push({ start: c.toISO(), end: e.toISO() }); c = e; }
+      else { if (g(c) <= ws) break; c = nWs(c); }
+    }
+  };
+  for (let b of busy) {
+    if (cur < b.s) proc(cur, b.s);
+    if (cur < b.e) cur = b.e;
+  }
+  if (cur < end) proc(cur, end);
+  return res;
+};
+export default findAvailableSlots;
+// Generation time: 134.745s
+// Result: PASS

tests/8_json_validator/outputs/anthropic_claude-opus-4.7 EFF_medium.js

@@ -0,0 +1,13 @@
+const validateJSON = async (data, schema) => {
+  const { default: Ajv } = await import('https://cdn.jsdelivr.net/npm/ajv@8/+esm');
+  const ajv = new Ajv({ allErrors: true, strict: false });
+  const validate = ajv.compile(schema);
+  const valid = validate(data);
+  return {
+    valid,
+    errors: valid ? [] : validate.errors.map(e => `${e.instancePath || '/'} ${e.message}`)
+  };
+};
+export default validateJSON;
+// Generation time: 2.748s
+// Result: PASS

tests/8_json_validator/outputs/deepseek_deepseek-v4-pro.js

@@ -0,0 +1,19 @@
+let _p
+async function validateJSON(d, s) {
+  if (!_p) _p = (async () => {
+    const {default: A} = await import('https://cdn.jsdelivr.net/npm/ajv@8.17.1/dist/ajv2020.esm.js')
+    return new A({allErrors: 1})
+  })()
+  try {
+    const a = await _p
+    const v = a.validate(s, d)
+    return v
+      ? {valid: 1, errors: []}
+      : {valid: 0, errors: a.errors.map(e => `${e.instancePath||'/'} ${e.message}`.trim())}
+  } catch (e) {
+    return {valid: 0, errors: [e.message]}
+  }
+}
+export default validateJSON;
+// Generation time: 54.656s
+// Result: FAIL

tests/8_json_validator/outputs/google_gemini-3.1-flash-lite.js

@@ -0,0 +1,18 @@
+async function validateJSON(data, schema) {
+  const { default: Ajv } = await import('https://esm.sh/ajv@8.12.0');
+  const { default: addFormats } = await import('https://esm.sh/ajv-formats@2.1.1');
+  
+  const ajv = new (Ajv)({ allErrors: true });
+  addFormats(ajv);
+
+  const validate = ajv.compile(schema);
+  const valid = validate(data);
+
+  return {
+    valid,
+    errors: valid ? [] : validate.errors.map(e => `${e.instancePath} ${e.message}`)
+  };
+}
+export default validateJSON;
+// Generation time: 1.038s
+// Result: PASS

tests/8_json_validator/outputs/moonshotai_kimi-k2.6.js

@@ -0,0 +1,17 @@
+let A;
+const validateJSON = async (json, schema) => {
+  try {
+    A ||= (await import('https://cdn.jsdelivr.net/npm/ajv@8.12.0/dist/ajv.mjs')).default;
+    const v = new A({ allErrors: !0, strict: !1 }).compile(schema);
+    const valid = v(json);
+    return {
+      valid,
+      errors: valid ? [] : v.errors.map(e => `${e.instancePath || '/'} ${e.message}`.trim())
+    };
+  } catch (e) {
+    return { valid: !1, errors: [e.message] };
+  }
+};
+export default validateJSON;
+// Generation time: 65.490s
+// Result: FAIL

tests/8_json_validator/outputs/openai_gpt-5.5 EFF_high.js

@@ -0,0 +1,101 @@
+const _jsonSchema = (() => {
+  const ajvs = new Map();
+  const vals = new WeakMap();
+  const opts = {
+    allErrors: true,
+    strict: false,
+    allowUnionTypes: true,
+    addUsedSchema: false,
+    validateFormats: true
+  };
+
+  const ver = s => {
+    s = `${s && typeof s == "object" && s.$schema || ""}`;
+    return s.includes("2020-12") ? "2020" : s.includes("2019-09") ? "2019" : "7";
+  };
+
+  const url = v => `https://esm.sh/ajv@8.17.1${v == "7" ? "" : `/dist/${v}`}?bundle`;
+
+  const load = v => {
+    let p = ajvs.get(v);
+    if (p) return p;
+
+    p = Promise.all([
+      import(url(v)),
+      import("https://esm.sh/ajv-formats@3.0.1?bundle&deps=ajv@8.17.1")
+    ]).then(([a, f]) => {
+      const Ajv = a.default || a.Ajv || a;
+      const addFormats = f.default || f.addFormats || f;
+      const ajv = new Ajv({ ...opts });
+      addFormats(ajv);
+      return ajv;
+    }).catch(e => {
+      ajvs.delete(v);
+      throw e;
+    });
+
+    ajvs.set(v, p);
+    return p;
+  };
+
+  const part = s => /^(0|[1-9]\d*)$/.test(s)
+    ? `[${s}]`
+    : /^[A-Za-z_$][\w$]*$/.test(s)
+      ? `.${s}`
+      : `[${JSON.stringify(s)}]`;
+
+  const loc = p => p
+    ? `$${p.split("/").slice(1).map(s => part(s.replace(/~1/g, "/").replace(/~0/g, "~"))).join("")}`
+    : "$";
+
+  const msg = e => {
+    const x = e.params || {};
+    const extra = e.keyword == "additionalProperties"
+      ? `: ${x.additionalProperty}`
+      : e.keyword == "required"
+        ? `: ${x.missingProperty}`
+        : "";
+
+    return `${loc(e.instancePath)} ${e.message || "is invalid"}${extra}`;
+  };
+
+  return { load, vals, ver, msg };
+})();
+
+async function validateJSON(data, schema) {
+  const isSchema = typeof schema == "boolean" ||
+    schema && typeof schema == "object" && !Array.isArray(schema);
+
+  if (!isSchema) {
+    return {
+      valid: false,
+      errors: ["Schema must be a JSON Schema object or boolean"]
+    };
+  }
+
+  try {
+    const cacheable = schema && typeof schema == "object";
+    let validate = cacheable && _jsonSchema.vals.get(schema);
+
+    if (!validate) {
+      const ajv = await _jsonSchema.load(_jsonSchema.ver(schema));
+      validate = ajv.compile(schema);
+      if (cacheable) _jsonSchema.vals.set(schema, validate);
+    }
+
+    const valid = validate(data);
+
+    return {
+      valid,
+      errors: valid ? [] : (validate.errors || []).map(_jsonSchema.msg)
+    };
+  } catch (e) {
+    return {
+      valid: false,
+      errors: [e?.message || String(e)]
+    };
+  }
+}
+export default validateJSON;
+// Generation time: 143.247s
+// Result: PASS

tests/8_json_validator/outputs/z-ai_glm-5.1.js

@@ -0,0 +1,9 @@
+async function validateJSON(json, schema) {
+  const { default: Ajv } = await import('https://esm.sh/ajv')
+  const ajv = new Ajv()
+  const valid = ajv.validate(schema, json)
+  return { valid, errors: valid ? [] : ajv.errors.map(e => e.message) }
+}
+export default validateJSON;
+// Generation time: 28.047s
+// Result: PASS

tests/9_stream_visualizer/outputs/anthropic_claude-opus-4.7 EFF_medium.js

@@ -0,0 +1,31 @@
+async function createStreamVisualizer(asyncIterable, options) {
+  const { maxPoints, alpha, width, height, yDomain } = options;
+  const d3 = await import("https://cdn.jsdelivr.net/npm/d3@7/+esm");
+  const data = [];
+  let ema;
+
+  for await (const { timestamp, value } of asyncIterable) {
+    ema = data.length ? alpha * value + (1 - alpha) * ema : value;
+    data.push({ timestamp, value, ema });
+    if (data.length > maxPoints) data.shift();
+  }
+
+  if (!data.length) return { data, path: "" };
+
+  const x = d3.scaleLinear()
+    .domain([data[0].timestamp, data[data.length - 1].timestamp])
+    .range([0, width]);
+
+  const y = d3.scaleLinear()
+    .domain(yDomain)
+    .range([height, 0]);
+
+  const path = d3.line()
+    .x(d => x(d.timestamp))
+    .y(d => y(d.ema))(data) ?? "";
+
+  return { data, path };
+}
+export default createStreamVisualizer;
+// Generation time: 6.848s
+// Result: PASS

tests/9_stream_visualizer/outputs/deepseek_deepseek-v4-pro.js

@@ -0,0 +1,23 @@
+const createStreamVisualizer = async (feed, { p = 1000, a = .25, w, h, y }) => {
+  const d3 = (await import('https://esm.sh/d3@7')).default
+  const data = []
+  for await (const { timestamp: t, value: v } of feed) {
+    const e = data.length ? a * v + (1 - a) * data[data.length - 1].ema : v
+    data.push({ timestamp: t, value: v, ema: e })
+    if (data.length > p) data.shift()
+  }
+  if (!data.length) return { data: [], path: '' }
+  const x = d3.scaleLinear()
+    .domain([data[0].timestamp, data[data.length - 1].timestamp])
+    .range([0, w])
+  const y = d3.scaleLinear()
+    .domain(y)
+    .range([h, 0])
+  const line = d3.line()
+    .x(d => x(d.timestamp))
+    .y(d => y(d.ema))
+  return { data, path: line(data) }
+}
+export default createStreamVisualizer;
+// Generation time: 71.592s
+// Result: FAIL

tests/9_stream_visualizer/outputs/google_gemini-3.1-flash-lite.js

@@ -0,0 +1,20 @@
+async function createStreamVisualizer(it, { maxPoints, alpha, width, height, yDomain }) {
+  const [{ scaleLinear, line, extent }] = await Promise.all([import('https://cdn.skypack.dev/d3')]);
+  const data = [];
+  let prevEma;
+
+  for await (const { timestamp, value } of it) {
+    prevEma = prevEma === undefined ? value : alpha * value + (1 - alpha) * prevEma;
+    data.push({ timestamp, value, ema: prevEma });
+    if (data.length > maxPoints) data.shift();
+  }
+
+  const x = scaleLinear().domain(extent(data, d => d.timestamp)).range([0, width]);
+  const y = scaleLinear().domain(yDomain).range([height, 0]);
+  const path = line().x(d => x(d.timestamp)).y(d => y(d.ema))(data);
+
+  return { data, path };
+}
+export default createStreamVisualizer;
+// Generation time: 1.461s
+// Result: PASS

tests/9_stream_visualizer/outputs/moonshotai_kimi-k2.6.js

@@ -0,0 +1,19 @@
+const createStreamVisualizer = async (stream, options = {}) => {
+  const { scaleLinear, line } = await import("https://cdn.jsdelivr.net/npm/d3@7/+esm");
+  const { maxPoints = 1e3, alpha = .2, width = 600, height = 400, yDomain } = options;
+  const data = [];
+  let ema;
+  for await (const { timestamp, value } of stream) {
+    ema = data.length ? alpha * value + (1 - alpha) * ema : value;
+    data.push({ timestamp, value, ema });
+    if (data.length > maxPoints) data.shift();
+  }
+  if (!data.length) return { data, path: "" };
+  const x = scaleLinear().domain([data[0].timestamp, data.at(-1).timestamp]).range([0, width]);
+  const y = scaleLinear().domain(yDomain).range([height, 0]);
+  const path = line().x(d => x(d.timestamp)).y(d => y(d.ema))(data);
+  return { data, path };
+};
+export default createStreamVisualizer;
+// Generation time: 68.920s
+// Result: PASS

tests/9_stream_visualizer/outputs/openai_gpt-5.5 EFF_high.js

@@ -0,0 +1,76 @@
+const d3$ = import("https://cdn.jsdelivr.net/npm/d3@7/+esm");
+
+async function createStreamVisualizer(asyncIterable, options = {}) {
+  const {
+    maxPoints,
+    alpha,
+    width,
+    height,
+    yDomain
+  } = options;
+
+  if (!asyncIterable?.[Symbol.asyncIterator])
+    throw new TypeError("asyncIterable must be an async iterable");
+
+  if (!Number.isInteger(maxPoints) || maxPoints < 1)
+    throw new RangeError("maxPoints must be a positive integer");
+
+  if (!Number.isFinite(alpha) || alpha < 0 || alpha > 1)
+    throw new RangeError("alpha must be between 0 and 1");
+
+  if (!Number.isFinite(width) || width <= 0 || !Number.isFinite(height) || height <= 0)
+    throw new RangeError("width and height must be positive numbers");
+
+  if (!Array.isArray(yDomain) || yDomain.length !== 2)
+    throw new TypeError("yDomain must be [min, max]");
+
+  const [yMin, yMax] = yDomain.map(Number);
+
+  if (!Number.isFinite(yMin) || !Number.isFinite(yMax) || yMin === yMax)
+    throw new RangeError("yDomain values must be finite and distinct");
+
+  const d3 = await d3$;
+  const data = [];
+  let ema;
+
+  for await (const item of asyncIterable) {
+    const source = Array.isArray(item)
+      ? { timestamp: item[0], value: item[1] }
+      : item;
+
+    const timestamp = +source?.timestamp;
+    const value = +source?.value;
+
+    if (!Number.isFinite(timestamp) || !Number.isFinite(value))
+      throw new TypeError("stream items must include finite timestamp and value");
+
+    ema = ema === undefined
+      ? value
+      : alpha * value + (1 - alpha) * ema;
+
+    data.push({ timestamp, value, ema });
+
+    if (data.length > maxPoints)
+      data.splice(0, data.length - maxPoints);
+  }
+
+  if (!data.length)
+    return { data, path: "" };
+
+  const x = d3.scaleLinear()
+    .domain([data[0].timestamp, data.at(-1).timestamp])
+    .range([0, width]);
+
+  const y = d3.scaleLinear()
+    .domain([yMin, yMax])
+    .range([height, 0]);
+
+  const path = d3.line()
+    .x(d => x(d.timestamp))
+    .y(d => y(d.ema))(data) || "";
+
+  return { data, path };
+}
+export default createStreamVisualizer;
+// Generation time: 27.168s
+// Result: PASS

tests/9_stream_visualizer/outputs/z-ai_glm-5.1.js

@@ -0,0 +1,14 @@
+async function createStreamVisualizer(s, {maxPoints: m, alpha: a, width: w, height: h, yDomain: [y0, y1]}) {
+  const {scaleLinear, line} = await import("https://cdn.jsdelivr.net/npm/d3@7/+esm")
+  let d = [], e
+  for await (const {timestamp: t, value: v} of s) {
+    d.push({timestamp: t, value: v, ema: e = d.length ? a * v + (1 - a) * e : v})
+    if (d.length > m) d.shift()
+  }
+  const x = scaleLinear([d[0].timestamp, d.at(-1).timestamp], [0, w]),
+        y = scaleLinear([y0, y1], [h, 0])
+  return {data: d, path: line().x(p => x(p.timestamp)).y(p => y(p.ema))(d)}
+}
+export default createStreamVisualizer;
+// Generation time: 93.516s
+// Result: PASS