quantid commited on
Commit
4546bbf
路
verified 路
1 Parent(s): d645d5d
README.md CHANGED
@@ -13,4 +13,4 @@ tags: ["news", "benchmark", "measurement", "statistics"]
13
 
14
  Measurement-first analysis of AI benchmarks, scientific computing and AI risk diagnostics, with open methods and live data.
15
 
16
- Latest: [Hugging Face Official Benchmarks: Daily Update 2026-10-20](https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-20.html), [Hugging Face Official Benchmarks: Daily Update 2026-10-18](https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-18.html), [Typed Decisions Leaderboard: A New #1 at 0.81](https://quantid-measurement-lab.static.hf.space/a/typed-decisions-new-number-one.html), [Measuring the Real Cost of an LLM Call: Tokens, Precision, and Honest Numbers](https://quantid-measurement-lab.static.hf.space/a/measuring-real-cost-llm-call.html), [Quantum Simulation and QEC Decoding on a GPU: The Open Quantum Challenge](https://quantid-measurement-lab.static.hf.space/a/quantum-challenge-simulation-and-qec.html)
 
13
 
14
  Measurement-first analysis of AI benchmarks, scientific computing and AI risk diagnostics, with open methods and live data.
15
 
16
+ Latest: [Why a Borda Score Is Harder to Game Than an Average](https://quantid-measurement-lab.static.hf.space/a/borda-rank-aggregation-why-it-is-hard-to-game-2026-10-20.html), [Hugging Face Official Benchmarks: Daily Update 2026-10-20](https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-20.html), [Hugging Face Official Benchmarks: Daily Update 2026-10-18](https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-18.html), [Typed Decisions Leaderboard: A New #1 at 0.81](https://quantid-measurement-lab.static.hf.space/a/typed-decisions-new-number-one.html), [Measuring the Real Cost of an LLM Call: Tokens, Precision, and Honest Numbers](https://quantid-measurement-lab.static.hf.space/a/measuring-real-cost-llm-call.html)
a/borda-rank-aggregation-why-it-is-hard-to-game-2026-10-20.html ADDED
@@ -0,0 +1,76 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <!doctype html><html lang="en"><head><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1">
2
+ <title>Why a Borda Score Is Harder to Game Than an Average | QuantID Measurement Lab</title><meta name="description" content="S1MB ranks 197 decision models by aggregating ranks over 137 benchmarks instead of averaging scores. Here is what that changes, and how to read the resulting table."><link rel="canonical" href="https://quantid-measurement-lab.static.hf.space/a/borda-rank-aggregation-why-it-is-hard-to-game-2026-10-20.html">
3
+ <meta property="og:type" content="article"><meta property="og:title" content="Why a Borda Score Is Harder to Game Than an Average | QuantID Measurement Lab"><meta property="og:description" content="S1MB ranks 197 decision models by aggregating ranks over 137 benchmarks instead of averaging scores. Here is what that changes, and how to read the resulting table."><meta property="og:url" content="https://quantid-measurement-lab.static.hf.space/a/borda-rank-aggregation-why-it-is-hard-to-game-2026-10-20.html">
4
+ <meta name="twitter:card" content="summary_large_image"><link rel="alternate" type="application/rss+xml" title="QuantID Measurement Lab" href="https://quantid-measurement-lab.static.hf.space/rss.xml">
5
+ <link rel="preconnect" href="https://fonts.googleapis.com"><link href="https://fonts.googleapis.com/css2?family=IBM+Plex+Sans:wght@400;600&family=IBM+Plex+Mono:wght@400;600&display=swap" rel="stylesheet">
6
+ <script type="application/ld+json">[{"@context": "https://schema.org", "@type": "NewsArticle", "headline": "Why a Borda Score Is Harder to Game Than an Average", "description": "S1MB ranks 197 decision models by aggregating ranks over 137 benchmarks instead of averaging scores. Here is what that changes, and how to read the resulting table.", "datePublished": "2026-10-20T11:20:00+09:00", "dateModified": "2026-10-20T11:20:00+09:00", "author": {"@type": "Organization", "name": "QuantID"}, "publisher": {"@type": "Organization", "name": "QuantID", "url": "https://www.quantid.co.kr"}, "mainEntityOfPage": "https://quantid-measurement-lab.static.hf.space/a/borda-rank-aggregation-why-it-is-hard-to-game-2026-10-20.html", "image": "", "keywords": "benchmarks, evaluation, ai, statistics", "inLanguage": "en"}]</script><style>
7
+ :root{--bg:#fbfaf6;--fg:#1d1b16;--mut:#6f695c;--ac:#e8590c;--card:#fff;--bd:#e4dfd3;--grid:rgba(0,0,0,.035)}
8
+ @media (prefers-color-scheme:dark){:root:not([data-theme=light]){--bg:#15140f;--fg:#ece7da;--mut:#a39c8b;--card:#1c1a14;--bd:#302c22;--grid:rgba(255,255,255,.04)}}
9
+ *{box-sizing:border-box}body{margin:0;color:var(--fg);font:16.5px/1.72 'IBM Plex Sans',system-ui,sans-serif;
10
+ background:var(--bg) linear-gradient(var(--grid) 1px,transparent 1px) 0 0/100% 24px,var(--bg) linear-gradient(90deg,var(--grid) 1px,transparent 1px) 0 0/24px 100%;background-blend-mode:normal}
11
+ a{color:var(--ac)}.w{max-width:1060px;margin:0 auto;padding:0 16px}.rd{max-width:1060px;margin:0 auto;padding:0 16px}
12
+ .mast{border-bottom:2px solid var(--fg);padding:26px 0 14px;display:flex;justify-content:space-between;align-items:flex-end;flex-wrap:wrap;gap:8px}
13
+ .mast a{font:700 26px/1 'IBM Plex Mono',monospace;color:var(--fg);text-decoration:none;letter-spacing:-.02em}.mast a i{color:var(--ac);font-style:normal}
14
+ .mast p{margin:0;font:13px 'IBM Plex Mono',monospace;color:var(--mut)}
15
+ .abs{font:15px/1.7 'IBM Plex Mono',monospace;color:var(--mut);border-left:3px solid var(--ac);padding:4px 0 4px 14px;margin:26px 0}
16
+ .idx{list-style:none;padding:0;margin:0;counter-reset:n}.idx li{counter-increment:n;border-bottom:1px dashed var(--bd)}
17
+ .idx a{display:grid;grid-template-columns:64px 1fr 120px;gap:14px;padding:18px 0;text-decoration:none;color:var(--fg)}
18
+ .idx a:before{content:"No." counter(n,decimal-leading-zero);font:600 13px 'IBM Plex Mono',monospace;color:var(--ac);padding-top:4px}
19
+ .idx h3{margin:0 0 4px;font:600 19px/1.35 'IBM Plex Sans'}.idx .d{font:12.5px 'IBM Plex Mono',monospace;color:var(--mut);text-align:right}
20
+ .meta{font:12.5px 'IBM Plex Mono',monospace;color:var(--mut)}.tag{font:12px 'IBM Plex Mono',monospace;color:var(--ac);margin-right:8px}.tag:before{content:"#"}
21
+ .paper{display:grid;grid-template-columns:1fr 240px;gap:36px}.spec{position:sticky;top:16px;align-self:start;background:var(--card);border:1px solid var(--bd);padding:14px;font:12.5px/1.6 'IBM Plex Mono',monospace}
22
+ .spec b{display:block;color:var(--ac);margin-top:8px}.spec{overflow-wrap:anywhere}
23
+ article h1{font:600 clamp(28px,4.5vw,40px)/1.2 'IBM Plex Sans';margin:28px 0 8px}article h2{font:600 22px/1.3 'IBM Plex Sans';margin-top:2em;padding-bottom:4px;border-bottom:1px solid var(--bd)}
24
+ article h2:before{content:"搂 ";color:var(--ac);font-family:'IBM Plex Mono'}article img{max-width:100%;border:1px solid var(--bd);background:#fff;padding:6px}
25
+ pre,code{font-family:'IBM Plex Mono',monospace}pre{background:var(--card);border:1px solid var(--bd);padding:14px;overflow:auto;font-size:13.5px}
26
+ table{border-collapse:collapse;display:block;overflow-x:auto;font:14px 'IBM Plex Mono',monospace}td,th{border:1px solid var(--bd);padding:6px 10px}th{background:var(--card)}
27
+ footer{border-top:2px solid var(--fg);margin-top:56px;padding:18px 0;font:12.5px 'IBM Plex Mono',monospace;color:var(--mut)}
28
+ @media(max-width:820px){.paper{grid-template-columns:1fr}.spec{position:static}.idx a{grid-template-columns:52px 1fr}.idx .d{display:none}}</style></head><body>
29
+ <header><div class="w"><div class="mast"><a href="https://quantid-measurement-lab.static.hf.space/">QuantID<i>/</i>Measurement Lab</a><p>Measuring AI, science and risk with reproducible methods</p></div></div></header><main><div class="rd"><div class="paper"><article><h1>Why a Borda Score Is Harder to Game Than an Average</h1><p class="meta">2026-10-20 路 QuantID</p><p><span class="tag">benchmarks</span><span class="tag">evaluation</span><span class="tag">ai</span><span class="tag">statistics</span></p><h2 id="the-question">The question</h2>
30
+ <p>Most leaderboards average a score across tasks. The <a href="https://huggingface.co/spaces/hotchpotch/S1MB-leaderboard">S1MB</a> decision-model board does something different: it ranks each of 137 benchmarks separately and aggregates the ranks. That is a Borda count.</p>
31
+ <p>The board currently carries 197 models, so the choice of aggregation decides who sits where. It is worth understanding before reading any position on it.</p>
32
+ <h2 id="averaging-rewards-spikes">Averaging rewards spikes</h2>
33
+ <p>Suppose a board has 100 tasks and you average the raw scores. One task has scores spread between 0.1 and 0.9, and the other 99 are packed between 0.80 and 0.84.</p>
34
+ <p>Improving on that one wide task by 0.3 moves your average by 0.003. Improving on every one of the other 99 by the full width of their spread moves it by about 0.04. Sounds like the many tasks win. But reverse it: a model that is mediocre everywhere and extraordinary on three high-variance tasks can land near the top of the average, because variance, not importance, decided the weighting.</p>
35
+ <p>Nobody chose that weighting. It fell out of how the metrics happened to be scaled.</p>
36
+ <h2 id="rank-aggregation-flattens-the-scale">Rank aggregation flattens the scale</h2>
37
+ <p>Under Borda, each task contributes the same thing: your position among the entrants. Winning a task by 0.3 and winning it by 0.001 both earn the same rank. Being last by a hair costs the same as being last by a mile.</p>
38
+ <p>Two consequences follow.</p>
39
+ <p><strong>You cannot buy a high placement with a few spikes.</strong> To finish first you have to place well on most of the 137 boards at once.</p>
40
+ <p><strong>Small real gains count.</strong> A method that reliably adds a little on many tasks gains more than one that adds a lot on a few. For an engineer picking a model for a production decision path, that is usually the property worth having.</p>
41
+ <h2 id="what-rank-aggregation-gives-up">What rank aggregation gives up</h2>
42
+ <p>It throws away margin. A board where the top five are separated by 0.001 on every task will still produce a confident-looking Borda ordering. The ranking is real, but the underlying differences may be inside noise.</p>
43
+ <p>So read both columns. S1MB publishes a task average next to the Borda score, and the two can disagree. On the current English board the top Borda entry is second on task average, and the model with the highest task average sits third on Borda. Neither column is wrong. They answer different questions.</p>
44
+ <table>
45
+ <thead>
46
+ <tr>
47
+ <th>Question</th>
48
+ <th>Column to read</th>
49
+ </tr>
50
+ </thead>
51
+ <tbody>
52
+ <tr>
53
+ <td>Who places well across the most tasks?</td>
54
+ <td>Borda</td>
55
+ </tr>
56
+ <tr>
57
+ <td>Who has the highest raw scores on average?</td>
58
+ <td>Task average</td>
59
+ </tr>
60
+ <tr>
61
+ <td>Is the gap meaningful?</td>
62
+ <td>Neither, you need the per-task spread</td>
63
+ </tr>
64
+ </tbody>
65
+ </table>
66
+ <h2 id="how-to-read-any-position-on-this-board">How to read any position on this board</h2>
67
+ <ol>
68
+ <li>Check which column the claim is quoting. "First on S1MB" without a column is incomplete.</li>
69
+ <li>Check the date. The board is live and 197 entries move.</li>
70
+ <li>Check coverage. A model scored on fewer tasks is not comparable to one scored on all 137.</li>
71
+ <li>Check whether entries are self-reported. Most public boards mix self-reported and verified numbers, and S1MB notes this.</li>
72
+ </ol>
73
+ <h2 id="a-note-on-reading-our-own-entries">A note on reading our own entries</h2>
74
+ <p>Models we publish appear on this board, including the current top Borda entry. We state the column, the date and the model count whenever we cite a position, and we do not describe our own measurement as independent verification. Readers should apply the four checks above to our entries exactly as they would to anyone else's.</p>
75
+ <p>Board: <a href="https://huggingface.co/spaces/hotchpotch/S1MB-leaderboard">S1MB leaderboard</a>, figures read 20 October 2026.</p></article><aside class="spec">SPECIMEN SHEET<b>Published</b>2026-10-20<b>Author</b>QuantID<b>Keywords</b>benchmarks, evaluation, ai, statistics<b>Method</b>Live public API data, reproducible code in text</aside></div></div></main>
76
+ <footer><div class="w">漏 QuantID 路 <a href="https://www.quantid.co.kr">quantid.co.kr</a> 路 <a href="https://quantid-measurement-lab.static.hf.space/rss.xml">RSS</a> 路 <a href="https://quantid-measurement-lab.static.hf.space/sitemap.xml">Sitemap</a></div></footer></body></html>
index.html CHANGED
@@ -26,5 +26,5 @@ pre,code{font-family:'IBM Plex Mono',monospace}pre{background:var(--card);border
26
  table{border-collapse:collapse;display:block;overflow-x:auto;font:14px 'IBM Plex Mono',monospace}td,th{border:1px solid var(--bd);padding:6px 10px}th{background:var(--card)}
27
  footer{border-top:2px solid var(--fg);margin-top:56px;padding:18px 0;font:12.5px 'IBM Plex Mono',monospace;color:var(--mut)}
28
  @media(max-width:820px){.paper{grid-template-columns:1fr}.spec{position:static}.idx a{grid-template-columns:52px 1fr}.idx .d{display:none}}</style></head><body>
29
- <header><div class="w"><div class="mast"><a href="https://quantid-measurement-lab.static.hf.space/">QuantID<i>/</i>Measurement Lab</a><p>Measuring AI, science and risk with reproducible methods</p></div></div></header><main><div class="w"><p class="abs">ABSTRACT. Measurement-first analysis of AI benchmarks, scientific computing and AI risk diagnostics, with open methods and live data.</p><ol class="idx"><li><a href="https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-20.html"><div><h3>Hugging Face Official Benchmarks: Daily Update 2026-10-20</h3><div class="meta">49 official benchmarks, 1,107 entries. 1 new #1, 0 new boards, +5 entries.</div></div><div class="d">2026-10-20</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-18.html"><div><h3>Hugging Face Official Benchmarks: Daily Update 2026-10-18</h3><div class="meta">49 official benchmarks, 1,102 entries. Baseline snapshot of every #1.</div></div><div class="d">2026-10-18</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/typed-decisions-new-number-one.html"><div><h3>Typed Decisions Leaderboard: A New #1 at 0.81</h3><div class="meta">A measurement note on the Typed Decisions official benchmark, which now has a new #1 at 0.81 holding both first and second place.</div></div><div class="d">2026-10-16</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/measuring-real-cost-llm-call.html"><div><h3>Measuring the Real Cost of an LLM Call: Tokens, Precision, and Honest Numbers</h3><div class="meta">How to measure the true cost of an LLM call beyond the sticker price per million tokens. Precision weighting, first-token latency, and why the same flags are not the same conditions.</div></div><div class="d">2026-10-12</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/quantum-challenge-simulation-and-qec.html"><div><h3>Quantum Simulation and QEC Decoding on a GPU: The Open Quantum Challenge</h3><div class="meta">Two measurable quantum tasks, simulation and QEC decoding, now run as an open challenge on classical GPUs with a public harness. 2,000 USD in prizes, reproducible scoring.</div></div><div class="d">2026-10-11</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/ai-native-the-evidence.html"><div><h3>The Evidence Behind an AI-Native Lab: What a Pre-AGI Organization Measures Up To</h3><div class="meta">Perfect AIME and HMMT, GPQA Diamond 94.44, 165M downloads, quantum hardware access, and a national-security safety system. The measured record behind an AI-native, pre-AGI lab.</div></div><div class="d">2026-10-11</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/eleven-number-one-benchmark-records.html"><div><h3>Eleven Number-One Records: Measuring a Model That Swept Math, Science, Law and Decisions</h3><div class="meta">One self-improving model family holds eleven public number-one benchmark records, from perfect AIME and HMMT scores to GPQA Diamond 94.44 and S1MB. Here are the measurements.</div></div><div class="d">2026-10-11</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/measuring-typed-decision-ai-s1mb.html"><div><h3>Measuring Typed Decision AI: What S1MB Actually Tests, and Who Leads It</h3><div class="meta">A 870M dollar round made typed, machine-native AI a headline. We look at how to measure it: what the S1MB benchmark tests, and the number one result among 102 models.</div></div><div class="d">2026-10-10</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/who-leads-hugging-face-official-benchmarks-concentration-gaps-hidden-entries.html"><div><h3>Who Leads Hugging Face&#x27;s Official Benchmarks? A Measurement of Concentration, Gaps and Hidden Entries</h3><div class="meta">We measured all 48 official Hugging Face benchmark leaderboards: Gini 0.71, top-10 orgs hold 49.5% of entries, 30.4% hidden by default.</div></div><div class="d">2026-10-05</div></a></li></ol></div></main>
30
  <footer><div class="w">漏 QuantID 路 <a href="https://www.quantid.co.kr">quantid.co.kr</a> 路 <a href="https://quantid-measurement-lab.static.hf.space/rss.xml">RSS</a> 路 <a href="https://quantid-measurement-lab.static.hf.space/sitemap.xml">Sitemap</a></div></footer></body></html>
 
26
  table{border-collapse:collapse;display:block;overflow-x:auto;font:14px 'IBM Plex Mono',monospace}td,th{border:1px solid var(--bd);padding:6px 10px}th{background:var(--card)}
27
  footer{border-top:2px solid var(--fg);margin-top:56px;padding:18px 0;font:12.5px 'IBM Plex Mono',monospace;color:var(--mut)}
28
  @media(max-width:820px){.paper{grid-template-columns:1fr}.spec{position:static}.idx a{grid-template-columns:52px 1fr}.idx .d{display:none}}</style></head><body>
29
+ <header><div class="w"><div class="mast"><a href="https://quantid-measurement-lab.static.hf.space/">QuantID<i>/</i>Measurement Lab</a><p>Measuring AI, science and risk with reproducible methods</p></div></div></header><main><div class="w"><p class="abs">ABSTRACT. Measurement-first analysis of AI benchmarks, scientific computing and AI risk diagnostics, with open methods and live data.</p><ol class="idx"><li><a href="https://quantid-measurement-lab.static.hf.space/a/borda-rank-aggregation-why-it-is-hard-to-game-2026-10-20.html"><div><h3>Why a Borda Score Is Harder to Game Than an Average</h3><div class="meta">S1MB ranks 197 decision models by aggregating ranks over 137 benchmarks instead of averaging scores. Here is what that changes, and how to read the resulting table.</div></div><div class="d">2026-10-20</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-20.html"><div><h3>Hugging Face Official Benchmarks: Daily Update 2026-10-20</h3><div class="meta">49 official benchmarks, 1,107 entries. 1 new #1, 0 new boards, +5 entries.</div></div><div class="d">2026-10-20</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-18.html"><div><h3>Hugging Face Official Benchmarks: Daily Update 2026-10-18</h3><div class="meta">49 official benchmarks, 1,102 entries. Baseline snapshot of every #1.</div></div><div class="d">2026-10-18</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/typed-decisions-new-number-one.html"><div><h3>Typed Decisions Leaderboard: A New #1 at 0.81</h3><div class="meta">A measurement note on the Typed Decisions official benchmark, which now has a new #1 at 0.81 holding both first and second place.</div></div><div class="d">2026-10-16</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/measuring-real-cost-llm-call.html"><div><h3>Measuring the Real Cost of an LLM Call: Tokens, Precision, and Honest Numbers</h3><div class="meta">How to measure the true cost of an LLM call beyond the sticker price per million tokens. Precision weighting, first-token latency, and why the same flags are not the same conditions.</div></div><div class="d">2026-10-12</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/quantum-challenge-simulation-and-qec.html"><div><h3>Quantum Simulation and QEC Decoding on a GPU: The Open Quantum Challenge</h3><div class="meta">Two measurable quantum tasks, simulation and QEC decoding, now run as an open challenge on classical GPUs with a public harness. 2,000 USD in prizes, reproducible scoring.</div></div><div class="d">2026-10-11</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/ai-native-the-evidence.html"><div><h3>The Evidence Behind an AI-Native Lab: What a Pre-AGI Organization Measures Up To</h3><div class="meta">Perfect AIME and HMMT, GPQA Diamond 94.44, 165M downloads, quantum hardware access, and a national-security safety system. The measured record behind an AI-native, pre-AGI lab.</div></div><div class="d">2026-10-11</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/eleven-number-one-benchmark-records.html"><div><h3>Eleven Number-One Records: Measuring a Model That Swept Math, Science, Law and Decisions</h3><div class="meta">One self-improving model family holds eleven public number-one benchmark records, from perfect AIME and HMMT scores to GPQA Diamond 94.44 and S1MB. Here are the measurements.</div></div><div class="d">2026-10-11</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/measuring-typed-decision-ai-s1mb.html"><div><h3>Measuring Typed Decision AI: What S1MB Actually Tests, and Who Leads It</h3><div class="meta">A 870M dollar round made typed, machine-native AI a headline. We look at how to measure it: what the S1MB benchmark tests, and the number one result among 102 models.</div></div><div class="d">2026-10-10</div></a></li><li><a href="https://quantid-measurement-lab.static.hf.space/a/who-leads-hugging-face-official-benchmarks-concentration-gaps-hidden-entries.html"><div><h3>Who Leads Hugging Face&#x27;s Official Benchmarks? A Measurement of Concentration, Gaps and Hidden Entries</h3><div class="meta">We measured all 48 official Hugging Face benchmark leaderboards: Gini 0.71, top-10 orgs hold 49.5% of entries, 30.4% hidden by default.</div></div><div class="d">2026-10-05</div></a></li></ol></div></main>
30
  <footer><div class="w">漏 QuantID 路 <a href="https://www.quantid.co.kr">quantid.co.kr</a> 路 <a href="https://quantid-measurement-lab.static.hf.space/rss.xml">RSS</a> 路 <a href="https://quantid-measurement-lab.static.hf.space/sitemap.xml">Sitemap</a></div></footer></body></html>
rss.xml CHANGED
@@ -1 +1 @@
1
- <?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>QuantID Measurement Lab</title><link>https://quantid-measurement-lab.static.hf.space/</link><description>Measurement-first analysis of AI benchmarks, scientific computing and AI risk diagnostics, with open methods and live data.</description><item><title>Hugging Face Official Benchmarks: Daily Update 2026-10-20</title><link>https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-20.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-20.html</guid><description>49 official benchmarks, 1,107 entries. 1 new #1, 0 new boards, +5 entries.</description><pubDate>2026-10-20T00:00:23+09:00</pubDate></item><item><title>Hugging Face Official Benchmarks: Daily Update 2026-10-18</title><link>https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-18.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-18.html</guid><description>49 official benchmarks, 1,102 entries. Baseline snapshot of every #1.</description><pubDate>2026-10-18T12:18:00+09:00</pubDate></item><item><title>Typed Decisions Leaderboard: A New #1 at 0.81</title><link>https://quantid-measurement-lab.static.hf.space/a/typed-decisions-new-number-one.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/typed-decisions-new-number-one.html</guid><description>A measurement note on the Typed Decisions official benchmark, which now has a new #1 at 0.81 holding both first and second place.</description><pubDate>2026-10-16T09:20:00+09:00</pubDate></item><item><title>Measuring the Real Cost of an LLM Call: Tokens, Precision, and Honest Numbers</title><link>https://quantid-measurement-lab.static.hf.space/a/measuring-real-cost-llm-call.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/measuring-real-cost-llm-call.html</guid><description>How to measure the true cost of an LLM call beyond the sticker price per million tokens. Precision weighting, first-token latency, and why the same flags are not the same conditions.</description><pubDate>2026-10-12T20:10:00+09:00</pubDate></item><item><title>Quantum Simulation and QEC Decoding on a GPU: The Open Quantum Challenge</title><link>https://quantid-measurement-lab.static.hf.space/a/quantum-challenge-simulation-and-qec.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/quantum-challenge-simulation-and-qec.html</guid><description>Two measurable quantum tasks, simulation and QEC decoding, now run as an open challenge on classical GPUs with a public harness. 2,000 USD in prizes, reproducible scoring.</description><pubDate>2026-10-11T00:20:00+09:00</pubDate></item><item><title>The Evidence Behind an AI-Native Lab: What a Pre-AGI Organization Measures Up To</title><link>https://quantid-measurement-lab.static.hf.space/a/ai-native-the-evidence.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/ai-native-the-evidence.html</guid><description>Perfect AIME and HMMT, GPQA Diamond 94.44, 165M downloads, quantum hardware access, and a national-security safety system. The measured record behind an AI-native, pre-AGI lab.</description><pubDate>2026-10-11T00:10:00+09:00</pubDate></item><item><title>Eleven Number-One Records: Measuring a Model That Swept Math, Science, Law and Decisions</title><link>https://quantid-measurement-lab.static.hf.space/a/eleven-number-one-benchmark-records.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/eleven-number-one-benchmark-records.html</guid><description>One self-improving model family holds eleven public number-one benchmark records, from perfect AIME and HMMT scores to GPQA Diamond 94.44 and S1MB. Here are the measurements.</description><pubDate>2026-10-11T00:00:00+09:00</pubDate></item><item><title>Measuring Typed Decision AI: What S1MB Actually Tests, and Who Leads It</title><link>https://quantid-measurement-lab.static.hf.space/a/measuring-typed-decision-ai-s1mb.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/measuring-typed-decision-ai-s1mb.html</guid><description>A 870M dollar round made typed, machine-native AI a headline. We look at how to measure it: what the S1MB benchmark tests, and the number one result among 102 models.</description><pubDate>2026-10-10T00:00:00+09:00</pubDate></item><item><title>Who Leads Hugging Face&#x27;s Official Benchmarks? A Measurement of Concentration, Gaps and Hidden Entries</title><link>https://quantid-measurement-lab.static.hf.space/a/who-leads-hugging-face-official-benchmarks-concentration-gaps-hidden-entries.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/who-leads-hugging-face-official-benchmarks-concentration-gaps-hidden-entries.html</guid><description>We measured all 48 official Hugging Face benchmark leaderboards: Gini 0.71, top-10 orgs hold 49.5% of entries, 30.4% hidden by default.</description><pubDate>2026-10-05T18:00:00+09:00</pubDate></item></channel></rss>
 
1
+ <?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>QuantID Measurement Lab</title><link>https://quantid-measurement-lab.static.hf.space/</link><description>Measurement-first analysis of AI benchmarks, scientific computing and AI risk diagnostics, with open methods and live data.</description><item><title>Why a Borda Score Is Harder to Game Than an Average</title><link>https://quantid-measurement-lab.static.hf.space/a/borda-rank-aggregation-why-it-is-hard-to-game-2026-10-20.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/borda-rank-aggregation-why-it-is-hard-to-game-2026-10-20.html</guid><description>S1MB ranks 197 decision models by aggregating ranks over 137 benchmarks instead of averaging scores. Here is what that changes, and how to read the resulting table.</description><pubDate>2026-10-20T11:20:00+09:00</pubDate></item><item><title>Hugging Face Official Benchmarks: Daily Update 2026-10-20</title><link>https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-20.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-20.html</guid><description>49 official benchmarks, 1,107 entries. 1 new #1, 0 new boards, +5 entries.</description><pubDate>2026-10-20T00:00:23+09:00</pubDate></item><item><title>Hugging Face Official Benchmarks: Daily Update 2026-10-18</title><link>https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-18.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-18.html</guid><description>49 official benchmarks, 1,102 entries. Baseline snapshot of every #1.</description><pubDate>2026-10-18T12:18:00+09:00</pubDate></item><item><title>Typed Decisions Leaderboard: A New #1 at 0.81</title><link>https://quantid-measurement-lab.static.hf.space/a/typed-decisions-new-number-one.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/typed-decisions-new-number-one.html</guid><description>A measurement note on the Typed Decisions official benchmark, which now has a new #1 at 0.81 holding both first and second place.</description><pubDate>2026-10-16T09:20:00+09:00</pubDate></item><item><title>Measuring the Real Cost of an LLM Call: Tokens, Precision, and Honest Numbers</title><link>https://quantid-measurement-lab.static.hf.space/a/measuring-real-cost-llm-call.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/measuring-real-cost-llm-call.html</guid><description>How to measure the true cost of an LLM call beyond the sticker price per million tokens. Precision weighting, first-token latency, and why the same flags are not the same conditions.</description><pubDate>2026-10-12T20:10:00+09:00</pubDate></item><item><title>Quantum Simulation and QEC Decoding on a GPU: The Open Quantum Challenge</title><link>https://quantid-measurement-lab.static.hf.space/a/quantum-challenge-simulation-and-qec.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/quantum-challenge-simulation-and-qec.html</guid><description>Two measurable quantum tasks, simulation and QEC decoding, now run as an open challenge on classical GPUs with a public harness. 2,000 USD in prizes, reproducible scoring.</description><pubDate>2026-10-11T00:20:00+09:00</pubDate></item><item><title>The Evidence Behind an AI-Native Lab: What a Pre-AGI Organization Measures Up To</title><link>https://quantid-measurement-lab.static.hf.space/a/ai-native-the-evidence.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/ai-native-the-evidence.html</guid><description>Perfect AIME and HMMT, GPQA Diamond 94.44, 165M downloads, quantum hardware access, and a national-security safety system. The measured record behind an AI-native, pre-AGI lab.</description><pubDate>2026-10-11T00:10:00+09:00</pubDate></item><item><title>Eleven Number-One Records: Measuring a Model That Swept Math, Science, Law and Decisions</title><link>https://quantid-measurement-lab.static.hf.space/a/eleven-number-one-benchmark-records.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/eleven-number-one-benchmark-records.html</guid><description>One self-improving model family holds eleven public number-one benchmark records, from perfect AIME and HMMT scores to GPQA Diamond 94.44 and S1MB. Here are the measurements.</description><pubDate>2026-10-11T00:00:00+09:00</pubDate></item><item><title>Measuring Typed Decision AI: What S1MB Actually Tests, and Who Leads It</title><link>https://quantid-measurement-lab.static.hf.space/a/measuring-typed-decision-ai-s1mb.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/measuring-typed-decision-ai-s1mb.html</guid><description>A 870M dollar round made typed, machine-native AI a headline. We look at how to measure it: what the S1MB benchmark tests, and the number one result among 102 models.</description><pubDate>2026-10-10T00:00:00+09:00</pubDate></item><item><title>Who Leads Hugging Face&#x27;s Official Benchmarks? A Measurement of Concentration, Gaps and Hidden Entries</title><link>https://quantid-measurement-lab.static.hf.space/a/who-leads-hugging-face-official-benchmarks-concentration-gaps-hidden-entries.html</link><guid>https://quantid-measurement-lab.static.hf.space/a/who-leads-hugging-face-official-benchmarks-concentration-gaps-hidden-entries.html</guid><description>We measured all 48 official Hugging Face benchmark leaderboards: Gini 0.71, top-10 orgs hold 49.5% of entries, 30.4% hidden by default.</description><pubDate>2026-10-05T18:00:00+09:00</pubDate></item></channel></rss>
sitemap.xml CHANGED
@@ -1 +1 @@
1
- <?xml version="1.0" encoding="UTF-8"?><urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9"><url><loc>https://quantid-measurement-lab.static.hf.space/</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-20.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-18.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/typed-decisions-new-number-one.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/measuring-real-cost-llm-call.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/quantum-challenge-simulation-and-qec.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/ai-native-the-evidence.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/eleven-number-one-benchmark-records.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/measuring-typed-decision-ai-s1mb.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/who-leads-hugging-face-official-benchmarks-concentration-gaps-hidden-entries.html</loc></url></urlset>
 
1
+ <?xml version="1.0" encoding="UTF-8"?><urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9"><url><loc>https://quantid-measurement-lab.static.hf.space/</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/borda-rank-aggregation-why-it-is-hard-to-game-2026-10-20.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-20.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/hf-official-benchmarks-daily-2026-10-18.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/typed-decisions-new-number-one.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/measuring-real-cost-llm-call.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/quantum-challenge-simulation-and-qec.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/ai-native-the-evidence.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/eleven-number-one-benchmark-records.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/measuring-typed-decision-ai-s1mb.html</loc></url><url><loc>https://quantid-measurement-lab.static.hf.space/a/who-leads-hugging-face-official-benchmarks-concentration-gaps-hidden-entries.html</loc></url></urlset>