Skip to content

Commit 7da65aa

Browse files
committed
add paper
1 parent b55eddf commit 7da65aa

1 file changed

Lines changed: 41 additions & 0 deletions

File tree

works.html

Lines changed: 41 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -70,6 +70,26 @@
7070
"https://huggingface.co/kadubon"
7171
],
7272
"publication": [
73+
{
74+
"@type": "ScholarlyArticle",
75+
"name": "AI Benchmark Half-Life in Recursive Corpora: A Theory of Validity Decay under Semantic Leakage and Regeneration",
76+
"genre": "Preprint",
77+
"url": "https://doi.org/10.5281/zenodo.18954286",
78+
"datePublished": "2026-03-11",
79+
"author": [
80+
{
81+
"@type": "Person",
82+
"name": "K Takahashi"
83+
}
84+
],
85+
"isPartOf": {
86+
"@type": "Periodical",
87+
"name": "Zenodo"
88+
},
89+
"citation": "K Takahashi (2026-03-11). AI Benchmark Half-Life in Recursive Corpora: A Theory of Validity Decay under Semantic Leakage and Regeneration. Zenodo. https://doi.org/10.5281/zenodo.18954286",
90+
"abstract": "This preprint develops a theory of AI benchmark half-life in recursive corpora under semantic leakage and regeneration, yielding validity-decay bounds and monitoring rules for evaluation systems whose items and solution traces re-enter public data. It models benchmark validity through discriminative power and construct validity, and derives jump-aware lifetime bounds, partial-identification results, portfolio design criteria, and safe sequential control under ambiguity and partial observability.",
91+
"keywords": "AI benchmark half-life, recursive corpora, semantic leakage, validity decay, benchmark contamination, construct validity, discriminative power, dynamic benchmarks, partial identification, sequential monitoring, lineage observability, model metrology"
92+
},
7393
{
7494
"@type": "ScholarlyArticle",
7595
"name": "When Should Inference Be Split? A Fixed-Budget Theory of Predictable Multi-Agent Advantage under Local Context Ceilings",
@@ -3737,6 +3757,27 @@ <h1>
37373757
Machine-readable endpoints: <a href="https://kadubon.github.io/github.io/feed.xml">RSS feed</a>, <a href="https://kadubon.github.io/github.io/CITATION.cff">CITATION.cff</a>, <a href="https://kadubon.github.io/github.io/robots.txt">robots.txt</a>, and <a href="https://kadubon.github.io/github.io/sitemap.xml">sitemap.xml</a>.
37383758
</p>
37393759
<ol class="publication-list">
3760+
<li>
3761+
<div class="publication-item">
3762+
<h3>
3763+
AI Benchmark Half-Life in Recursive Corpora: A Theory of Validity Decay under Semantic Leakage and Regeneration
3764+
</h3>
3765+
<p>
3766+
<span class="publication-meta">
3767+
Preprint | Published: 2026-03-11
3768+
</span>
3769+
<a href="https://doi.org/10.5281/zenodo.18954286" target="_blank">
3770+
DOI: 10.5281/zenodo.18954286
3771+
</a>
3772+
</p>
3773+
<p class="publication-abstract">
3774+
Abstract: This preprint develops a theory of AI benchmark half-life in recursive corpora under semantic leakage and regeneration, yielding validity-decay bounds and monitoring rules for evaluation systems whose items and solution traces re-enter public data. It models benchmark validity through discriminative power and construct validity, and derives jump-aware lifetime bounds, partial-identification results, portfolio design criteria, and safe sequential control under ambiguity and partial observability.
3775+
</p>
3776+
<p class="publication-keywords">
3777+
Keywords: AI benchmark half-life, recursive corpora, semantic leakage, validity decay, benchmark contamination, construct validity, discriminative power, dynamic benchmarks, partial identification, sequential monitoring, lineage observability, model metrology
3778+
</p>
3779+
</div>
3780+
</li>
37403781
<li>
37413782
<div class="publication-item">
37423783
<h3>

0 commit comments

Comments
 (0)