Metraipolis › archive › dc_537c8e5cd956

Are Benchmarks Reliable? Toward Structural Diagnosis via Sample-Level Capability Boundaries — arXiv:2609.33196

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.33196
License (as archived)
CC0-1.0
Archived
2026-10-07T20:56:31.693Z
SHA-256 of the archived text
c402341ccd5e2e80c30261c82d87cb249b006acdd28595fdf922ef8d21d91948
Stable URL of this page
https://metraipolis.com/documento/dc_537c8e5cd956

How to cite

Are Benchmarks Reliable? Toward Structural Diagnosis via Sample-Level Capability Boundaries — arXiv:2609.33196 — arXiv (en), https://arxiv.org/abs/2609.33196 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_537c8e5cd956 (Metraipolis archive dc_537c8e5cd956, sha256 c402341ccd5e2e80c30261c82d87cb249b006acdd28595fdf922ef8d21d91948)
BibTeX
@misc{dc537c8e5cd956,
  title = {Are Benchmarks Reliable? Toward Structural Diagnosis via Sample-Level Capability Boundaries — arXiv:2609.33196},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.33196},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_537c8e5cd956; sha256 c402341ccd5e2e80c30261c82d87cb249b006acdd28595fdf922ef8d21d91948},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.33196. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Evaluating large language models (LLMs) relies heavily on benchmark scores, yet aggregate metrics can obscure whether benchmark samples reliably support model comparison. We introduce \textbf{BSDProbe}, a sample-level framework for \emph{benchmark structural diagnosis} that estimates capability boundaries from repeated-response trajectories along ordered model axes. BSDProbe summarizes samples by boundary position, boundary width, boundary-signal validity, and order consistency, then aggregates them into benchmark-level structural profiles. Experiments on six benchmarks show that benchmark reliability is axis-conditioned and heterogeneous: GSM8K and MATH exhibit the most stable measurement structures, MMLU and TriviaQA are relatively stable but heterogeneous, while GPQA and PopQA show stronger axis-conditioned risks. These profiles remain consistent across Qwen3, Qwen2.5, and cross-model axes. BSDProbe further selects compact high-value subsets whose model discriminability reaches up to $8.58\times$ that of the full benchmark. These results suggest that reliable benchmark use requires examining sample-level capability boundaries beyond leaderboard scores.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.33196 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T20:29:01.174Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).