analiza aiqbench
The benchmark tests whether frontier LLMs can screen research proposals for methodological soundness before resources are spent. Across 12 models, standard prompting produced a 74.0% mean false-positive rate, and aggressive prompting cut false approvals only by raising false rejections. For readers building autonomous research agents, this suggests current LLMs are not yet reliable as standalone gatekeepers for scientific rigor, and that human or verified review remains necessary.
SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones?
arXiv:2605.30329v2 Announce Type: replace Abstract: Autonomous AI research agents aim to speed up scientific discovery by automating the research pipeline from hypothesis generation to peer review. However, existing benchmarks rarely test a fundamental bottleneck whether LLMs can judge the methodological viability of a research idea $\textbf{before}$ investing time and computational resources. Without a reliable ``first-gate'' filter, autonomous agents risk scaling flawed methodology rather than accelerating meaningful science. We introduce $\textbf{SoundnessBench}$, a curated benchmark of 1,099 real research proposals, with source verified extraction, reviewer-derived proxy labels, and proposal-only expert judgments on a subset of 327 proposals. Across 12 frontier LLMs, our evaluation reveals a pervasive $\textbf{optimism bias}$: standard prompting yields a $74.0\%$ mean false-positive rate. Aggressive prompting reduces false approvals but substantially increases false rejections. Additional analyses show that optimism persists under controls for public corpus contamination, and surface features, as well as in literature-grounded agent evaluation. These results suggest a limitation in judging scientific soundness, indicating that current LLMs are not yet reliable as standalone gatekeepers for scientific rigor.
Title:SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones?
Abstract:Autonomous AI research agents aim to speed up scientific discovery by automating the research pipeline from hypothesis generation to peer review. However, existing benchmarks rarely test a fundamental bottleneck whether LLMs can judge the methodological viability of a research idea $\textbf{before}$ investing time and computational resources. Without a reliable ``first-gate'' filter, autonomous agents risk scaling flawed methodology rather than accelerating meaningful science. We introduce $\textbf{SoundnessBench}$, a curated benchmark of 1,099 real research proposals, with source verified extraction, reviewer-derived proxy labels, and proposal-only expert judgments on a subset of 327 proposals. Across 12 frontier LLMs, our evaluation reveals a pervasive $\textbf{optimism bias}$: standard prompting yields a $74.0\%$ mean false-positive rate. Aggressive prompting reduces false approvals but substantially increases false rejections. Additional analyses show that optimism persists under controls for public corpus contamination, and surface features, as well as in literature-grounded agent evaluation. These results suggest a limitation in judging scientific soundness, indicating that current LLMs are not yet reliable as standalone gatekeepers for scientific rigor.
Machine Learning (cs.LG)
arXiv:2605.30329 [cs.LG]
(or arXiv:2605.30329v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2605.30329 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 269 cuvinte