SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones?

• URL: http://arxiv.org/abs/2605.30329v1
• Date: Thu, 28 May 2026 17:57:37 GMT
• Title: SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones?
• Abstract summary: We introduce SoundnessBench, a curated benchmark of 1,099 machine-learning research proposals reconstructed from ICLR submissions.<n>SoundnessBench should be interpreted as a benchmark for recoverable proposal-stage soundness rather than exact prediction of full-paper review outcomes.
• Score: 51.154921661608675
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Autonomous AI research agents aim to accelerate scientific discovery by automating the research pipeline, from hypothesis generation to peer review. However, existing benchmarks rarely test a fundamental bottleneck: whether Large Language Models can judge the methodological viability of a research idea before expending time and computational resources. We introduce SoundnessBench, a curated benchmark of 1,099 machine-learning research proposals reconstructed from ICLR submissions, labeled with reviewer soundness sub-scores, and audited against source papers. SoundnessBench should be interpreted as a benchmark for recoverable proposal-stage soundness rather than exact prediction of full-paper review outcomes. Across 12 frontier LLMs, we find a pervasive optimism bias: under standard prompting, models frequently rate low-soundness proposals as sound, while aggressive prompting largely shifts errors from false positives to false negatives. Additional controls for public-corpus contamination, paper-identifying phrases, surface features, and human audit quality suggest that this behavior is not explained by a single confounder. Our results indicate that current LLMs are not yet reliable as standalone first-gate evaluators for scientific rigor.

Related papers

• NovGauge: A Fine-Grained Benchmark for Diagnosing LLMs' Capability in Paper Novelty Assessment [54.4265627247104]
  Large language models (LLMs) are increasingly used in peer review at major AI conferences.<n>Existing benchmarks assess novelty as a single holistic score.<n>We present NovGauge, a human-anchored benchmark for fine-grained novelty assessment diagnosis.
  arXiv  Detail & Related papers  (2026-09-10T08:34:56Z)
• AgentIdeaBench: Benchmarking Scientific Ideation in the Agent Era [55.37558486468993]
  We introduce AgentIdeaBench, a benchmark that evaluates scientific ideation under two matched settings, static observation and active exploration.<n>We report matched Static-Active evaluations for 33 LLMs across 40 densely scored subfields spanning five disciplines.<n>Active exploration reveals considerably more capability headroom, and that headroom is unevenly distributed across models.
  arXiv  Detail & Related papers  (2026-09-07T15:19:13Z)
• Understanding and Mitigating Spurious Signal Amplification in Test-Time Reinforcement Learning for Math Reasoning [37.981949917083746]
  Test-time reinforcement learning (TTRL) always adapts models at inference time via pseudo-labeling.<n>We propose a unified framework, Debiased and Denoised test-time Reinforcement Learning, to mitigate spurious signals.
  arXiv  Detail & Related papers  (2026-04-23T06:32:08Z)
• Learning to Predict Future-Aligned Research Proposals with Language Models [59.79457676644722]
  We build a time-consistent dataset of 17,771 papers from targets and their pre-cutoff citations.<n>To train models, we build a time-consistent dataset of 17,771 papers from targets and their pre-cutoff citations.<n>Across Llama-3.1 and Qwen2.5 models, future-aligned tuning improves future alignment over unaligned baselines.
  arXiv  Detail & Related papers  (2026-03-28T05:41:15Z)
• Proof of Time: A Benchmark for Evaluating Scientific Idea Judgments [12.46263502996029]
  PoT is a semi-verifiable benchmarking framework for large language models.<n>It links scientific idea judgments to downstream signals that become observable later.<n>PoT supports scalable evaluation of agents on future-facing scientific idea judgment tasks.
  arXiv  Detail & Related papers  (2026-01-12T14:55:37Z)
• Pre-review to Peer review: Pitfalls of Automating Reviews using Large Language Models [1.8349858105838042]
  Large Language Models are versatile general-task solvers, and their capabilities can truly assist people with scholarly peer review as textitpre-review agents.<n>While incredibly beneficial, automating academic peer-review, as a concept, raises concerns surrounding safety, research integrity, and the validity of the academic peer-review process.
  arXiv  Detail & Related papers  (2025-12-14T09:56:07Z)
• BadScientist: Can a Research Agent Write Convincing but Unsound Papers that Fool LLM Reviewers? [21.78901120638025]
  We investigate whether fabrication-oriented paper generation agents can deceive multi-model LLM review systems.<n>Our generator employs presentation-manipulation strategies requiring no real experiments.<n>Despite provably sound aggregation mathematics, integrity checking systematically fails.
  arXiv  Detail & Related papers  (2025-10-20T18:37:11Z)
• GEAR: A General Evaluation Framework for Abductive Reasoning [21.08814504507274]
  GEAR (General Evaluation for Abductive Reasoning) is a general-purpose, fully automated, transparent, and label-free evaluation paradigm.<n>GEAR scores hypothesis sets by three metrics: consistency (each hypothesis explains the observations), generalizability (consistent hypotheses make meaningful predictions on unseen inputs), and diversity (the set covers distinct predictions and patterns).
  arXiv  Detail & Related papers  (2025-09-28T22:22:28Z)
• Automatic Evaluation Metrics for Artificially Generated Scientific Research [3.9845810840390743]
  We investigate two automatic evaluation metrics, specifically citation count prediction and review score prediction.<n>Our findings reveal that citation count prediction is more viable than review score prediction, and predicting scores is more difficult purely from the research hypothesis than from the full paper.
  arXiv  Detail & Related papers  (2025-02-14T14:56:14Z)
• HyPoradise: An Open Baseline for Generative Speech Recognition with Large Language Models [81.56455625624041]
  We introduce the first open-source benchmark to utilize external large language models (LLMs) for ASR error correction. The proposed benchmark contains a novel dataset, HyPoradise (HP), encompassing more than 334,000 pairs of N-best hypotheses. LLMs with reasonable prompt and its generative capability can even correct those tokens that are missing in N-best list.
  arXiv  Detail & Related papers  (2023-09-27T14:44:10Z)
• Neural Importance Sampling for Rapid and Reliable Gravitational-Wave Inference [59.040209568168436]
  We first generate a rapid proposal for the Bayesian posterior using neural networks, and then attach importance weights based on the underlying likelihood and prior. This provides (1) a corrected posterior free from network inaccuracies, (2) a performance diagnostic (the sample efficiency) for assessing the proposal and identifying failure cases, and (3) an unbiased estimate of the Bayesian evidence. We carry out a large study analyzing 42 binary black hole mergers observed by LIGO and Virgo with the SEOBNRv4PHM and IMRPhenomHMXP waveform models.
  arXiv  Detail & Related papers  (2022-10-11T18:00:02Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.