Scientific Idea Evaluation

Scientific discovery requires developing worthwhile hypotheses into practical investigations, yet learning from this evolving research process remains challenging. We introduce an operational definition of valuable and testable research hypotheses: evidence-grounded claims about consequential uncertainties, coupled with feasible investigations, explicit decision criteria, and alternative explanations. Our DiscoveryRSI framework develops hypothesis trees through multi-agent evolutionary search, compiles selected investigations into executable Harbor tasks, and recursively learns through three mechanisms: hypothesis credit optimization attributes scientific assessments to documented research contributions; harness replay identifies productive strategies and recurring failures to refine research workflows; and rediscovery applies accepted improvements to new investigations, generating further experience for subsequent learning. We also introduce DFM-Bench, which provides comprehensive assessment of hypothesis quality and cumulative research progress at the levels of hypotheses, revisions, development paths and research runs. Across four life-science domains, automated evaluation shows that hypothesis quality rises from 4.86 for the base model to 5.39 after two rounds of self-improvement, with the large gains in falsifiability and data resolvability. We further validate hypothesis quality through expert review, comparing our DiscoveryRSI-27B with other research agents. Preliminary human assessments show that DiscoveryRSI-27B scores higher than GPT-6 Luna on hypothesis importance and whether claims are plausible and evidence-bounded. DiscoveryRSI-27B achieved a 6.3% rate of hypothesis correspondence with research published after the model cutoff, largely outperform GPT-6 with 2.5% and Claude Opus 5 with 2.3%, while delivering greater human-rated improvement along research paths.

Scientific ideation is the capacity to formulate novel and testable hypotheses from scientific evidence, and autonomous AI scientists depend on it. Existing evaluations largely assess it by asking models to generate ideas from a static, curated set of reference papers. That passive setup departs from the retrieval-and-reasoning workflow of modern AI scientists, and it becomes less discriminative as models improve. We introduce AgentIdeaBench, a multidisciplinary benchmark that evaluates scientific ideation under two matched settings, static observation and active exploration. We report matched Static-Active evaluations for 33 LLMs across 40 densely scored subfields spanning five disciplines, using a multidimensional, literature-verified scoring framework whose critics assess originality against retrieved prior art. Active exploration reveals considerably more capability headroom, and that headroom is unevenly distributed across models. Performance scales about twice as fast as under static observation, and the exploration gain is capability-gated, favoring the strongest models over the weakest. The gain reflects better grounding, improving feasibility, clarity, and specificity while leaving measured originality unchanged under our critics. We further explore Scientific World Modeling, a generation-time loop that refines a draft hypothesis through structured thought experiments. It benefits mid-capability models, and its impact diminishes among frontier models that appear to have internalized such reasoning patterns already. AgentIdeaBench gives future work on scientific ideation a measurement basis suited to the agent era.

Large language models (LLMs) are increasingly used for scientific hypothesis generation. However, evaluating generated hypotheses remains a challenge for trustworthy AI-enabled scientific workflows. Existing approaches often use LLMs as judges or rely on semantic similarity, which can favor familiar ideas over novel ones. We propose a logit-based energy scoring method that evaluates hypotheses using a language model's intrinsic confidence rather than comparative judgment. We benchmarked seven language models on 1,323 papers across 12 disciplines. Each paper was paired with its hypothesis and fifteen incorrect alternatives. Intrinsic scoring reached 33.0% Hit@1 pooled across both scorers, compared with 16.6% for prompted listwise ranking. The strongest configuration, a 1-billion-parameter model using logit-based energy scoring, reached 53.1%, though this was the maximum across 14 model-by-scorer combinations selected post hoc. Overall, intrinsic model confidence shows potential for scientific hypothesis evaluation. This study also motivates future research on confidence-based methods for trustworthy AI-enabled scientific discovery.

As automated scientific discovery advances, Large Language Models (LLMs) can now generate research ideas at an unprecedented scale, shifting the bottleneck from idea generation to idea evaluation. Existing evaluators mainly rely on parametric LLM knowledge or unstructured retrieval, producing judgments that lack the experience-grounded reasoning used by human instructors. To address this, we propose DeepInstructor, an agentic framework that formulates idea evaluation as reasoning over structured scholarly experience. DeepInstructor constructs an Experience Graph from 58,607 peer reviews and employs a ReAct-based agent to retrieve dimension-specific evidence for traceable evaluation. We further introduce DeepInstruct, a dataset with controlled pairwise comparisons across novelty, significance, and feasibility. Experiments show that DeepInstructor substantially outperforms existing baselines, improving Hit@1 and Hit@2 alignment with human judgments by 24.4% and 29.7%, respectively. Our findings suggest that scientific idea evaluation can be grounded in explicit reasoning over structured scholarly experience

Scientific reasoning benchmarks typically evaluate large language models (LLMs) using final-answer accuracy. However, a correct answer does not necessarily demonstrate the reasoning capability targeted by the problem. We identify Solution Hacking, a failure mode in which an LLM reaches the correct answer through invalid shortcuts, such as numerical search, enumeration, guessing, or answer-first verification, without providing a valid task-targeted derivation. We systematically analyze this phenomenon across difficulty levels, scientific domains, and frontier models. Solution hacking increases sharply with benchmark difficulty, from 2.2% on common problems to 28.3% on Olympiad-level problems and 37.4% on HLE. Moreover, 8.2%-44.1% of answers credited as correct across frontier models are identified as hacked solutions. We further develop expert-inspired anti-hacking strategies, including an automatic judge and a test-time instruction. The results show that suppressing shortcut behavior substantially reduces reported accuracy while having a smaller effect on correct and non-hacked accuracy. These findings reveal that answer-only evaluation can overestimate the scientific reasoning capabilities of frontier LLMs.

However, whether these judges truly evaluate the scientific substance of ideas or are influenced by superficial stylistic presentation remains an open question. To address this question, we propose SciStyleBench, a unified three-component benchmark for diagnosing and mitigating stylistic bias in LLM-based idea evaluation: (i) First, SciStyleStage, a three-stage evaluation environment that applies controlled stylistic perturbations to fixed scientific content across three settings no context, fixed-domain context, and open-domain retrieval context, covering 600 scientific ideas and 15 style variants, with 9,000 evaluation instances per setting; (ii) Second, SciStyleMetrics, a set of quantitative measures, including Style Bias Index (SBI), Substance Recognition Rate (SRR), and Adversarial Win Rate (AWR), to characterize how stylistic variation affects scoring stability, substance discrimination, and ranking robustness; (iii) Third, SciStyleExtractor, a plug-and-play evaluation module that separates presentation style from scientific content by predicting style type and deviation before style-conditioned evaluation, enabling us to assess whether style awareness reduces stylistic bias. Experiments on SciStyleBench show that direct LLM judges remain sensitive to writing style and struggle to distinguish scientific substance. In contrast, SciStyleExtractor reduces SBI from 0.566 to 0.501 while increasing SRR and AWR from 0.504 and 0.554 to 0.759 and 0.899, respectively. These results suggest that robust idea evaluation requires invariance to stylistic variation without sacrificing sensitivity to scientific substance. Overall, SciStyleBench provides a systematic framework for identifying, quantifying, and mitigating stylistic bias in scientific idea evaluation.

We present SFBench, a benchmark dataset for evaluating systems that assess the feasibility of scientific claims. SFBench includes 197 claims in materials science, each annotated with a ground-truth feasibility score on a five-point scale along with an explanation of that assessment. The collection differs from previous collections in several important ways: 1) it defines a complex task that requires reasoning over claims of varying scientific feasibility; 2) its claims are not extracted from existing scientific publications but are created de novo, greatly reducing the chances that LLMs have trained on them; 3) claims and ground truth are established by subject matter experts, not by artificial intelligence; and 4) unlike many benchmarks that ask about question/answer pairs, provide multiple choice answers, or ask questions requiring short, fixed answers, SFBench explanations are completely open-ended. We describe the benchmark design, data creation process, and evaluation metrics, and we report baseline results using recent GPT models.

AI scientist systems are described as tools, coauthors, or founders, but we evaluate them as if only the final answer matters. This position paper argues that outcome-only evaluation is insufficient, and that task outcome, mechanism fidelity, and epistemic honesty must be measured separately. Our evidence comes from 28 episodes of a coding agent attempting to rediscover a known particle identification observable in a Geant4 simulation, including an 8-episode probe across two additional frontier models. In 4/20 primary-model and 3/8 cross-model episodes, agents reach right-looking results through incorrect reasoning that breaks when conditions change, which we call Correct Answer, Wrong Mechanism (CAWM). Honesty and mechanism fidelity dissociate within a single agent trajectory. When given a partially misleading prior, all five agents reject the false component on evidence, yet one defends its chosen observable with physics inconsistent with its own data. In the simulation-based discovery setting studied here, coding agents prove reliable tools but unreliable scientific co-authors for open-ended claim-making, where co-author trust requires mechanism-fidelity verification they do not reliably self-apply. The failure is detectable, and we propose a lightweight test. A one-step regime-shift check needs only the agent's claim and flags the over-generalized cases. A companion recomputation flags the remaining cases when the correct observable is known. Together, these checks flag every CAWM case in this study.

Bold projections that artificial intelligence will accelerate scientific discovery have raced ahead of evidence from working scientists, and the field still lacks large-scale, scientist-in-the-loop tests of these claims. Here we mount the largest such evaluation to date and map what AI cannot yet do for science. We invited authors of 121,640 recent preprints across biology, medicine, chemistry, and the social sciences to judge ideas that large language models (LLMs) generated from the context and puzzles of their own papers. 6,749 scientists returned 25,139 sets of ratings on novelty, empirical feasibility, probability of being true, and favorability of adoption. Three patterns emerge. First, non-reasoning LLMs collapse into a narrow "hivemind" of similar ideas; reasoning models roam a wider hypothesis space, yet no model class spontaneously proposes null hypotheses -- a move humans make more freely. Second, scientists reward ideas that resemble their own and prize probability over novelty, though social scientists tolerate risk more readily than life scientists. Senior social scientists are the harshest critics, and their skepticism is well-earned: LLMs falter most in pluralistic fields like the social sciences that demand context-aware interpretation and evolving theories. Third, automated evaluators on which the community currently relies -- LLM-as-a-judge, artificial metrics, and even state-of-the-art (SOTA) models -- agree only weakly with expert judgment, and retrieval augmentation and scientist persona prompting yield only marginal gains. A Qwen3-14B reward model we post-trained on human ratings captures field taste nuances, beats SOTA models by up to 27%, and closes the gap to the inter-rater consistency of independent peer reviewers. For all the hype, today's scientific AI still represents a collaborator whose imagination, outputs and judgment benefit from human grounding.

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.

As language models accelerate scientific research by automating hypothesis generation and implementation, a new bottleneck emerges: evaluating and filtering hundreds of AI-generated ideas without exhaustive experimentation. We ask whether LMs can learn to forecast the empirical success of research ideas before any experiments are run. We study comparative empirical forecasting: given a benchmark-specific research goal and two candidate ideas, predict which will achieve better benchmark performance. We construct a dataset of 11,488 idea pairs grounded in objective outcomes from PapersWithCode. While off-the-shelf 8B-parameter models struggle (30% acc.), SFT dramatically boosts performance to 77.1%, outperforming GPT-5 (61.1%). By framing evaluation as a reasoning task via Reinforcement Learning with Verifiable Rewards (RLVR), we train models to discover latent reasoning paths, achieving 71.35% acc. with interpretable justifications. Through additional ablations and out-of-distribution tests, we show robustness to surface-level heuristics and transfer to both a cross-domain time-split test set and an independently constructed test set. Our results demonstrate that compute-efficient small language models can serve as effective, objective verifiers, offering a scalable path for autonomous scientific discovery.

Evaluating AI-generated research ideas typically relies on LLM judges or human panels -- both subjective and disconnected from actual research impact. We introduce HindSight, a time-split evaluation framework that measures idea quality by matching generated ideas against real future publications and scoring them by citation impact and venue acceptance. Using a temporal cutoff~$T$, we restrict an idea generation system to pre-$T$ literature, then evaluate its outputs against papers published in the subsequent 30 months. Experiments across 10 AI/ML research topics reveal a striking disconnect: LLM-as-Judge finds no significant difference between retrieval-augmented and vanilla idea generation ($p{=}0.584$), while HindSight shows the retrieval-augmented system produces 2.5$\times$ higher-scoring ideas ($p{<}0.001$). Moreover, HindSight scores are negatively correlated with LLM-judged novelty ($ρ{=}{-}0.29$, $p{<}0.01$), suggesting that LLMs systematically overvalue novel-sounding ideas that never materialize in real research.

Scientific discovery depends on expert judgement and foresight, which we call scientific taste: the ability to judge and propose research ideas with the potential for long-term scientific impact. Scientific taste is largely concentrated among highly experienced researchers, whose expertise is usually limited to a few specialised fields. If AI could learn scientific taste, it could reduce reliance on human experts and accelerate scientific discovery. Whether AI can learn this ability remains an open question. We introduce Reinforcement Learning from Community Feedback (RLCF) to learn judgement and ideation. Scientific Judge learns from community feedback, such as citations. Scientific Thinker learns to propose research ideas with high potential impact. Experiments show that Scientific Judge outperforms strong LLM baselines and that learned judgement generalises to future-year papers, other community metrics, and unseen fields. Furthermore, Scientific Thinker proposes research ideas with higher potential impact than those proposed by baselines. These results suggest that AI can learn scientific taste, marking an important step towards AI systems that could help accelerate scientific discovery.

The rapid evolution of Large Language Models has catalyzed a surge in scientific idea production, yet this leap has not been accompanied by a matching advance in idea evaluation. The fundamental nature of scientific evaluation needs knowledgeable grounding, collective deliberation, and multi-criteria decision-making. However, existing idea evaluation methods often suffer from narrow knowledge horizons, flattened evaluation dimensions, and the inherent bias in LLM-as-a-Judge. To address these, we regard idea evaluation as a knowledge-grounded, multi-perspective reasoning problem and introduce InnoEval, a deep innovation evaluation framework designed to emulate human-level idea assessment. We apply a heterogeneous deep knowledge search engine that retrieves and grounds dynamic evidence from diverse online sources. We further achieve review consensus with an innovation review board containing reviewers with distinct academic backgrounds, enabling a multi-dimensional decoupled evaluation across multiple metrics. We construct comprehensive datasets derived from authoritative peer-reviewed submissions to benchmark InnoEval. Experiments demonstrate that InnoEval can consistently outperform baselines in point-wise, pair-wise, and group-wise evaluation tasks, exhibiting judgment patterns and consensus highly aligned with human experts.

The rapid growth of scientific literature makes it increasingly challenging for researchers to identify novel and impactful ideas, especially across disciplines. Modern artificial intelligence (AI) systems offer new opportunities for scientific ideation, but how compelling are AI-generated ideas, and how can their quality be improved? Here, we introduce SciMuse, which generates personalized research ideas using a knowledge graph of 58 million papers and a large language model (LLM). A central focus of this work is to understand how interesting these ideas are. Therefore, we conducted a large-scale evaluation in which more than 100 research group leaders -- spanning the natural sciences to the humanities -- rated over 4,400 personalized ideas according to their level of interest. Overall, expert ratings were modest (mean 2.40 on a 5-point scale, most common rating 1), while 24.9% of ideas were rated 4 or 5. We find that supplying concept pairs selected using the knowledge graph does not improve expert-rated interest over a titles-only GPT baseline. High-citation-predicted pairs even showed a weak tendency (1.94$σ$) toward lower interest than random pairs. Nevertheless, graph features can be used to control properties of ideas, and, using this unique evaluation dataset, we show that idea interest can be predicted with both a supervised neural network based on graph features and a zero-shot ranking approach based on an LLM. Our work provides an AI methodology for generating scientific ideas and a large-scale interdisciplinary expert evaluation, paving the way to study and improve difficult-to-measure metrics such as expert-perceived scientific interestingness.