mlsanityData to 7 Oct 2026

Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researchers

arXiv 2409.04109 ↗First posted 6 Sep 2024cs.CL cs.AI cs.CYcited by 348

On the map: LLM agents and tool use → Web, search, and research agents

Abstract (arXiv)

Recent advancements in large language models (LLMs) have sparked optimism about their potential to accelerate scientific discovery, with a growing number of works proposing research agents that autonomously generate and validate new ideas. Despite this, no evaluations have shown that LLM systems can take the very first step of producing novel, expert-level ideas, let alone perform the entire research process. We address this by establishing an experimental design that evaluates research idea generation while controlling for confounders and performs the first head-to-head comparison between expert NLP researchers and an LLM ideation agent. By recruiting over 100 NLP researchers to write novel ideas and blind reviews of both LLM and human ideas, we obtain the first statistically significant conclusion on current LLM capabilities for research ideation: we find LLM-generated ideas are judged as more novel (p < 0.05) than human expert ideas while being judged slightly weaker on feasibility. Studying our agent baselines closely, we identify open problems in building and evaluating research agents, including failures of LLM self-evaluation and their lack of diversity in generation. Finally, we acknowledge that human judgements of novelty can be difficult, even by experts, and propose an end-to-end study design which recruits researchers to execute these ideas into full projects, enabling us to study whether these novelty and feasibility judgements result in meaningful differences in research outcome.

Key facts. Cited by 348 papers in the map (data through 7 Oct 2026). Placed in Web, search, and research agents, one of 16 directions in LLM agents and tool use. Builds on 35 of its 56 references that are in the map. First posted on arXiv 6 Sep 2024.

Builds on

35 of 56 references.

Closest ancestors · 7

Same direction, newest first

  1. 2024MLR-Copilot: Autonomous Machine Learning Research based on Large Language Models Agents26 Aug
  2. 2024BLADE: Benchmarking Language Model Agents for Data-Driven Science19 Aug
  3. 2024DiscoveryBench: Towards Data-Driven Discovery with Large Language Models1 Jul
  4. 2024ResearchArena: Benchmarking Large Language Models' Ability to Collect and Organize Information as Research Agents13 Jun
  5. 2024DS-Agent: Automated Data Science by Empowering Large Language Models with Case-Based Reasoning27 Feb
  6. 2024InfiAgent-DABench: Evaluating Agents on Data Analysis Tasks10 Jan
  7. 2024MARG: Multi-Agent Review Generation for Scientific Papers8 Jan

Other references · 28

  1. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,160
  2. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  3. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,317
  4. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Modelscited by 3,155
  5. 2022Least-to-Most Prompting Enables Complex Reasoning in Large Language Modelscited by 1,397
  6. 2022Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Taskscited by 1,057
  7. 2024The AI Scientist: Towards Fully Automated Open-Ended Scientific Discoverycited by 1,002
  8. 2024Large Language Monkeys: Scaling Inference Compute with Repeated Samplingcited by 946
  9. 2023REPLUG: Retrieval-Augmented Black-Box Language Modelscited by 786
  10. 2023Chain-of-Verification Reduces Hallucination in Large Language Modelscited by 317
  11. 2022Prompting GPT-3 To Be Reliablecited by 258
  12. 2024ResearchAgent: Iterative Research Idea Generation over Scientific Literature with Large Language Modelscited by 200
16 more
  1. 2023Does Writing with Language Models Reduce Content Diversity?cited by 190
  2. 2024SciCode: A Research Coding Benchmark Curated by Scientistscited by 172
  3. 2023Take a Step Back: Evoking Reasoning via Abstraction in Large Language Modelscited by 157
  4. 2023Active Prompting with Chain-of-Thought for Large Language Modelscited by 154
  5. 2024The Prompt Report: A Systematic Survey of Prompt Engineering Techniquescited by 115
  6. 2023Large Language Models as Analogical Reasonerscited by 107
  7. 2023System 2 Attention (is something you might need too)cited by 85
  8. 2023Goal Driven Discovery of Distributional Differences via Language Descriptionscited by 61
  9. 2022Describing Differences between Text Distributions with Natural Languagecited by 54
  10. 2024Autonomous LLM-driven research from data to human-verifiable research paperscited by 50
  11. 2024LitSearch: A Retrieval Benchmark for Scientific Literature Searchcited by 47
  12. 2024CiteME: Can Language Models Accurately Cite Scientific Claims?cited by 45
  13. 2023Has the Machine Learning Review Process Become More Arbitrary as the Field Has Grown? The NeurIPS 2021 Consistency Experimentcited by 31
  14. 2024How AI Ideas Affect the Creativity, Diversity, and Evolution of Human Ideas: Evidence From a Large, Dynamic Experimentcited by 22
  15. 2024MASSW: A New Dataset and Benchmark Tasks for AI-Assisted Scientific Workflowscited by 19
  16. 2024Shared Imagination: LLMs Hallucinate Alikecited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 41; 2025: 161; 2026: 146.

348 papers.

Sort:

By area:

  1. 2025AlphaEvolve: A coding agent for scientific and algorithmic discoverycited by 763
  2. 2025Agent Laboratory: Using LLM Agents as Research Assistantscited by 405
  3. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 381
  4. 2025PaperBench: Evaluating AI's Ability to Replicate AI Researchcited by 286
  5. 2025MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attentioncited by 248
  6. 2024ScienceAgentBench: Toward Rigorous Assessment of Language Agents for Data-Driven Scientific Discoverycited by 208
  7. 2025Survey on Evaluation of LLM-based Agentscited by 190
  8. 2024CycleResearcher: Improving Automated Research via Automated Reviewcited by 142
  9. 2025MLGym: A New Framework and Benchmark for Advancing AI Research Agentscited by 111
  10. 2025SWE-Lancer: Can Frontier LLMs Earn 1 Million from Real-World Freelance Software Engineering?cited by 99
  11. 2025DeepReview: Improving LLM-based Paper Review with Human-like Deep Thinking Processcited by 97
  12. 2024Chain of Ideas: Revolutionizing Research Via Novel Idea Development with LLM Agentscited by 94
38 more
  1. 2025AI-Researcher: Autonomous Scientific Innovationcited by 88
  2. 2025From Automation to Autonomy: A Survey on Large Language Models in Scientific Discoverycited by 86
  3. 2025Jointly Reinforcing Diversity and Quality in Language Model Generationscited by 79
  4. 2025From AI for Science to Agentic Science: A Survey on Autonomous Scientific Discoverycited by 74
  5. 2024MOOSE-Chem: Large Language Models for Rediscovering Unseen Chemistry Scientific Hypothesescited by 68
  6. 2025Paper2Code: Automating Code Generation from Scientific Papers in Machine Learningcited by 67
  7. 2024A Systematic Survey on Large Language Models for Algorithm Designcited by 65
  8. 2025AI Research Agents for Machine Learning: Search, Exploration, and Generalization in MLE-benchcited by 65
  9. 2025AgentRxiv: Towards Collaborative Autonomous Researchcited by 64
  10. 2024Many Heads Are Better Than One: Improved Scientific Idea Generation by A LLM-Based Multi-Agent Systemcited by 63
  11. 2024Benchmarking Distributional Alignment of Large Language Modelscited by 58
  12. 2025ResearchBench: Benchmarking LLMs in Scientific Discovery via Inspiration-Based Task Decompositioncited by 57
  13. 2025MLR-Bench: Evaluating AI Agents on Open-Ended Machine Learning Researchcited by 57
  14. 2024Nova: An Iterative Planning and Search Approach to Enhance Novelty and Diversity of LLM Generated Ideascited by 52
  15. 2025Towards Robust Agentic CUDA Kernel Benchmarking, Verification, and Optimizationcited by 52
  16. 2024Echoes in AI: Quantifying lack of plot diversity in LLM outputscited by 47
  17. 2025ScienceBoard: Evaluating Multimodal Autonomous Agents in Realistic Scientific Workflowscited by 47
  18. 2025ResearchQA: Evaluating Scholarly Question Answering at Scale Across 75 Fields with Survey-Mined Questions and Rubricscited by 44
  19. 2024Human Creativity in the Age of LLMs: Randomized Experiments on Divergent and Convergent Thinkingcited by 43
  20. 2025SciReplicate-Bench: Benchmarking LLMs in Agent-driven Algorithmic Reproduction from Research Paperscited by 42
  21. 2025Earth-Agent: Unlocking the Full Landscape of Earth Observation with Agentscited by 42
  22. 2024Towards Scientific Discovery with Generative AI: Progress, Opportunities, and Challengescited by 41
  23. 2025Transforming Science with Large Language Models: A Survey on AI-assisted Scientific Discovery, Experimentation, Content Generation, and Evaluationcited by 41
  24. 2026EvoScientist: Towards Multi-Agent Evolving AI Scientists for End-to-End Scientific Discoverycited by 41
  25. 2024IdeaSynth: Iterative Research Idea Development Through Evolving and Composing Idea Facets with Literature-Grounded Feedbackcited by 40
  26. 2024DrugAgent: Automating AI-aided Drug Discovery Programming through LLM Multi-Agent Collaborationcited by 40
  27. 2025Automated Hypothesis Validation with Agentic Sequential Falsificationscited by 40
  28. 2024Human-LLM Compound System for Scientific Ideation through Facet Recombination and Novelty Evaluationcited by 39
  29. 2024ResearchTown: Simulator of Human Research Communitycited by 35
  30. 2025HypoBench: Towards Systematic and Principled Benchmarking for Hypothesis Generationcited by 34
  31. 2025ResearchCodeBench: Benchmarking LLMs on Implementing Novel Machine Learning Research Codecited by 34
  32. 2025Theoretical Physics Benchmark (TPBench) -- a Dataset and Study of AI Reasoning Capabilities in Theoretical Physicscited by 33
  33. 2025Evaluating Sakana's AI Scientist: Bold Claims, Mixed Results, and a Promising Future?cited by 32
  34. 2026Scientific production in the era of Large Language Modelscited by 30
  35. 2025AutoReproduce: Automatic AI Experiment Reproduction with Paper Lineagecited by 29
  36. 2025LMR-BENCH: Evaluating LLM Agent's Ability on Reproducing Language Modeling Researchcited by 29
  37. 2025Roll the dice & look before you leap: Going beyond the creative limits of next-token predictioncited by 28
  38. 2025MLRC-Bench: Can Language Agents Solve Machine Learning Research Challenges?cited by 27

50 most cited of 348 All 348 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026Thinking Like a Scientist? A Structural Study of LLM-Generated Research Methods
  2. 2025Deep Ideation: Designing LLM Agents to Generate Novel Research Ideas on Scientific Concept Network
  3. 2026Contemporary AI lacks the imagination to diverge or negate in science
  4. 2025Large Language Models for Scientific Idea Generation: A Creativity-Centered Survey
  5. 2026Orchestrating LLM Agents for Scientific Research: A Pilot Study of Multiple Choice Question (MCQ) Generation and Evaluation
  6. 2025InnovatorBench: Evaluating Agents' Ability to Conduct Innovative LLM Research
  7. 2023Large Language Models are Zero Shot Hypothesis Proposers
  8. 2025Towards Scientific Intelligence: A Survey of LLM-based Scientific Agents