mlsanityData to 7 Oct 2026

AI Scientists Fail Without Strong Implementation Capability

arXiv 2506.01372 ↗First posted 2 Jun 2025cs.AI cs.CL cs.LGcited by 20

Not on the map. Why?

Abstract (arXiv)

The emergence of Artificial Intelligence (AI) Scientist represents a paradigm shift in scientific discovery, with large language models (LLMs) taking the lead as the primary executor in the entire scientific workflow from idea generation to experiment implementation. Recent AI Scientist studies demonstrate sufficient capabilities for independent scientific discovery, with the generated research reports gaining acceptance at the ICLR 2025 workshop and ACL 2025, arguing that a human-level AI Scientist, capable of uncovering phenomena previously unknown to humans, may be imminent. Despite this substantial progress, AI Scientist has yet to produce a groundbreaking achievement in the domain of computer science on par with automated scientific tools. Based on extensive quantitative evidence from existing benchmarks in complex engineering tasks and a systematic evaluation assess 28 research papers generated by five advanced AI Scientist systems, we argue that the fundamental bottleneck for AI Scientists lies in their capability to execute the requisite verification procedures. Current AI Scientist systems lack the execution capabilities needed to execute rigorous experiments and produce high-quality scientific papers. To better illustrate the root cause of this implementation gap, we provide an in-depth discussion on the fundamental limitations of AI Scientist. This position paper aims to call for the participants in the community to bridge the implementation gap.

Key facts. Cited by 20 papers in the map (data through 7 Oct 2026). Builds on 75 of its 95 references that are in the map. First posted on arXiv 2 Jun 2025.

Builds on

75 of 95 references.

References · 75

  1. 2025Qwen3 Technical Reportcited by 8,195
  2. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  3. 2020Scaling Laws for Neural Language Modelscited by 6,411
  4. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
  5. 2013Distributed Representations of Words and Phrases and their Compositionalitycited by 3,462
  6. 2023SWE-bench: Can Language Models Resolve Real-World GitHub Issues?cited by 3,193
  7. 2024LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Codecited by 1,906
  8. 2023Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generationcited by 1,332
  9. 2025s1: Simple test-time scalingcited by 1,311
  10. 2023A Survey on Evaluation of Large Language Modelscited by 1,227
  11. 2024The AI Scientist: Towards Fully Automated Open-Ended Scientific Discoverycited by 1,002
  12. 2024Large Language Model based Multi-Agents: A Survey of Progress and Challengescited by 752
38 more
  1. 2025BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agentscited by 616
  2. 2025Why Do Multi-Agent LLM Systems Fail?cited by 559
  3. 2024A Survey on RAG Meeting LLMs: Towards Retrieval-Augmented Large Language Modelscited by 450
  4. 2024MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineeringcited by 429
  5. 2025Agent Laboratory: Using LLM Agents as Research Assistantscited by 405
  6. 2025LLMs Get Lost In Multi-Turn Conversationcited by 400
  7. 2025The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Searchcited by 375
  8. 2025Accelerating scientific discovery with Co-Scientistcited by 353
  9. 2024Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researcherscited by 348
  10. 2022Large Language Models are Better Reasoners with Self-Verificationcited by 304
  11. 2025PaperBench: Evaluating AI's Ability to Replicate AI Researchcited by 286
  12. 2025Model Context Protocol (MCP): Landscape, Security Threats, and Future Research Directionscited by 257
  13. 2024Scaling Large Language Model-based Multi-Agent Collaborationcited by 235
  14. 2025A Survey on Test-Time Scaling in Large Language Models: What, How, Where, and How Well?cited by 215
  15. 2025AIDE: AI-Driven Exploration in the Space of Codecited by 206
  16. 2025ZeroSearch: Incentivize the Search Capability of LLMs without Searchingcited by 203
  17. 2024Chain of Agents: Large Language Models Collaborating on Long-Context Taskscited by 201
  18. 2025When More is Less: Understanding Chain-of-Thought Length in LLMscited by 182
  19. 2023SciMON: Scientific Inspiration Machines Optimized for Noveltycited by 174
  20. 2024CycleResearcher: Improving Automated Research via Automated Reviewcited by 142
  21. 2024AutoSurvey: Large Language Models Can Automatically Write Surveyscited by 116
  22. 2024CORE-Bench: Fostering the Credibility of Published Research Through a Computational Reproducibility Agent Benchmarkcited by 115
  23. 2024Survey on Large Language Model-Enhanced Reinforcement Learning: Concept, Taxonomy, and Methodscited by 114
  24. 2025DeepReview: Improving LLM-based Paper Review with Human-like Deep Thinking Processcited by 97
  25. 2024SWE-Bench+: Enhanced Coding Benchmark for LLMscited by 95
  26. 2024Chain of Ideas: Revolutionizing Research Via Novel Idea Development with LLM Agentscited by 94
  27. 2024SciAgents: Automating scientific discovery through multi-agent intelligent graph reasoningcited by 90
  28. 2025A Survey of AI Agent Protocolscited by 77
  29. 2024AutoFlow: Automated Workflow Generation for Large Language Model Agentscited by 76
  30. 2025Superintelligent Agents Pose Catastrophic Risks: Can Scientist AI Offer a Safer Path?cited by 70
  31. 2024MOOSE-Chem: Large Language Models for Rediscovering Unseen Chemistry Scientific Hypothesescited by 68
  32. 2025Paper2Code: Automating Code Generation from Scientific Papers in Machine Learningcited by 67
  33. 2024MLR-Copilot: Autonomous Machine Learning Research based on Large Language Models Agentscited by 64
  34. 2025PaSa: An LLM Agent for Comprehensive Academic Paper Searchcited by 61
  35. 2025ResearchBench: Benchmarking LLMs in Scientific Discovery via Inspiration-Based Task Decompositioncited by 57
  36. 2025CodeScientist: End-to-End Semi-Automated Scientific Discovery with Code-based Experimentationcited by 54
  37. 2024Nova: An Iterative Planning and Search Approach to Enhance Novelty and Diversity of LLM Generated Ideascited by 52
  38. 2021Inconsistency in Conference Peer Review: Revisiting the 2014 NeurIPS Experimentcited by 49

50 most cited of 75 All 75 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 9; 2026: 11.

20 papers.

Sort:

By area:

  1. 2025Memento: Fine-tuning LLM Agents without Fine-tuning LLMscited by 119
  2. 2025The Illusion of Diminishing Returns: Measuring Long Horizon Execution in LLMscited by 62
  3. 2025DeepScientist: Advancing Frontier-Pushing Scientific Findings Progressivelycited by 56
  4. 2025RECODE-H: A Benchmark for Research Code Development with Interactive Human Feedbackcited by 24
  5. 2026AutoFigure: Generating and Refining Publication-Ready Scientific Illustrationscited by 20
  6. 2025Scientific Algorithm Discovery by Augmenting AlphaEvolve with Deep Researchcited by 16
  7. 2026ResearchGym: Evaluating Language Model Agents on Real-World AI Researchcited by 15
  8. 2025Jr. AI Scientist and Its Risk Report: Autonomous Scientific Exploration from a Baseline Papercited by 10
  9. 2026Can We Predict Before Executing Machine Learning Agents?cited by 10
  10. 2026From Fluent to Verifiable: Claim-Level Auditability for Deep Research Agentscited by 7
  11. 2026MLReplicate: Benchmarking Autonomous Research Systems for Machine Learning Reproducibilitycited by 6
  12. 2026Can LLMs Write Reliable Rubrics? A Meta-Evaluation for Experiment Reproductioncited by 6
8 more
  1. 2026NanoResearch: Co-Evolving Skills, Memory, and Policy for Personalized Research Automationcited by 4
  2. 2025AI for Scientific Discovery is a Social Problemcited by 3
  3. 2025ReX-MLE: The Autonomous Agent Benchmark for Medical Imaging Challengescited by 3
  4. 2026Overview of the NLPCC 2026 Shared Task 11: Agent-Based Experiment Reproduction from Scientific Paperscited by 2
  5. 2025The Variance Paradox: How AI Reduces Diversity but Increases Noveltycited by 1
  6. 2026verdi: retrieval is not transfer for continual world model optimizationcited by 0
  7. 2026Personalized Auto-Research: Towards a True AI Co-Scientistcited by 0
  8. 2026SocioVerse2: A Longitudinal Dynamic Social Simulation Framework under a Human-AI Co-evolutionary Paradigmcited by 0