mlsanityData to 14 Oct 2026

AI Research Agents Narrow Scientific Exploration

arXiv 2605.27905 ↗First posted 27 May 2026cs.CLcited by 4

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

Abstract (arXiv)

AI research agents now support large-scale AI-assisted scientific discovery. We examine whether AI-generated ideas broaden scientific exploration or primarily reinforce existing work. Using five agent frameworks and five large language models, we generate 219,655 ideas for different scientific fields. Across experiments, four consistent patterns emerge. First, AI-generated ideas are more concentrated than human-authored papers within the same research area. Second, they remain much closer to starting literature than later human follow-on work does. Third, AI-generated ideas align less with future human research. Last, AI-generated ideas are located in lower-impact regions of the historical scientific landscape. Overall, current AI research agents appear better suited to local elaboration than to broadening scientific exploration.

Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). Placed in Web, search, and research agents, one of 16 directions in LLM agents and tool use. Builds on 25 of its 70 references that are in the map. First posted on arXiv 27 May 2026.

Builds on

25 of 70 references.

Closest ancestors · 2

Same direction, newest first

  1. 2025Agent Laboratory: Using LLM Agents as Research Assistants8 Jan
  2. 2024Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researchers6 Sep

Other references · 23

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2025Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Modelscited by 1,261
  3. 2023ChatGPT Outperforms Crowd-Workers for Text-Annotation Taskscited by 870
  4. 2020SPECTER: Document-level Representation Learning using Citation-informed Transformerscited by 430
  5. 2025Accelerating scientific discovery with Co-Scientistcited by 401
  6. 2024ResearchAgent: Iterative Research Idea Generation over Scientific Literature with Large Language Modelscited by 210
  7. 2023SciMON: Scientific Inspiration Machines Optimized for Noveltycited by 179
  8. 2026Towards End-to-End Automation of AI Researchcited by 159
  9. 2024Classic GNNs are Strong Baselines: Reassessing GNNs for Node Classificationcited by 85
  10. 2024GTA: A Benchmark for General Tool Agentscited by 77
  11. 2024Polynormer: Polynomial-Expressive Graph Transformer in Linear Timecited by 65
  12. 2024CS-Bench: A Comprehensive Benchmark for Large Language Models towards Computer Science Masterycited by 30
11 more
  1. 2013Tradition and Innovation in Scientists' Research Strategiescited by 26
  2. 2024VisMin: Visual Minimal-Change Understandingcited by 23
  3. 2022Dynamic Sparse Network for Time Series Classification: Learning What to "see''cited by 22
  4. 2019The Diversity-Innovation Paradox in Sciencecited by 21
  5. 2024Learning Long Range Dependencies on Graphs via Random Walkscited by 18
  6. 2020Neural Embeddings of Scholarly Periodicals Reveal Complex Disciplinary Organizationscited by 16
  7. 2024NarrativeBridge: Enhancing Video Captioning with Causal-Temporal Narrativecited by 10
  8. 2025Hermes 4 Technical Reportcited by 8
  9. 2024TACT: Advancing Complex Aggregative Reasoning with Information Extraction Toolscited by 5
  10. 2022Aging and the Narrowing of Scientific Innovationcited by 1
  11. 2024Diffusing to the Top: Boost Graph Neural Networks with Minimal Hyperparameter Tuningcited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 4.

4 papers.

  1. 2026AutoResearch: Insight In, Hallucination Outcited by 1
  2. 2026Large Discovery Models: Empirically-grounded Model-Based Open-Ended Searchcited by 0
  3. 2026Social Influence and the Allocation of Scientific Attention in AI Populationscited by 0
  4. 2026Large Language Models Extend Known Research But Do Not Create New Research Directionscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026AgentIdeaBench: Benchmarking Scientific Ideation in the Agent Era
  2. 2026Contemporary AI lacks the imagination to diverge or negate in science
  3. 2024The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery
  4. 2026AI Scientists as Engines of Discovery: A Case for Development within Reformed Institutions
  5. 2025Transforming Science with Large Language Models: A Survey on AI-assisted Scientific Discovery, Experimentation, Content Generation, and Evaluation
  6. 2025AI-Researcher: Autonomous Scientific Innovation
  7. 2026Position: AI Agents in Scientific Teams Should Be Studied as Human-Agent Systems
  8. 2025Virtuous Machines: Towards Artificial General Science