arXiv AI

Why and How to Do Science When AI Is Better at It: A Framework Grounded in Human Understanding and Direction

The paper explores why humans should continue conducting science as AI becomes increasingly capable, arguing that science’s essential goal is durable, collective human understanding of the world for human ends and values. It proposes a two-part test—human understanding and human direction—to decide when research tasks can be delegated to AI, applying this framework across all research phases, including peer review and training. The authors emphasize that this guidance remains relevant even if AI outperforms humans, focusing on the impact of AI on human understanding and accountability rather than on tool type or current AI limitations.

arXiv AI
Sep 18

ScientistTwo: Pioneering the Human Knowledge Frontier with Autonomous AI

The paper introduces ScientistTwo, a fully autonomous multi‑agent framework that takes a scientific problem, establishes baselines, generates hypotheses, and coordinates specialized agents to conduct an end‑to‑end discovery cycle without human intervention. It rigorously tests and refines its methods through automated experiments, ablation studies, and a closed‑loop peer‑review engine. Benchmarking against top conferences (ICLR, ICML, NeurIPS) shows that ScientistTwo produces expert‑level, publishable papers and codebases that outperform human state‑of‑the‑art models and receive higher review ratings under automated AI review.

By Jaehyun Nam, Jinsung Yoon, Yanzhou Pan, Yubo Wang, Rui Meng, Parthasarathy Ranganathan, Tomas Pfister
arXiv AI
Aug 19

ASI-Bench: At the Dawn of Artificial Superintelligence

ASI‑Bench is a new benchmark that evaluates AI systems on their ability to conduct innovative exploration and autonomous scientific research across 11 domains, using 60 project‑level tasks. It progressively removes human methodological guidance to test whether AI can independently select methods, execute research, and produce verifiable results. Results from 18 state‑of‑the‑art agent–model configurations show a sharp performance drop when guidance is reduced, indicating current systems still rely heavily on human input.

By Junwei Zhou, Zhen Sun, Binyu Li, Jiangyu Zhou, Yuexi Pan, Hengyu Wang, Honghe Ren, Xiaohan Jia, Xueyang Zhou, Xiaoyu Cao, Yongchao Chen, Yuanning Feng, Junhao Wu, Cheng Zhang, Sijia Chen, Haoyu Xue, Chengsong You, Huan Wang, Koutian Wu, Peigan Gao, Jiakun Wu, Wenzhe Li, Ergan Shang, Qingyuan Zheng, Jingjing Zhou, Ruixuan Jia, Yan Xu, Hongrui Zhang, Xiao-Han Ma, Zhengxiang Cheng, Yuexing Hao, Liting Mai, Xianglin Ji, Wenjun Zhang, Zhuofan Chen, Yixiao Huang, Chi Wang, Wenyue Hua, Yilun Hao, Yuantao Zhai, Ziyan Zhao, Jingyan Xie
arXiv Computation and Language
Nov 4

IdeaTaste: Do LLMs Share the Research Taste of Human Researchers?

IdeaTaste is a benchmark that evaluates large language models (LLMs) on their ability to generate research proposals that align with what human researchers actually publish. The study compares proposals generated by six contemporary LLMs against real papers from four research areas, finding that LLMs achieve only about 26% of the citation matching and 18% of the idea matching of human communities. Even larger LLM populations and richer field contexts do not close this gap, indicating that LLMs struggle more with translating research directions into concrete approaches than with selecting the right topics.

By Shanshan Zhong, Chen Xu, Jingjie Ning, Zhengzhong Liu, Aviral Kumar, Chenyan Xiong
arXiv AI
Oct 27

Closing the empirical loop: autonomous AI agents conduct end-to-end research with human participants

The paper reports that a domain‑agnostic, agentic AI system can autonomously carry out the full scientific workflow—from hypothesis generation to data collection and manuscript preparation—by conducting three online psychological studies involving 288 participants. The AI designed experiments on visual working memory, mental rotation, and imagery vividness, coded analysis pipelines in over eight hours, and produced completed manuscripts, achieving methodological rigor comparable to experienced researchers. The work demonstrates the feasibility of AI‑driven scientific discovery while highlighting limitations in conceptual nuance and raising questions about scientific credit and understanding.

By Gabrielle Wehr, Reuben Rideaux, Amaya J. Fox, David R. Lightfoot, Jason Tangen, Jason B. Mattingley, Shane E. Ehrhardt
arXiv AI
Oct 16

From Research Goals to Scientific Closure: Evaluating Agents Through Evidence-Supported Physics Research

The paper introduces PhysClosureBench, a benchmark of 50 expert‑verified physics research tasks that test whether AI agents can organize investigations and produce evidence‑supported conclusions. Each task specifies a research goal and constraints, while the agent must design the investigation and validate evidence. Seven leading LLMs were evaluated, with Claude‑Opus‑5 scoring highest on local scientific accomplishment but achieving full research closure (RC4) only 32.7% of the time, revealing a gap between local success and complete closure.

By Yijia Mao, Song Dai, Tao Chen, Xiangyu Lu, Linqian Zeng, Chenxi Hu, Sitong Zhou, Dongyang He, Xiaodan Mao, Linxuan Zhang, Tong Zhu, Yafu Li, Fangchen Yu, Peng Ye, Feng He, Yu Cheng, Yun Luo
arXiv AI
Aug 26

Ethical LLM-Assisted Research: A Framework for Responsible Delegation, Verification, and Epistemic Value

The paper proposes a normative framework for ethical use of large language models (LLMs) in scientific research, treating reasoning as a distributed process where human control remains essential for epistemic legitimacy. It introduces key constructs—content origin, human verification, responsibility assignment, accountable ownership, and epistemic outcome—to separate claim provenance from verification and responsibility. The authors argue that the ethical boundary hinges on adequate verification and accountable human ownership, and they propose an "epistemic audit" to document delegation, verification, provenance, and responsibility for transparent, reviewable AI-assisted reasoning.

By Kalin Stoyanov
arXiv AI
Oct 2

Science or Slop?: Benchmarking and Mitigating Scientific Slop in AI-Generated Papers

The paper "Science or Slop?: Benchmarking and Mitigating Scientific Slop in AI-Generated Papers" introduces SciSlopBench, a dataset of 390 AI‑generated papers paired with human‑written counterparts, and defines six measures across Structure, Argument, and Artifacts to detect scientific slop. The authors show that these measures can identify AI papers with 85.9% accuracy and that higher slop correlates with lower ICLR ratings and distinguishes rejected from accepted papers. They also propose SciSlopHarness, a framework that guides a fixed LLM to revise only evidence‑supported sections, reducing the AI‑human gap by 63% without human reference targets.

By Yerim Oh, Young-Jun Lee, Jaewoo Ahn, Gunhee Kim, Dongyeop Kang