Autonomous Research Agents: A Survey of AI Scientists and the Verification Gap

Summary

Surveys 35 AI-scientist systems with full-text coding of 26 entries along seven audit dimensions including novelty verification and result-selection disclosure, and reports that code release is now common (83% of 24 runnable systems) while reproducibility-grade and claim-verification artifacts are not (38% release seeds or execution traces; 38% report any novelty-verification method). Its central claim is that the field's bottleneck is no longer whether agents can complete research tasks but whether reviewers can verify the claims those agents produce. The field-level warrant for making transparency and traceability first-class outcome variables rather than accuracy alone, and the most current (Aug 2026) survey of research-agent systems.

Contribution

Queued — not yet read.

Method

Queued — not yet read.

Relevance to RISE

Surfaced by the 2026-08-29 literature sweep as a high-priority candidate for the Architecture as Epistemology (ISR) paper (angle: epistemic quality of LLM outputs; MAS failure and cost analyses). Verified against: Opened https://arxiv.org/abs/2608.05179 (arXiv abstract page, v1; 2026 submission)

Critique / open questions

Queued — not yet assessed.

Key quotes

Large language model (LLM) agents are increasingly used across the scientific research lifecycle: ideation, literature search, experiment design and execution, analysis, manuscript drafting, and review. (abstract)

End-to-end AI scientist systems can now produce paper-like manuscripts, but their claims are often harder to verify than their code is to run. (abstract)