ScientistOne: Towards Human-Level Autonomous Research via Chain-of-Evidence
For researchers and developers of autonomous AI systems, this work addresses the critical problem of output verifiability, providing a framework and system that eliminates common failure modes like hallucinated references and unreproducible results.
Autonomous research agents produce outputs with verifiability failures like fabricated citations and unreproducible scores. The authors introduce Chain-of-Evidence (CoE), a verifiability framework, and ScientistOne, a system that maintains evidence chains, achieving zero hallucinated references (0/337), perfect score verification (12/12), and high method-code alignment (14/15), matching or exceeding human experts on five tasks and generalizing to six additional tasks.
Autonomous research agents produce competitive solutions and professional-looking manuscripts, yet their outputs contain verifiability failures undetectable by surface-level evaluation: fabricated citations, unreproducible scores, and method descriptions that diverge from the implementation. We address this through three contributions. First, Chain-of-Evidence (CoE), a verifiability framework requiring every claim to be traceable to its evidence source. Second, ScientistOne, an end-to-end autonomous research system that maintains evidence chains by construction throughout literature review, solution discovery, and paper writing. Third, CoE Audit, a post-hoc audit whose four integrity checks -- score verification, specification violation, reference verification, and method-code alignment -- apply uniformly to all systems. Across 75 papers spanning five systems and five frontier research tasks, every baseline exhibits at least one systematic failure mode: hallucinated reference rates reach 21%, score verification passes in as few as 42% of papers, and method-code alignment ranges from 20% to 80%. ScientistOne achieves zero hallucinated references (0/337), perfect score verification (12/12), and the highest method-code alignment (14/15), while matching or exceeding human expert performance on all five tasks. ScientistOne further generalizes to six additional tasks spanning medical imaging, fine-grained recognition, 3D perception, and language modeling, achieving state-of-the-art on Parameter Golf and gold medals on MLE-Bench tasks where baselines fail entirely.