PhAI Labs releases DiscoveryRSI research agent and 27B model
PhAI Labs, working with several universities and research institutions, proposed DiscoveryRSI and trained DiscoveryRSI-27B using a hypothesis-credit attribution mechanism, releasing the paper, code and models. The system stores evolving hypotheses and test plans in a research hypothesis tree and is paired with DFM-Bench, which evaluates hypothesis revisions, research paths and whole research episodes. The team reports that DiscoveryRSI's group scored 6.3% in a literature cross-check, the highest of five model groups, while the Claude Opus 5 group scored 2.3% and ranked last; it cautions that group sample sizes differ and the comparison is descriptive, so it does not prove the hypotheses were adopted or validated by later research.
The work tries to make a research agent's self-improvement traceable and reviewable through hypothesis trees and contribution attribution, with code and models released for other researchers to test.
机器之心 2026-10-19 12:02 北京
PhAI Labs 是一家面向下一代智能系统的新型研究组织（Neo Lab），长期关注人工智能与科学发现深度融合的前沿方向。
机器之心发布
Claude，竟然排在最后？
#PhAI Labs#DiscoveryRSI#DiscoveryRSI-27B#DFM-Bench#Claude Opus 5#GPT-5.6 Sol#GPT-6 Luna#DeepSeek#GLM#Stanford University#MIT#Harvard Medical School