Multi-Agent Autonomous Lab Turns Rejected Proposals Into Search Guidance
A multi-agent autonomous experimentation paper from Institute of Science Tokyo demonstrates an agent-driven research pipeline: a PI agent leading the research rejected 75 proposals over roughly 30 hours, while a manager-role agent could identify recurring rejected ideas and explicitly exclude those directions from subsequent instructions. As a result, the next 9 research proposals explored 7 previously untouched directions.
Confirmed
- The key mechanism reported in the paper is that "criticism changes the next search step": rejected proposals become information rather than just a denial on paper.
- krishnan's reading of the paper is that this reject–record–exclude loop is central to how multi-agent systems can effectively explore new directions.
Why it matters
- This mechanism gives autonomous experimentation systems an explicit way to avoid repeating the same dead ends, allowing the search to keep pushing toward unexplored directions.
- krishnan proposed industrial applications based on this (he explicitly noted these are suggested directions, not results from the paper): in troubleshooting, record which explanations have been tested, what ruled them out, and what evidence would justify reopening them.
- He further offered a testable experimental design: compare an agent's "fresh retry" on the same problem against a "retry with explicit rejection records," to see whether the latter explores the solution space more effectively.
2026-10-31 ~ 2026-10-31 · 5 related posts
Primary sources
- [source] PI agent rejected 75 proposals in 30 hours; manager banned recurring bad ideas — krishnan · 2026-10-31
- Next 9 proposals explored 7 new directions: rejection became information — krishnan · 2026-10-31
- Testing agents with explicit rejection records: a cheaper path to verified troubleshooting answers — krishnan · 2026-10-31
- [source] krishnan on autonomous lab agents: criticism changed the next search, not just verdicts — krishnan · 2026-10-31
- krishnan proposes logging ruled-out troubleshooting hypotheses for future reopening — krishnan · 2026-10-31