Negative Knowledge as Failure-aware Shared Memory for AutoResearch

2026-06-18

AI Summary

This work addresses the underutilization of failed experiments as reusable knowledge assets in AI-assisted scientific research. It proposes a “negative knowledge memory layer” that employs curation agents to structure and store unsuccessful experimental attempts in a shared repository, enabling downstream research agents to explicitly accept or reject these records when designing new experiments. For the first time, structured negative knowledge is treated as an independent knowledge asset, explicitly maintained and transferred across tasks. Integrated into the AutoResearch multi-agent framework, the approach supports the generation, typologized storage, and retrieval of negative knowledge. Evaluated on ScienceAgentBench and nonlinear mathematical physics PDE tasks, the method solves novel problems that baseline systems cannot address—using fewer tokens—and substantially improves cross-problem exploration efficiency.

Abstract

AI-assisted research systems generate many failed attempts, but those failures rarely become a durable, shared knowledge asset. We propose a negative knowledge memory layer: a curator agent converts each failed attempt into a bounded, typed record in a shared bank, and a downstream research agent explicitly adopts or rejects those records before proposing its next experiment. We evaluate this layer in two settings: same-task retry on ScienceAgentBench and cross-task scientific research on two nonlinear math-physics PDE problems. The negative knowledge layer outperforms vanilla AutoResearch baselines while using fewer tokens; agents with the negative knowledge bank solve new tasks that all baselines fail to solve in PDE systems research. We also show that the previous negative knowledge bank can transfer and enhance AutoResearch on different PDE problems. These results suggest that structured negative knowledge is a knowledge asset that should be explicitly maintained in broader AI-engaged scientific research beyond a memory-compression or debugging aid, alongside positive findings, as a collective infrastructure for scientific memory. Code is available at https://github.com/hch-wang/Negative_Knowledge.

Problem

• negative knowledge
• AI-assisted research
• failed attempts
• shared memory
• scientific research

Innovation

• negative knowledge
• shared memory
• AutoResearch
• failure-aware
• scientific reasoning