A monitor trained on agent uncertainty statistics triggers rollbacks that improve multi-agent LLM collaboration across three test environments.
In ICML 2024 Workshop on Mechanistic Interpretability
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Preventing Rogue Agents Improves Multi-Agent Collaboration
A monitor trained on agent uncertainty statistics triggers rollbacks that improve multi-agent LLM collaboration across three test environments.