ReDAct defers LLM agent decisions to a larger model based on small-model uncertainty exceeding a threshold, achieving equivalent performance to full large-model use at lower cost in environments like ALFWorld and MiniGrid.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
ReDAct: Uncertainty-Aware Deferral for LLM Agents
ReDAct defers LLM agent decisions to a larger model based on small-model uncertainty exceeding a threshold, achieving equivalent performance to full large-model use at lower cost in environments like ALFWorld and MiniGrid.