Adding dynamics, reward, and policy feedback to LLM prompts degrades their already poor decision-making performance in MiniGrid navigation tasks.
Advances in Neural Information Processing Systems 36, 73383–73394 (2023)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Feedback-Induced Performance Decline in LLM-Based Decision-Making
Adding dynamics, reward, and policy feedback to LLM prompts degrades their already poor decision-making performance in MiniGrid navigation tasks.