method · first seen 11 Oct 2026
LLM as value function
Using a pretrained language model with in-context learning to estimate state values without training a separate critic.
Mentions per week
PapersCodeDiscussion
Earliest sources
- LLM Value Functions for Credit Assignment in Agentic RLPapers · 11 Oct 2026 · Kaijie Zhu, Weixiang Yan, Omar Attia, Abbas Kazerouni, Alexander Toshev, Andrew…
Latest sources
- LLM Value Functions for Credit Assignment in Agentic RLPapers · 11 Oct 2026 · Kaijie Zhu, Weixiang Yan, Omar Attia, Abbas Kazerouni, Alexander Toshev, Andrew…