TY  - RPRT
TI  - Policy-Conditioned Counterfactual Credit for Verifiable Reinforcement Learning of Long-Horizon Language Agents
AU  - Renwei Meng
PY  - 2026
UR  - https://arxiv.org/abs/2606.05263
ID  - 2606.05263
ER  -