TY  - RPRT
TI  - Trust the Batch, On- or Off-Policy: Adaptive Policy Optimization for RL Post-Training
AU  - Rasool Fakoor
AU  - Murdock Aubry
AU  - Nicholas Stranges
AU  - Alexander J. Smola
PY  - 2026
UR  - https://arxiv.org/abs/2605.12380
ID  - 2605.12380
ER  -