TY  - RPRT
TI  - ABC: Advantage-Based Control Variates for Reinforcement Learning with Verifiable Rewards
AU  - Hsiao-Ru Pan
AU  - Florent Draye
AU  - Bernhard Schölkopf
PY  - 2026
UR  - https://arxiv.org/abs/2609.36058
ID  - 2609.36058
ER  -