TY  - RPRT
TI  - Soft-SVeRL: Self-Verified Reinforcement Learning with Soft Rewards
AU  - Saurabh Dash
AU  - Pierre Clavier
AU  - John Dang
AU  - Matthias Galle
AU  - Marzieh Fadaee
AU  - Ahmet Üstün
AU  - Beyza Ermis
PY  - 2026
UR  - https://arxiv.org/abs/2605.28561
ID  - 2605.28561
ER  -