Reward shaping
Technique · Training · 2 items since May 2026
Turning sparse outcome rewards into denser, better-targeted signals during RL.
Related
- Organizations
Trends
Items
- Jul 25
Meituan's CAST turns solver values into rewards
Converts a game solver's change in state value into a dense turn-level reward for RLVR, provably equivalent to logit-free on-policy distillation, beating trained baselines on Sokoban, Minesweeper, and Rush Hour and transferring zero-shot to ALFWorld and WebShop.
- May 16
Error diversity predicts RLVR training success
EDAS amplifies advantage penalties for dominant repeated errors and softens them for rare exploratory failures, adding an average 6.29 points over DAPO on Qwen3-8B across seven math benchmarks as a drop-in for any RLVR algorithm.