Showing cs.LGShow all
2 papers · 1 filter
cs.LG2026
The Cancellation Hypothesis in Critic-Free RL: From Outcome Rewards to Token Credits
Tianhao Cheng, Zeyu Huang, Zihan Qiu +5
A commonly accepted explanation of critic-free RL for LLMs, based on sequence-level rewards, is that it reinforces successful rollouts with a positive advantage while penalizing fa…
cs.LG2026
DARC: Disagreement-Aware Alignment via Risk-Constrained Decoding
Mingxi Zou, Jiaxiang Chen, Junfan Li +4
Preference-based alignment methods (e.g., RLHF, DPO) typically optimize a single scalar objective, implicitly averaging over heterogeneous human preferences. In practice, systemati…