conftrace_

overoptimization

3 papers

Explore in graph

Co-occurring keywords · papers mentioning both

rlhf (2) reward model (2) kl divergence budget (1) policy confidence (1) generalizability (1) iterative data smoothing (1) batch-wise sum-to-zero regularization (1) overfitting (1) large language model (1) alignment quality (1)

Papers

ConfPO: Exploiting Policy Model Confidence for Critical Token Selection in Preference Optimization ICML 2025 On the Robustness of Reward Models for Language Model Alignment ICML 2025 Iterative Data Smoothing: Mitigating Reward Overfitting and Overoptimization in RLHF ICML 2024