NeurIPS 2026 reviews are up: reviews, scores, rebuttals and meta-reviews for 7,877 papers. See the highest rated
acceptodds: the papers of ICLR 2027
47 results for “keyword:"rlhf"” in ICLR 2027. 11 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]Rethinking KL Regularization in RLHF: From Wrong Value Estimation to Correct Gradient Optimization▼1361.33
[PDF]First-Order In-Context Shapley: Scalable Data Auditing for LLM Alignment and Evaluation▼5.8k2.33
[PDF]Scoring Higher, Answering Worse: Mitigating Reward Hacking in Rubric-Based RL via Protocol-Level Rubrics▲5103.00
[PDF]One Self, Many Masks: Language Models Have a Self That Persists Beneath Their Personas▼1.2k1.67
[PDF]Stop Training on Pairs You've Already Learned: Policy-Aware Pair Selection for Preference Optimization▼5.8k2.33
[PDF]Theoretical Tensions in RLHF: Reconciling Empirical Success with Inconsistencies in Social Choice Theory▼1361.33
[PDF]Which Preferences to Train On? End-to-End Multi-Objective Alignment with an Adversarial Preference Distribution▲3952.67
[PDF]Zeroth-Order Momentum and Adaptive Methods for Nonsmooth Policy Optimization from Human Preferences▲5.4k3.00
[PDF]ADAPTRLHF: ADAPTIVE KL-CONSTRAINED REIN FORCEMENT LEARNING FROM HUMAN FEEDBACK WITH GROUP-NORMALIZED AD VANTAGE ESTIMATION▲11.00
[PDF]The Comparison Graph Determines What RLHF Can Learn: Spectral Identifiability, Error Bounds, and a Pre-Training Diagnostic for Preference Data▲3.0k2.33
[PDF]STABLE ALIGNMENT: ADAPTIVE KL CONTROL WITH GROUP-NORMALIZED ADVANTAGE ESTIMATION FOR RLHF▲11.00
[PDF]A Preference Is a Spectrum, Not a Scalar: PRISM for Structured Ordinal Reward Modeling▼4.3k2.00
[PDF]Localizing Failures: Hallucination-Grounded Preference Learning for Faithful Generative Super-Resolution▲5103.00
[PDF]Categorical Regression in REBEL: Rate Preservation, Boundary Bias, and No General Robustness▼1.2k1.67