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
158 results for “keyword:"grpo"” in ICLR 2027. 42 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]Winner-Take-All: Where Exponential Entropy Collapse Comes From in Reinforcement Learning for LLMs▲5.4k3.00
[PDF]Reward Reasoning for Its Gain over a Snap Judgment: Inverse-RL Cold Start of GUI Agents with an Adversarial Critic▲5.4k3.00
[PDF]Future-KL Regularized GRPO: Process-Level Credit Assignment from -Divergence Regularization▲143.33
[PDF]Matched Compute Is Not Matched Signal: Where the Gradient Goes in Group-Relative Reinforcement Learning▼1361.33
[PDF]DataLens: Predicting Finite-Step Data Utility for Data-Efficient LLM Reinforcement Learning▼4.3k2.00
[PDF]ProSearch: Teaching LLMs When (Not) to Search over Streaming Inputs via Reinforcement Learning▼5.8k2.33
[PDF]Cool the Sampler, Not the Learner: Sampling Temperature Moves the Staleness Cliff of Importance-Corrected GRPO▼4.3k2.00
[PDF]PEPFM-TANRL: FINE-TUNING A MULTI-MANIFOLD FLOW MATCHING MODEL FOR PEPTIDE DESIGN VIA TANGENT-SPACE GRPO▼1.2k1.67
[PDF]Where and When: Structure-Grounded Credit Assignment for Cross-Modal Reinforcement Learning▼4.5k2.67
[PDF]Task-Aware Configuration Selection for Preference Optimization via Early Training Dynamics▲3.0k2.33
[PDF]From-Base RLVR for Verify-Only Judges: A Rule-Based Reward Turns Base Models into Trajectory Auditors▲11.00
Withdrawn or desk rejected (42)
[PDF]Verifier-Induced Strategy Collapse: Family-Correlated False Negatives Eliminate Valid Strategies While Capability Metrics ImproveDesk rejected
[PDF]TimeQuest: Scaling Inference-Time Compute in Video-LLMs via Decoupled Temporal NavigationWithdrawn2.00
[PDF]Gold-LP: Data Selection for Vision-Language Reinforcement Learning with Verifiable RewardsWithdrawn1.67