On-Policy or Off-Policy Learning? A Systematic Study of Distillation Dynamics
Paper • 2609.35259 • Published • 202
Rania Bouazizi
rbouazizi
AI & ML interests
reward hacking, speech translation, curriculum learning
Recent Activity
liked a dataset 1 day ago
Gene829/gene-reinforcement-learning-instruct liked a dataset 1 day ago
Srishti280992/repro-exact-unlearning-in-reinforcement-learning-traces liked a model 1 day ago
zxy0307/AI_06_reinforcement_Learning_baizhanOrganizations
None yet
upvoted 5
papers 1 day ago
Sharpening Tax in Post-Training
Paper • 2610.01509 • Published • 111
Thinking Outside the Box: Can Language Models Rely on External Guidance Selectively?
Paper • 2609.39578 • Published • 72
TERRA: Terrain-Aware Reconstruction, Retargeting and Control for Musculoskeletal Locomotion
Paper • 2609.38653 • Published • 39
GUI-HARVEST: Self-Improving GUI Agents through Evidence-Driven Harness Evolution
Paper • 2610.00948 • Published • 22
upvoted 3
papers 2 days ago
TRACE: Rollout-Guided Quantization-Aware Training for FP4 Reinforcement Learning of MoE Language Models
Paper • 2610.07767 • Published • 92
Rationale-Guided Policy Optimization: Learning to Reason with Adaptive Rationale Scaffolding
Paper • 2610.07342 • Published • 21
TRIAGE: Direction-Aware Mismatch Stabilization of Native NVFP4 Reinforcement Learning
Paper • 2610.07043 • Published • 43
upvoted 3
papers 3 days ago
Towards Looped Models Done Right, Part II: Rethinking at Fixed Points
Paper • 2610.06833 • Published • 33
LoGRA: Scaling LLM Reinforcement Learning with Low-Rank Gradient Sketches
Paper • 2610.06647 • Published • 148
Self-Generated Feedback Destabilizes Test-Time Training: A Causal Decomposition of Long-Horizon Adaptation
Paper • 2610.05076 • Published • 26
upvoted 3
papers 4 days ago
MetaRubric: Learning to Reward for Rubric-Based Reinforcement Learning
Paper • 2610.02824 • Published • 33
LexReward: A Taxonomy-Driven Reward Framework for Legal Language Models
Paper • 2609.39071 • Published • 66
Collective Bias Mitigation via Model Routing and Collaboration
Paper • 2610.03240 • Published • 20
upvoted 2
papers 5 days ago
upvoted 2
papers 6 days ago
upvoted 2
papers 8 days ago