How Much Training-Inference Mismatch Can On-Policy Distillation Tolerate?
Paper • 2611.11450 • Published • 13
🤝 Open to Collab
Songlin Jiang
HollowMan6
AI & ML interests
AI Infrastructure, Reinforcement Learning
Recent Activity
authored a paper 2 days ago
How Much Training-Inference Mismatch Can On-Policy Distillation Tolerate? upvoted a paper 2 days ago
How Much Training-Inference Mismatch Can On-Policy Distillation Tolerate? upvoted a paper about 1 month ago
Learning Discriminative Geometry for Drifting ModelsOrganizations
None yet
upvoted a paper 2 days ago
upvoted 7
papers about 1 month ago
Learning Discriminative Geometry for Drifting Models
Paper • 2610.04703 • Published • 6
NeMo-DCR: Bit-Exact Delta-Compressed Refit for Scalable Agentic RL at Trillion-Parameter Scale
Paper • 2610.08430 • Published • 21
On the Scaling of PEFT: Towards Million Personal Models of Trillion Parameters
Paper • 2606.02437 • Published • 104
MinT: Managed Infrastructure for Training and Serving Millions of LLMs
Paper • 2605.13779 • Published • 183
LongStraw: Long-Context RL Beyond 2M Tokens under a Fixed GPU Budget
Paper • 2607.14952 • Published • 77
Macaron-V1: Towards Open Continual Learning with Self-Improvement and Mixture-of-LoRA
Paper • 2608.09819 • Published • 113
Nereus: Adaptive Parallelism for LLM Post-Training
Paper • 2609.34645 • Published • 12
upvoted 3
papers about 1 year ago
OpenRLHF: An Easy-to-use, Scalable and High-performance RLHF Framework
Paper • 2405.11143 • Published • 40
QUASAR: Quantum Assembly Code Generation Using Tool-Augmented LLMs via Agentic RL
Paper • 2510.00967 • Published • 13
Fine-Tuning Large Language Models on Quantum Optimization Problems for Circuit Generation
Paper • 2504.11109 • Published • 3