Jiang, Li

Offline RL with No OOD Actions: In-Sample Learning via Implicit Value Regularization

Most offline reinforcement learning (RL) methods suffer from the trade-off between improving the policy to surpass the behavior policy and constraining the policy to limit the deviation from the...

DELTA: Decoupling Latent Heterogeneity in Asymmetric Low-Rank Compression

PCoMoE: Shifting MoE Inference from Monolithic Expert Selection to Fine-Grained Path Composition

Seer: Efficient KV Cache Management for LLM Acceleration via Mitigating Semantic Search Inefficiency

Tibetan Number Identification Based on Classification of Number Components in Tibetan Word Segmentation

AIR-based light clients for supporting Moses engine training

FlexQuant: A Flexible and Efficient Dynamic Precision Switching Framework for LLM Quantization

Calibrated Speculative Decoding: Frequency-Guided Candidate Selection for Efficient Inference