Fei Chao

Out of the Memory Barrier: A Highly Memory-Efficient Training System for LLMs with Million-Token Contexts

Flow Caching for Autoregressive Video Generation

Adaptive Routing of Experts via Fuzzy Rule Interpolation for Efficient Multimodal Learning

FlexibleLLM: Making Low-Bit Quantization for Large Language Models More Flexible and Efficient

UniCalli: A Unified Diffusion Framework for Column-Level Generation and Recognition of Chinese Calligraphy

Event-Anchored Frame Selection for Efficient Long-Video Understanding

RoPK: A Head-Level Key Cache Channel Pruning Method for Efficient Long-Context LLM Inference

Modality-Aware Quantization: Balancing Visual and Textual Fidelity in Multimodal Compression

Continuous Test-Time Adaptation of Vision-Language Models

Beyond Proxy Metrics: A New Evaluation Framework for LLM Compression by Directly Measuring Generative Faithfulness

SplitQuant: Efficient Low-Bit Quantization for Diffusion Transformers via In-Channel Dimension Splitting

Spotlight Attention: Towards Efficient LLM Generation via Non-linear Hashing-based KV Cache Retrieval

Determining Layer-wise Sparsity for Large Language Models Through a Theoretical Perspective

Video-RAG: Visually-aligned Retrieval-Augmented Long Video Comprehension

GS-Bias: Global-Spatial Bias Learner for Single-Image Test-Time Adaptation of Vision-Language Models

polybasic Speculative Decoding Through a Theoretical Perspective

Discovering Important Experts for Mixture-of-Experts Models Pruning Through a Theoretical Perspective

Low Rank Quantization Adaptation for Large Language Model

Automated Fine-Grained Mixture-of-Experts Quantization

Mitigating Gradient Interference for Efficient Sparse Fine-Tuning of Large Language Models

Extreme composite compression of large language models through joint optimization

Multi-Perspective Test-Time Prompt Tuning for Global, Local Visuals, and Language

Towards Efficient Automatic Self-Pruning of Large Language Models

Polybasic Speculative Decoding Under a Theoretical Perspective