Xiang Liu

Reasoning Language Model Inference Serving Unveiled: An Empirical Study

DiffAdapt: Difficulty-Adaptive Reasoning for Token-Efficient LLM Inference

Zero in on Faithful Anchors: High-Fidelity Visual Token Condensation for Multimodal Large Language Models

Beyond Single Embeddings: Capturing Diverse Targets with Multi-Query Retrieval

OracleKV: Oracle Guidance for Question-Independent KV Cache Eviction

AnTKV: Anchor Token-Aware Ultra-Low-Bit Vector Quantization for KV Cache in Large Language Models

Mediater: Memory-efficient LLM Merging with Less Parameter Conflicts and Uncertainty Based Routing

Is Privacy Always Prioritized over Learning? Probing LLMs' Value Priority Belief under External Perturbations

Can LLMs Maintain Fundamental Abilities under KV Cache Compression?

FlowKV: Enhancing Multi-Turn Conversational Coherence in LLMs via Isolated Key-Value Cache Management

Can Compressed LLMs Truly Act? An Empirical Evaluation of Agentic Capabilities in LLM Compression

ChunkKV: Semantic-Preserving KV Cache Compression for Efficient Long-Context LLM Inference

ChunkKV: Semantic-Preserving KV Cache Compression for Efficient Long-Context LLM Inference