Hai Jin

RLAP-CLIP: Continual Multimodal Learning with Prototype Adaptation and Difficulty-Aware Routing

Vulcan: Crafting Compact Class-Specific Vision Transformers For Edge Intelligence

Figma2Code: Automating Multimodal Design to Code in the Wild

APT: Towards Universal Scene Graph Generation via Plug-in Adaptive Prompt Tuning

AgentRace: Benchmarking Efficiency in LLM Agent Frameworks

Vanish into Thin Air: Cross-prompt Universal Adversarial Attacks for SAM2

Sim-LLM: Optimizing LLM Inference at the Edge through Inter-Task KV Reuse

Dataflow-Guided Neuro-Symbolic Language Models for Type Inference

FEDNET: FREQUENCY ENHANCED DECOMPOSED NETWORK FOR OUT-OF-DISTRIBUTION TIME SERIES CLASSIFICATION

CodeSync: Synchronizing Large Language Models with Dynamic Code Evolution at Scale