Tianxin Wei

Inference Scaling of LLM Ensembling: Bridging Token Spaces with Token Translation

MC-Search: Evaluating and Enhancing Multimodal Agentic Search with Structured Long Reasoning Chains

Seeing but Not Believing: Probing the Disconnect Between Visual Attention and Answer Correctness in VLMs

NIRVANA: Structured Pruning Reimagined for Large Language Models Compression

Geometri-Disentangelment Unlearning

ReMix: Reinforcement Routing for Mixtures of LoRAs in LLM Finetuning

WAPITI: A Watermark for Finetuned Open-Source LLMs

TRQA: Time Series Reasoning Question And Answering Benchmark

Graph4MM: Weaving Multimodal Learning with Structural Information

Automatic Task-aware Instruction Optimizer for Black-box LLMs

Breaking Silos: Adaptive Model Fusion Unlocks Better Time Series Forecasting

Robust Watermarking for Diffusion Models: A Unified Multi-Dimensional Recipe

WAPITI: A Watermark for Finetuned Open-Source LLMs