LaVi Lab

[ICCV 2025] "Fine-grained Spatiotemporal Grounding on Egocentric Videos"

Implementation of "IntentEdit: Multi-Agent Reasoning for Intent-Driven Complex Image Editing"

Official code for "Rethinking Chain-of-Thought Reasoning for Videos"

The code for paper 'Learning from Videos for 3D World: Enhancing MLLMs with 3D Vision Geometry Priors'

[ICCV 2025] Official code for "AIM: Adaptive Inference of Multi-Modal LLMs via Token Merging and Pruning"

[CVPR 2025] The code for paper ''Video-3D LLM: Learning Position-Aware Video Representation for 3D Scene Understanding''.

[Findings of ACL 2025] "C2LEVA: Toward Comprehensive and Contamination-Free Language Model Evaluation"

[EMNLP 2023 Demo] "CLEVA: Chinese Language Models EVAluation Platform"

[ACL 2025] Official code for ''Learning to Reason from Feedback at Test-Time''.