[ICCVW 25] LLaVA-MORE: A Comparative Study of LLMs and Visual Backbones for Enhanced Visual Instruction Tuning

The official repo of our work "Pensieve: Retrospect-then-Compare mitigates Visual Hallucination"

FaceXBench: Evaluating Multimodal LLMs on Face Understanding

The Official Code Repo for EgoOrientBench [CVPR25]

Vision-Zephyr: a multimodal LLM for Visual Commonsense Reasoning—CLIP-ViT + Zephyr-7B with visual prompting; code, training scripts, and VCR evaluation.