LLaVA

LLaVA is a dataset/benchmark used to evaluate the performance of Multimodal Large Language Models (MLLMs) on vision-language tasks, focusing on their internal visual representations and interpretability.

Papers using LLaVA (10)

Cascaded Sparse Autoencoders Learn Multi-Level Visual Concepts in Multimodal LLMs2026Behind Maya: Building A Multilingual Vision Language Model2025DeepSight: Bridging Depth Maps and Language with a Depth-Driven Multimodal Model2026HiPP-Prune: Hierarchical Preference-Conditioned Structured Pruning for Vision-Language Models2026Seeing Right but Saying Wrong: Inter- and Intra-Layer Refinement in MLLMs without Training2026Gated Relational Alignment via Confidence-based Distillation for Efficient VLMs2026Safe-llava: A Privacy-preserving Vision-language Dataset And Benchmark For Biometric Safety2025DeepInsert: Early Layer Bypass for Efficient and Performant Multimodal Understanding2025LVLM-Compress-Bench: Benchmarking the Broader Impact of Large Vision-Language Model Compression2025Maya: An Instruction Finetuned Multilingual Multimodal Model2024