SEED-Bench

SEED-Bench Card Benchmark details Benchmark type: SEED-Bench is a large-scale benchmark to evaluate Multimodal Large Language Models (MLLMs). It consists of 19K multiple choice questions with accurate human annotations, which covers 12 evaluation dimensions including the comprehension of both the image and video modali

Papers using SEED-Bench (9)

Make Your LVLM KV Cache More Lightweight2026LLMind: Bio-inspired Training-free Adaptive Visual Representations for Vision-Language Models2026Same Answer, Different Representations: Hidden instability in VLMs2026Vision to Geometry: 3D Spatial Memory for Sequential Embodied MLLM Reasoning and Exploration2025HybridToken-VLM: Hybrid Token Compression for Vision-Language Models2025InternLM-XComposer: A Vision-Language Large Model for Advanced Text-image Comprehension and Composition2023 · 31 citesEE-MLLM: A Data-Efficient and Compute-Efficient Multimodal Large Language Model2024 · 1 citesDynamic Multimodal Evaluation with Flexible Complexity by Vision-Language Bootstrapping2024Exploring Multi-Grained Concept Annotations for Multimodal Large Language Models2024