VLM2Vec: Training Vision-Language Models for Massive Multimodal Embedding Tasks

Benchmark Model Rank Results
representation-learning-on-mmeb-v2VLM2Vec-V1-Qwen2VL-2B–MMEB-v2 Score: 42.69
representation-learning-on-mmeb-v2VLM2Vec-V1-Qwen2VL-7B–MMEB-v2 Score: 47.0
representation-learning-on-mmeb-v3VLM2Vec-V1-Qwen2VL-2B–MMEB-v3 Score: 17.53
representation-learning-on-mmeb-v3VLM2Vec-V1-Qwen2VL-7B–MMEB-v3 Score: 19.29