Paper summary
Paper summary: Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling
InternVL 2.5 keeps InternVL 2.0's architecture but improves training, testing and data, and studies scaling of vision encoders, LLMs, data and test-time settings; it's reported to rival GPT-4o and Claude-3.5-Sonnet.
- Same architecture, better recipe: the gains come from training strategy, test-time configuration and data quality.
- The paper reports it as the first open-source MLLM to pass 70% on the MMMU benchmark.