Evaluation records for new models in OpenVLMRecords

[saxenarohit]

Thank you for sharing the evaluation kit and the evaluation records.

Would it be possible to update the records in OpenVLMRecords to include newer models such as InternVL3?

https://huggingface.co/datasets/VLMEval/OpenVLMRecords/tree/main/mmeval

[FangXinyu-0913]

Hi @saxenarohit, the newest evaluation result is already updated. You can download it from http://opencompass.openxlab.space/assets/OpenVLM.json

[saxenarohit]

Hi @FangXinyu-0913
Thanks. The OpenVLM JSON only provides average results. I wanted to check whether detailed results like the per-sample predictions found in the mmeval folder are also available somewhere for newer models.