List your company

Physical Reasoning on PACC (test)

56.2CoherenceGemini 2.5 Flash
-2.24812.92628.143.274May 6, 2026

Evaluation results

MethodLinks
Gemini 2.5 FlashModel Category=Proprietary, Evaluation Protocol=zero-shot, Frame Sampling=16 frames2026.0556.252.955.64045.22527.24040.91
PhyARModel Category=Open-Source, Backbone=VideoLLaMA3-7B, Evaluation Protocol=fine-tuned, Frame Sampling=16 frames, Training Strategy=LoRA, Epochs=32026.055052.922.25028.621.468.27041.56
GPT-4oModel Category=Proprietary, Evaluation Protocol=zero-shot, Frame Sampling=16 frames2026.0543.841.255.65035.721.445.54038.31
Qwen3.5-35BModel Category=Open-Source, Precision=16-bit, Evaluation Protocol=zero-shot, Frame Sampling=16 frames2026.056.217.611.13016.710.79.11013.64
Flash-VStream-7BModel Category=Open-Source, Precision=16-bit, Evaluation Protocol=zero-shot, Frame Sampling=16 frames2026.050000004.500.65
Video-ChatGPT-7BModel Category=Open-Source, Precision=16-bit, Evaluation Protocol=zero-shot, Frame Sampling=16 frames2026.0505.90000000.65
InternVL2.5-8BModel Category=Open-Source, Precision=16-bit, Evaluation Protocol=zero-shot, Frame Sampling=16 frames2026.0505.90104.80002.6
Video-LLaVA-7BModel Category=Open-Source, Precision=16-bit, Evaluation Protocol=zero-shot, Frame Sampling=16 frames2026.0500004.87.1002.6
MiniCPM-V 2.6Model Category=Open-Source, Precision=16-bit, Evaluation Protocol=zero-shot, Frame Sampling=16 frames2026.0505.90104.83.64.503.9
VideoLLaMA2-7BModel Category=Open-Source, Precision=16-bit, Evaluation Protocol=zero-shot, Frame Sampling=16 frames2026.0505.901011.93.64.505.84
VideoLLaMA3-7BModel Category=Open-Source, Precision=16-bit, Evaluation Protocol=zero-shot, Frame Sampling=16 frames2026.0505.901014.33.69.1107.79