List your company

Open-ended Reasoning on private CCTV 80 videos (test)

39.47Rescaled BertScore F1+ Dashcam SFT
14.40620.91327.4233.927May 21, 2026

Evaluation results

MethodLinks
+ Dashcam SFTTraining=SFT2026.0539.47
+ CCTV SFTTraining=SFT2026.0539.45
+ RLTraining=SFT+RL2026.0537.29
Gemini 3.1 FlashTraining=0-shot2026.0523.2
Gemini 2.5 ProTraining=0-shot2026.0515.6
CR2Training=0-shot2026.0515.37