List your company

Proactive Assistance on Pro2Bench Ego4D

86G-Mean F1Qwen3.6-VL-27B
4.8825.944768.06Jun 3, 2026

Evaluation results

MethodLinks
Qwen3.6-VL-27BFT=true, PWR=true, Architecture=split duplex+planner2026.068664
Llama 4-MaverickFT=true, PWR=true, Architecture=split duplex+planner2026.068574
Llama 4-MaverickFT=true, Architecture=monolithic2026.067048
Claude Opus 4.6Architecture=monolithic, Mode=zero-shot2026.065437
Gemini 3.1 ProArchitecture=monolithic, Mode=zero-shot2026.065136
Qwen3.6-VL-27BArchitecture=monolithic2026.065143
Llama 4-MaverickArchitecture=monolithic2026.065135
Qwen3-VL-235BArchitecture=monolithic2026.065037
Qwen3.6-VL-27BFT=true, Architecture=monolithic2026.065035
GPT-5.2Architecture=monolithic, Mode=zero-shot2026.068—