List your company

Cross-task physical reasoning on PHYRE Set 3 (test)

52Pass@1Human
-0.41613.19226.840.408Jul 7, 2026

Evaluation results

MethodLinks
HumanCategory=Oracle2026.075273.282
DQN-expertCategory=Oracle2026.0733.841.246.6
+EG+VAORABase Model=Qwen3VL-8B-SFT, Protocol=Expert-Guided + VAORA2026.073045.651.2
Gemini-3.1-ProCategory=Closed-Source API2026.0727.848.455.4
+EG (Expert-Guided)Base Model=Qwen3VL-8B-SFT, Protocol=Expert-Guided2026.0721.83439.2
Qwen3VL-8B (SFT)Category=Open-Sourced Baselines, Protocol=SFT2026.0717.636.845.2
Gemini-3.1-FlashCategory=Closed-Source API2026.0714.229.437.6
InternVL-3.5-8BCategory=Open-Sourced Baselines2026.07510.415.4
GPT-5.4Category=Closed-Source API2026.074.48.614
Claude-Sonnet-4.6Category=Closed-Source API2026.0737.410.8
Qwen3VL-8B (GRPO)Category=Open-Sourced Baselines, Protocol=GRPO2026.072.4610.8
Qwen3VL-8B (SFT+GRPO)Category=Open-Sourced Baselines, Protocol=SFT+GRPO2026.072.27.810.2
Qwen3VL-8BCategory=Open-Sourced Baselines2026.071.647.4