List your company
ResearchFormerly WizWand

Puzzle Reasoning on FrozenLake

82Success Rateo4-mini
11.2829.644866.36May 5, 2026
Updated 6mo ago

Evaluation results

MethodLinks
o4-mini2026.0582
Claude 4.5 Sonnet2026.0580
GLANCE-FullBackbone=Qwen2.5-VL-3B, Reward Strategy=Dense Extrinsic, Reasoning Strategy=World Model Reasoning for Visual States2026.0578
Gemini 2.5 Pro2026.0578
GLANCE-BaseBackbone=Qwen2.5-VL-3B, Reward Strategy=Sparse Extrinsic, Reasoning Strategy=World Model Reasoning Strategy2026.0573
VAGEN-FullBackbone=Qwen2.5-VL-3B, Reward Strategy=Dense Extrinsic, Reasoning Strategy=World Model Reasoning for Visual States2026.0572
VAGEN-BaseBackbone=Qwen2.5-VL-3B, Reward Strategy=Sparse Extrinsic, Reasoning Strategy=World Model Reasoning Strategy2026.0571
GLANCE w/ Turn-PPOBackbone=Qwen2.5-VL-3B, Reward Strategy=Turn-level PPO, Reasoning Strategy=World Model Reasoning Strategy2026.0570
Claude 3.7 Sonnet2026.0569
Turn-PPO w/ MaskBackbone=Qwen2.5-VL-3B, Reward Strategy=Turn-level PPO, Reasoning Strategy=World Model Reasoning Strategy2026.0568
GRPO w/ MaskBackbone=Qwen2.5-VL-3B, Reward Strategy=RL Baseline, Reasoning Strategy=World Model Reasoning Strategy2026.0557
GPT-4o2026.0554
Qwen2.5-VL-72BBackbone=Qwen2.5-VL-72B2026.0544
Vanilla-PPOBackbone=Qwen2.5-VL-3B, Reward Strategy=RL Baseline, Reasoning Strategy=World Model Reasoning Strategy2026.0521
VLM-R1-3BBackbone=VLM-R1-3B2026.0515
Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2026.0514
Qwen2.5-VL-3BBackbone=Qwen2.5-VL-3B2026.0514