List your company

Maze Solving on Searchformer maze (test)

6Plan AccuracyCorrect Traces
-0.0321.5343.14.666May 19, 2025

Evaluation results

MethodLinks
Correct TracesBackbone=Qwen3-8B-base, Fine-tuning regime=Correct Traces2025.0566.6
Swapped TracesBackbone=Qwen3-8B-base, Fine-tuning regime=Swapped Traces2025.050.70
Base ModelBackbone=Qwen3-8B-base, Fine-tuning regime=None2025.050.20
Solution-onlyBackbone=Qwen3-8B-base, Fine-tuning regime=Solution-only2025.050.20