List your company

Maze Solving on Wilson maze (test)

100Plan AccuracyCorrect Traces
-2.64824.00150.6577.299May 19, 2025

Evaluation results

MethodLinks
Correct TracesBackbone=Qwen3-8B-base, Fine-tuning regime=Correct Traces2025.05100100
Solution-onlyBackbone=Qwen3-8B-base, Fine-tuning regime=Solution-only2025.0599.70
Swapped TracesBackbone=Qwen3-8B-base, Fine-tuning regime=Swapped Traces2025.05980
Base ModelBackbone=Qwen3-8B-base, Fine-tuning regime=None2025.051.30