List your company

Maze Solving on Kruskal maze (test)

99.9Plan AccuracyCorrect Traces
-2.43624.13250.777.268May 19, 2025

Evaluation results

MethodLinks
Correct TracesBackbone=Qwen3-8B-base, Fine-tuning regime=Correct Traces2025.0599.9100
Solution-onlyBackbone=Qwen3-8B-base, Fine-tuning regime=Solution-only2025.0598.50
Swapped TracesBackbone=Qwen3-8B-base, Fine-tuning regime=Swapped Traces2025.0598.20
Base ModelBackbone=Qwen3-8B-base, Fine-tuning regime=None2025.051.50