List your company

Constrained Path Planning on Textualized Gridworld Long targets (test)

8.4Success RateSAE-opt
2.684.1655.657.135May 21, 2026

Evaluation results

MethodLinks
SAE-optModel=Qwen3-4B-Instruct2026.058.421.6
SAE-SSVModel=Qwen3-4B-Instruct2026.058.224.4
Dense-optModel=Qwen3-4B-Instruct2026.057.824.7
SAE-opt-anchModel=Qwen3-4B-Instruct2026.057.418.9
ITIModel=Qwen3-4B-Instruct2026.056.519.2
SAE-optModel=LLaMA3.1-8B-Instruct2026.055.942.2
CAAModel=Qwen3-4B-Instruct2026.055.619.9
SAE-optModel=Phi-3-Mini-4K-Instruct2026.055.252.2
Dense-optModel=LLaMA3.1-8B-Instruct2026.05548.8
CAAModel=Phi-3-Mini-4K-Instruct2026.054.938.2
SAE-SSVModel=Phi-3-Mini-4K-Instruct2026.054.955.8
Dense-optModel=Phi-3-Mini-4K-Instruct2026.054.955.6
SAE-opt-anchModel=LLaMA3.1-8B-Instruct2026.054.947
SAE-SSVModel=LLaMA3.1-8B-Instruct2026.054.848.8
CAAModel=LLaMA3.1-8B-Instruct2026.054.133.9
ITIModel=LLaMA3.1-8B-Instruct2026.05443.7
SAE-opt-anchModel=Phi-3-Mini-4K-Instruct2026.05455.5
ITIModel=Phi-3-Mini-4K-Instruct2026.053.340.3
ICLModel=Qwen3-4B-Instruct2026.053.276
ICLModel=LLaMA3.1-8B-Instruct2026.053.262.2
ICLModel=Phi-3-Mini-4K-Instruct2026.052.989.1