Robotic Planning on LEMMA Stack and Pass tasks (test)
75Success RateSG-CoT
Evaluation Results
| Method | Links | |
|---|---|---|
| SG-CoTLLM Backbone=Gemini-2.5-Flash2026.03 | 75 | |
| ProgPromptLLM Backbone=Gemini-2.5-Flash2026.03 | 59 | |
| SG-CoTLLM Backbone=Qwen3-VL-2B2026.03 | 59 | |
| InnerMonoLLM Backbone=Gemini-2.5-Flash2026.03 | 48 | |
| InnerMonoLLM Backbone=Qwen3-VL-2B2026.03 | 45 | |
| CLARALLM Backbone=Gemini-2.5-Flash2026.03 | 45 | |
| ProgPromptLLM Backbone=Qwen3-VL-2B2026.03 | 44 | |
| CLARALLM Backbone=Qwen3-VL-2B2026.03 | 39 |