Agentic Rollback and Planning on Primary grid n=1,008
3.79Quality ScoreOracle
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| OracleLLM=DeepSeek-V32026.04 | 3.79 | 0 | 0 | 18 | |
| Full RestartLLM=DeepSeek-V32026.04 | 3.17 | 11.41 | 3.44 | 30.2 | |
| AbsorberLLM=DeepSeek-V32026.04 | 3.07 | 0.78 | 0.78 | 27.6 | |
| Checkpoint-5LLM=DeepSeek-V32026.04 | 2.96 | 3.35 | 2.26 | 28 | |
| LangGraph-InterruptLLM=DeepSeek-V32026.04 | 2.9 | 0.74 | 0.74 | 29.4 | |
| NaiveLLM=DeepSeek-V32026.04 | 2.78 | 0 | 0 | 28.2 | |
| IgnoreLLM=DeepSeek-V32026.04 | 1.99 | 0 | 0 | 22.9 |