Climate Policy Analysis on GCAM tasks (100 tasks)
0.6Latency Range MinPiERN-GCAM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PiERN-GCAMLLM=Qwen2.5-0.5B-Instruct, Hardware=NVIDIA GeForce RTX 4090 GPU, Routing mechanism=token-level routing, Modules=text-to-computation module and pre-trained neural surrogate experts2025.09 | 0.6 | 125 | 2,889 | |
| Agent-ExpertLLM=Qwen2.5-0.5B-Instruct, Orchestration framework=AutoGen 0.10.4, Execution=lightweight pre-trained expert models2025.09 | 1.9 | 249 | 1,443 | |
| Agent-LegacyLLM=Qwen2.5-0.5B-Instruct, Orchestration framework=AutoGen 0.10.4, Execution=original GCAM C++ kernel2025.09 | 3,605 | 126.73 | 0.79 |