Formulation design on Formulation design
92.3ValidityGPT-5
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT-5Backbone model=GPT-52026.04 | 92.3 | 92.2 | |
| Claude-3.5Backbone model=Claude-3.52026.04 | 88.6 | 82.6 | |
| Qwen3-14B w SciDCBackbone model=Qwen3-14B, Framework configuration=w SciDC2026.04 | 75.5 | 68.3 | |
| Qwen3-4B w SciDCBackbone model=Qwen3-4B, Framework configuration=w SciDC2026.04 | 71 | 43.4 | |
| Qwen3-4B w/o KBackbone model=Qwen3-4B, Framework configuration=w/o K2026.04 | 56.8 | 56.8 | |
| Qwen3-14BBackbone model=Qwen3-14B2026.04 | 50.9 | 50.4 | |
| Qwen3-4BBackbone model=Qwen3-4B2026.04 | 43.4 | 43 | |
| Qwen3-14B w/o KBackbone model=Qwen3-14B, Framework configuration=w/o K2026.04 | 18.4 | 18.4 |