Genetic Circuit Design on Cello-ID
66.2Task Success Rate (TSR)Qwen3-8B RLVF-H-C
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-8B RLVF-H-CTraining Protocol=GenCircuit-RL (Hierarchical-Curriculum)2026.05 | 66.2 | |
| Gemma-3-12B RLVF-H-CTraining Protocol=GenCircuit-RL (Hierarchical-Curriculum)2026.05 | 59 | |
| Qwen3-8B RLVF-HIERTraining Protocol=RLVF-Hierarchical2026.05 | 58.3 | |
| Llama-3.1-8B RLVF-H-CTraining Protocol=GenCircuit-RL (Hierarchical-Curriculum)2026.05 | 57.4 | |
| Qwen3-8B RLVF-BINTraining Protocol=RLVF-Binary2026.05 | 50.4 | |
| Qwen3-8B SFTTraining Protocol=Supervised fine-tuning2026.05 | 46.2 | |
| Gemma-3-12B SFTTraining Protocol=Supervised fine-tuning2026.05 | 41.2 | |
| Llama-3.1-8B SFTTraining Protocol=Supervised fine-tuning2026.05 | 39.8 | |
| OPUS 4.5Few-shot Setting=5-shot2026.05 | 36.5 | |
| OPUS 4.5Few-shot Setting=0-shot2026.05 | 25.4 | |
| Qwen3-8B BASEFew-shot Setting=0-shot2026.05 | 8.2 |