Structural Understanding on PolyBench 1.0 (test)
87EMPolyLM (Qwen-2.5-14B)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PolyLM (Qwen-2.5-14B)category=PolyBench, CoT=true2026.01 | 87 | 37 | 4.63 | |
| PolyLM (Phi-4-14B)category=PolyBench, CoT=false2026.01 | 80 | 13 | 4.43 | |
| PolyLM (Phi-4-14B)category=PolyBench, CoT=true2026.01 | 80 | 18 | 4.72 | |
| GPT-5category=Closed-source, CoT=false2026.01 | 73 | 51 | 4.88 | |
| PolyLM (Qwen-2.5-14B)category=PolyBench, CoT=false2026.01 | 70 | 11 | 3.46 | |
| Phi-4-14Bcategory=Off-the-shelf, CoT=false2026.01 | 67 | 12 | 4.43 | |
| PolyLM (ChemLLM-7B)category=PolyBench, CoT=true2026.01 | 67 | 32 | 4.3 | |
| PolyLM (Qwen-2.5-7B)category=PolyBench, CoT=true2026.01 | 60 | 24 | 3.91 | |
| GPT-4ocategory=Closed-source, CoT=false2026.01 | 60 | 29 | 3.71 | |
| PolyLM (Qwen-2.5-7B)category=PolyBench, CoT=false2026.01 | 53 | 27 | 4.53 | |
| 4o-Minicategory=Closed-source, CoT=false2026.01 | 53 | 17 | 3.95 | |
| Qwen-2.5-14Bcategory=Off-the-shelf, CoT=false2026.01 | 50 | 11 | 4.15 | |
| Qwen-2.5-7Bcategory=Off-the-shelf, CoT=false2026.01 | 47 | 12 | 3.78 | |
| ChemLLM-7Bcategory=Domain aligned, CoT=false2026.01 | 47 | 11 | 3.34 | |
| PolyLM (ChemLLM-7B)category=PolyBench, CoT=false2026.01 | 47 | 13 | 4.1 | |
| Sonnetcategory=Closed-source, CoT=false2026.01 | 47 | 26 | 4.74 | |
| ether0-24Bcategory=Domain aligned, CoT=false2026.01 | 38 | 7 | 3.26 | |
| Llamole-8Bcategory=Domain aligned, CoT=false2026.01 | 35 | 9 | 2.5 | |
| LlaSMol-8Bcategory=Domain aligned, CoT=false2026.01 | 27 | 8 | 3.34 |