Terminology Understanding on Chemistry Domain Dataset
80.2Recall@10LBR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| LBRBackbone=Qwen2-1.5B, Attention Architecture=Causal, Training Paradigm=LLM+IB-GL+CL2026.01 | 80.2 | 72.7 | 87.82 | |
| LBRBackbone=Qwen2.5-1.5B, Attention Architecture=Causal, Training Paradigm=LLM+IB-GL+CL2026.01 | 79.7 | 72.3 | 87.91 | |
| GTEBackbone=Qwen2-1.5B, Attention Architecture=Bi., Training Paradigm=LLM+CL2026.01 | 76.6 | 67.7 | 65.26 | |
| LBRBackbone=Llama3.2-1B, Attention Architecture=Causal, Training Paradigm=LLM+IB-GL+CL2026.01 | 72.2 | 61.9 | 84.53 | |
| LLM2VecBackbone=SLLaMA-1.3B, Attention Architecture=Bi., Training Paradigm=LLM+CL2026.01 | 71.2 | 61.5 | 79.34 | |
| BGEBackbone=XLMR-Large, Attention Architecture=Bi., Training Paradigm=LLM+CL2026.01 | 68.9 | 55.7 | 78.35 | |
| SFT+CLBackbone=Qwen2-1.5B, Attention Architecture=Causal, Training Paradigm=LLM+GL+CL2026.01 | 44.9 | 40.5 | 64.47 | |
| SFT+CLBackbone=Qwen2.5-1.5B, Attention Architecture=Causal, Training Paradigm=LLM+GL+CL2026.01 | 43.6 | 39.1 | 60.68 | |
| ChemLLMBackbone=InternLM-2B, Attention Architecture=Causal, Training Paradigm=LLM+GL2026.01 | 23.7 | 16.8 | 11.012 | |
| HuaTuoBackbone=LLaMA-7B, Attention Architecture=Causal, Training Paradigm=LLM+GL2026.01 | 20.1 | 18.8 | 31.29 | |
| Qwen2.5Backbone=Qwen2.5-1.5B, Attention Architecture=Causal, Training Paradigm=LLM+GL2026.01 | 14.6 | 11.5 | 18.311 | |
| Qwen2Backbone=Qwen2-1.5B, Attention Architecture=Causal, Training Paradigm=LLM+GL2026.01 | 10 | 7.8 | 20.01 |