Functional Group Reasoning (Numeric Regression) on FGBench Comparison
48.344RMSELFM2-2.6B-MMAI
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| LFM2-2.6B-MMAITraining=MT, Reasoning=/think2026.03 | 48.344 | 100 | |
| LFM2-2.6B-MMAITraining=ST, Reasoning=/think2026.03 | 50.598 | 100 | |
| LFM2-2.6B-MMAITraining=ST, Reasoning=/no think2026.03 | 53.045 | 100 | |
| LFM2-2.6B-MMAITraining=MT, Reasoning=/no think2026.03 | 53.127 | 100 | |
| DeepSeek-3-2Class=Open-weight LLM2026.03 | 55.504 | 64.3 | |
| LFM2-2.6BClass=Liquid Foundation Model2026.03 | 61.556 | 100 | |
| GPT-5-1Class=Proprietary LLM2026.03 | 63.233 | 100 | |
| Sonnet-4-5Class=Proprietary LLM2026.03 | 64.468 | 100 | |
| Llama-3.1Parameters=70B2026.03 | 64.887 | 94.3 | |
| ChemLLM-7BClass=Foundational LLM for Chemistry2026.03 | 65.428 | 51.4 | |
| Qwen2.5-7BClass=Open-weight LLM2026.03 | 65.471 | 22.3 | |
| GPT-5-2Class=Proprietary LLM2026.03 | 66.223 | 100 | |
| Opus-4-5Class=Proprietary LLM2026.03 | 71.679 | 100 | |
| Llama-3.1Parameters=8B2026.03 | 80.566 | 76.1 | |
| Llama-3-8B-MolInstClass=Foundational LLM for Chemistry2026.03 | 138.654 | 83.7 | |
| LlaSMol-Mistral-7BClass=Foundational LLM for Chemistry2026.03 | 245.298 | 92.4 |