Semantic Textual Similarity on BIOSSES (test)
0.748Similarity ScoreBest-Layer
Evaluation Results
| Method | Links | |
|---|---|---|
| Best-LayerBase Model=Pythia-410m2026.03 | 0.748 | |
| Best-LayerBase Model=Llama3-8B2026.03 | 0.7283 | |
| Best-LayerBase Model=Gemma2-2B2026.03 | 0.7202 | |
| WeightedBase Model=Pythia-410m2026.03 | 0.6779 | |
| Last-LayerBase Model=Pythia-410m2026.03 | 0.673 | |
| Last-LayerBase Model=Llama3-8B2026.03 | 0.67 | |
| FC-EncoderBase Model=Gemma2-2B, GNN Variant=GCN2026.03 | 0.6602 | |
| WeightedBase Model=Gemma2-2B2026.03 | 0.6527 | |
| Cayley-EncoderBase Model=Gemma2-2B, GNN Variant=GIN2026.03 | 0.6407 | |
| FC-EncoderBase Model=Llama3-8B, GNN Variant=GCN2026.03 | 0.6281 | |
| Cayley-EncoderBase Model=Gemma2-2B, GNN Variant=GCN2026.03 | 0.6197 | |
| MLP Last-LayerBase Model=Pythia-410m2026.03 | 0.5968 | |
| FC-EncoderBase Model=Pythia-410m, GNN Variant=GCN2026.03 | 0.5865 | |
| Last-LayerBase Model=Gemma2-2B2026.03 | 0.5844 | |
| DWAttBase Model=Gemma2-2B2026.03 | 0.5708 | |
| Cayley-EncoderBase Model=Pythia-410m, GNN Variant=GIN2026.03 | 0.5659 | |
| MLP Last-LayerBase Model=Llama3-8B2026.03 | 0.5535 | |
| Cayley-EncoderBase Model=Llama3-8B, GNN Variant=GIN2026.03 | 0.5532 | |
| FC-EncoderBase Model=Gemma2-2B, GNN Variant=GIN2026.03 | 0.5488 | |
| MLP Last-LayerBase Model=Gemma2-2B2026.03 | 0.541 | |
| WeightedBase Model=Llama3-8B2026.03 | 0.5282 | |
| FC-EncoderBase Model=Pythia-410m, GNN Variant=GIN2026.03 | 0.5214 | |
| FC-EncoderBase Model=Llama3-8B, GNN Variant=GIN2026.03 | 0.4962 | |
| DWAttBase Model=Llama3-8B2026.03 | 0.4688 | |
| DWAttBase Model=Pythia-410m2026.03 | 0.4529 | |
| Set-EncoderBase Model=Pythia-410m2026.03 | 0.4472 | |
| Cayley-EncoderBase Model=Pythia-410m, GNN Variant=GCN2026.03 | 0.3932 | |
| Set-EncoderBase Model=Llama3-8B2026.03 | 0.3296 | |
| Set-EncoderBase Model=Gemma2-2B2026.03 | 0.3164 | |
| Cayley-EncoderBase Model=Llama3-8B, GNN Variant=GCN2026.03 | 0.2325 |