Triplet Accuracy on Deliberation Evaluation Suite GSC, Remesh, Polis (test)
81.8AbGST5-XL + DPT
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ST5-XL + DPTBackbone=ST5-XL, Tuning=Decorrelated Preference Tuning2026.05 | 81.8 | 76.4 | 65.9 | 67.5 | 65.6 | 69.5 | 69.8 | 65.1 | 68.4 | 62.5 | 61.9 | 68.6 | |
| ST5-XLBackbone=ST5-XL2026.05 | 81.1 | 70.5 | 69.7 | 65.8 | 61.1 | 69.9 | 60.7 | 57.6 | 56.4 | 55.5 | 69.2 | 65.2 | |
| OAI text-emb-3-largeProvider=OpenAI2026.05 | 79.6 | 74.5 | 56.1 | 63.9 | 61.1 | 71.4 | 58.1 | 57.7 | 51.6 | 50 | 68.1 | 62.9 | |
| ST5-LargeBackbone=ST5-Large2026.05 | 75.5 | 65.5 | 57.2 | 64.6 | 59.9 | 69.2 | 58.8 | 58.1 | 49.9 | 56.8 | 72.2 | 62.5 | |
| Voyage-4Provider=Voyage AI2026.05 | 72.3 | 66.6 | 62.1 | 58.4 | 56.7 | 67 | 54.9 | 57.4 | 50.8 | 50.1 | 64.5 | 60.1 | |
| Voyage-4-liteProvider=Voyage AI2026.05 | 70.1 | 64.5 | 61.3 | 58.2 | 56.8 | 65.4 | 53.3 | 57 | 51.2 | 49.3 | 65.5 | 59.3 | |
| Voyage-4-largeProvider=Voyage AI2026.05 | 69.9 | 69 | 63.7 | 60 | 57.9 | 69 | 55.4 | 58.3 | 50.3 | 48.3 | 63.6 | 60.5 | |
| ST5-BaseBackbone=ST5-Base2026.05 | 68.6 | 62.5 | 47.1 | 63.7 | 59 | 66.6 | 58.4 | 56.9 | 51.1 | 56 | 68.8 | 59.9 | |
| Voyage-3-largeProvider=Voyage AI2026.05 | 68.3 | 61 | 60.6 | 57.2 | 57.5 | 67.6 | 58.5 | 57.6 | 50.7 | 49.8 | 63.5 | 59.3 | |
| GTE-largeVariant=Non-instruct2026.05 | 67.6 | 65.6 | 59.5 | 60.2 | 56.3 | 67.7 | 54.7 | 55.5 | 46.2 | 51.6 | 69.5 | 59.5 | |
| BGE-large-enVersion=v1.5, Prefix=Represent this sentence...2026.05 | 64.7 | 59.5 | 56.5 | 62.5 | 57.6 | 68 | 54.2 | 55.5 | 53.2 | 52.9 | 66.3 | 59.2 | |
| e5-large-v2Prefix=query/passage2026.05 | 63.9 | 63.6 | 51.8 | 65.1 | 59.4 | 70.5 | 53.5 | 57.1 | 56.2 | 57.5 | 66.1 | 60.4 | |
| MxBAI-embed-largePrefix=Represent this sentence...2026.05 | 63.9 | 62.4 | 55.1 | 62.7 | 59.9 | 69.2 | 54.2 | 56.2 | 50.5 | 54.5 | 70 | 59.9 | |
| OAI text-emb-3-smallProvider=OpenAI2026.05 | 63.8 | 65.3 | 51.9 | 62.1 | 58.2 | 70.8 | 56.2 | 57 | 50.3 | 50.4 | 66 | 59.3 | |
| Voyage-3Provider=Voyage AI2026.05 | 63.8 | 50.4 | 57.5 | 58 | 54.9 | 65.2 | 53 | 55.5 | 50.6 | 45.8 | 65.2 | 56.4 | |
| BGE-SparseCL (arguana)Pooling=mean2026.05 | 62.4 | 61.7 | 42.7 | 59.2 | 57.2 | 66.1 | 50.1 | 56.1 | 57.5 | 52.6 | 63 | 57.2 | |
| all-distilrobertaArchitecture=distilroberta2026.05 | 61.8 | 55.8 | 53.4 | 59.5 | 57.5 | 67.6 | 53.7 | 55.4 | 50.5 | 48.5 | 65.8 | 57.2 | |
| Qwen2-1.5B-instructBackbone=Qwen2-1.5B, Tuning=Instruction-tuned2026.05 | 59.6 | 50.8 | 49.7 | 56.5 | 54.9 | 58 | 48.5 | 50.9 | 49.4 | 50.5 | 62.7 | 53.8 | |
| StanceAware-SBERTArchitecture=SBERT2026.05 | 58.5 | 58.2 | 36.4 | 64.8 | 59.7 | 68.2 | 68.1 | 56.1 | 52.5 | 52.3 | 68.3 | 58.5 | |
| Paraphrase-MiniLMBase Model=MiniLM2026.05 | 58.2 | 67.8 | 51.7 | 59.7 | 58.3 | 67.5 | 51.7 | 56.4 | 53.3 | 52.4 | 64.7 | 58.3 | |
| all-MiniLM-L12Layers=122026.05 | 57.8 | 57.1 | 47.6 | 59.2 | 57.8 | 67.2 | 52 | 55.3 | 49.9 | 48.8 | 64.1 | 56.1 | |
| all-mpnet-baseArchitecture=mpnet2026.05 | 57.1 | 55.6 | 60.4 | 58.5 | 57 | 69.2 | 54.4 | 56.4 | 47.1 | 52.4 | 68 | 57.8 | |
| Arctic-embed-lVersion=v2.0, Prefix=query/passage2026.05 | 56.4 | 62.4 | 58.5 | 57.6 | 56.2 | 67.5 | 53.4 | 57.6 | 50.2 | 47.3 | 64.2 | 57.4 | |
| NLI-mpnetBase Model=mpnet2026.05 | 56.1 | 66.9 | 47.6 | 65.6 | 61 | 69.5 | 58.3 | 56.9 | 52.8 | 55.5 | 68.7 | 59.9 | |
| all-MiniLM-L6Layers=62026.05 | 55.4 | 52.3 | 48.9 | 60.2 | 57.6 | 67.1 | 53.6 | 55.1 | 48.4 | 48.1 | 65.4 | 55.7 | |
| Stella-en-1.5BVersion=v5, Backbone=Stella-1.5B, Tuning=Instruction-tuned2026.05 | 43.5 | 50.2 | 65.3 | 55 | 55.1 | 54 | 53.4 | 49.1 | 47.4 | 57 | 49.7 | 52.7 |