Natural Language Inference on SICK
91.58AccuracyIndividual Task
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Individual TaskBackbone=Llama-3, Evaluation Protocol=fine-tuned on single tasks2026.04 | 91.58 | — | — | |
| FinetunedBackbone=LLAMA-3 8B, LoRA rank=162026.03 | 90.9 | — | — | |
| KnOTS + ENMPBackbone=Llama-3, Scaling factor (λ)=1.1, Top-k (%)=90, ENMP (Elective Neuron Modification Protocol)=true2026.04 | 88.47 | — | — | |
| Full dataset tuningTraining=Full dataset (FD)2023.05 | 88.1 | — | — | |
| CoreSpace + ENMPBackbone=Llama-3, Scaling factor (λ)=0.5, ENMP (Elective Neuron Modification Protocol)=true2026.04 | 87.92 | — | — | |
| BERT + AlignerBackbone=BERT-base, Evaluation Protocol=Fine-tuned, Alignment Feature=Conditional alignment probability2021.06 | 87.2 | — | — | |
| TSV + ENMPBackbone=Llama-3, Scaling factor (λ)=0.6, ENMP (Elective Neuron Modification Protocol)=true2026.04 | 86.45 | — | — | |
| BERTBackbone=BERT-base, Evaluation Protocol=Fine-tuned2021.06 | 86.2 | — | — | |
| TIES + ENMPBackbone=Llama-3, Scaling factor (λ)=1.2, Top-k (%)=80, ENMP (Elective Neuron Modification Protocol)=true2026.04 | 84.92 | — | — | |
| DARE + ENMPBackbone=Llama-3, Scaling factor (λ)=1.1, Top-k (%)=80, Drop rate (p)=0.1, ENMP (Elective Neuron Modification Protocol)=true2026.04 | 84.56 | — | — | |
| TA + ENMPBackbone=Llama-3, Scaling factor (λ)=0.3, ENMP (Elective Neuron Modification Protocol)=true2026.04 | 83.88 | — | — | |
| KnOTSBackbone=Llama-3, Scaling factor (λ)=1.1, Top-k (%)=902026.04 | 82.12 | — | — | |
| CoreSpaceBackbone=Llama-3, Scaling factor (λ)=0.52026.04 | 81.74 | — | — | |
| TSVBackbone=Llama-3, Scaling factor (λ)=0.62026.04 | 81.37 | — | — | |
| TABackbone=Llama-3, Scaling factor (λ)=0.32026.04 | 80.6 | — | — | |
| 32-shot tuningTraining=32-shot (FS)2023.05 | 78 | — | — | |
| Training-free Weight Recycling (L^T_KD)Distillation Data=Unlabeled Task Data, Source Dataset=MNLI2023.05 | 75.2 | — | — | |
| Training-free Weight Recycling (L^W_KD)Distillation Data=Wikipedia, Source Dataset=MNLI2023.05 | 74.3 | — | — | |
| TIESBackbone=Llama-3, Scaling factor (λ)=1.2, Top-k (%)=802026.04 | 73.99 | — | — | |
| DAREBackbone=Llama-3, Scaling factor (λ)=1.1, Top-k (%)=80, Drop rate (p)=0.12026.04 | 71.14 | — | — | |
| MICBackbone=BERT, Dimension=5122026.05 | 70.84 | — | — | |
| MICBackbone=BERT, Dimension=7682026.05 | 70.75 | — | — | |
| MICBackbone=BERT, Dimension=2562026.05 | 70.51 | — | — | |
| MICBackbone=BERT, Dimension=1282026.05 | 70.29 | — | — | |
| MICBackbone=TinyBERT 6L, Dimension=7682026.05 | 70.14 | — | — | |
| MICBackbone=TinyBERT 6L, Dimension=5122026.05 | 70.13 | — | — | |
| Unsup SimCSEBackbone=TinyBERT 6L, Dimension=7682026.05 | 69.93 | — | — | |
| MICBackbone=TinyBERT 6L, Dimension=2562026.05 | 69.92 | — | — | |
| Unsup SimCSEBackbone=TinyBERT 6L, Dimension=5122026.05 | 69.87 | — | — | |
| MICBackbone=TinyBERT 6L, Dimension=1282026.05 | 69.66 | — | — | |
| MRLBackbone=TinyBERT 6L, Dimension=7682026.05 | 69.53 | — | — | |
| MRLBackbone=TinyBERT 6L, Dimension=5122026.05 | 69.47 | — | — | |
| MICBackbone=BERT, Dimension=642026.05 | 69.09 | — | — | |
| MRLBackbone=BERT, Dimension=7682026.05 | 68.94 | — | — | |
| Unsup SimCSEBackbone=TinyBERT 6L, Dimension=2562026.05 | 68.93 | — | — | |
| Unsup SimCSEBackbone=TinyBERT 6L, Dimension=1282026.05 | 68.91 | — | — | |
| MRLBackbone=TinyBERT 6L, Dimension=1282026.05 | 68.88 | — | — | |
| MRLBackbone=TinyBERT 6L, Dimension=2562026.05 | 68.85 | — | — | |
| MICBackbone=TinyBERT 6L, Dimension=642026.05 | 68.82 | — | — | |
| MRLBackbone=BERT, Dimension=5122026.05 | 68.33 | — | — | |
| ESEBackbone=BERT, Dimension=7682026.05 | 68.19 | — | — | |
| MRLBackbone=TinyBERT 6L, Dimension=642026.05 | 68.17 | — | — | |
| Unsup SimCSEBackbone=BERT, Dimension=7682026.05 | 68.15 | — | — | |
| ESEBackbone=TinyBERT 6L, Dimension=7682026.05 | 67.76 | — | — | |
| ESEBackbone=TinyBERT 6L, Dimension=5122026.05 | 67.69 | — | — | |
| Unsup SimCSEBackbone=TinyBERT 6L, Dimension=642026.05 | 67.5 | — | — | |
| Unsup SimCSEBackbone=BERT, Dimension=5122026.05 | 67.41 | — | — | |
| MRLBackbone=BERT, Dimension=2562026.05 | 67.23 | — | — | |
| MICBackbone=BERT, Dimension=322026.05 | 67.08 | — | — | |
| ESEBackbone=BERT, Dimension=5122026.05 | 67.04 | — | — | |
| MICBackbone=TinyBERT 6L, Dimension=322026.05 | 67 | — | — | |
| ESEBackbone=TinyBERT 6L, Dimension=2562026.05 | 67 | — | — | |
| ESEBackbone=TinyBERT 6L, Dimension=642026.05 | 66.92 | — | — | |
| ESEBackbone=BERT, Dimension=1282026.05 | 66.74 | — | — | |
| ESEBackbone=TinyBERT 6L, Dimension=1282026.05 | 66.62 | — | — | |
| MRLBackbone=BERT, Dimension=1282026.05 | 66.57 | — | — | |
| ESEBackbone=BERT, Dimension=2562026.05 | 66.43 | — | — | |
| MRLBackbone=BERT, Dimension=642026.05 | 66.11 | — | — | |
| Unsup SimCSEBackbone=BERT, Dimension=2562026.05 | 65.72 | — | — | |
| MICBackbone=BERT, Dimension=162026.05 | 65.09 | — | — | |
| MRLBackbone=TinyBERT 6L, Dimension=322026.05 | 65.08 | — | — | |
| ESEBackbone=TinyBERT 6L, Dimension=322026.05 | 64.98 | — | — | |
| Unsup SimCSEBackbone=TinyBERT 6L, Dimension=322026.05 | 64.69 | — | — | |
| MICBackbone=TinyBERT 6L, Dimension=162026.05 | 64.3 | — | — | |
| MRLBackbone=BERT, Dimension=322026.05 | 63.34 | — | — | |
| ESEBackbone=TinyBERT 6L, Dimension=162026.05 | 62.56 | — | — | |
| ESEBackbone=BERT, Dimension=322026.05 | 62.36 | — | — | |
| ESEBackbone=BERT, Dimension=642026.05 | 62.07 | — | — | |
| ESEBackbone=BERT, Dimension=162026.05 | 61.5 | — | — | |
| Unsup SimCSEBackbone=TinyBERT 6L, Dimension=162026.05 | 60.66 | — | — | |
| Demonstration learningMode=Zero-shot with context2023.05 | 60.5 | — | — | |
| MRLBackbone=BERT, Dimension=162026.05 | 59.35 | — | — | |
| Unsup SimCSEBackbone=BERT, Dimension=1282026.05 | 59.33 | — | — | |
| MRLBackbone=TinyBERT 6L, Dimension=162026.05 | 59.07 | — | — | |
| DPS MixTraining Dataset=MNLI2022.11 | 58.18 | — | — | |
| DPS DenseTraining Dataset=SNLI2022.11 | 56.23 | — | — | |
| DPS DenseTraining Dataset=MNLI2022.11 | 55.76 | — | — | |
| CHILD-TUNINGDTraining Dataset=MNLI2022.11 | 55.69 | — | — | |
| DPS MixTraining Dataset=SNLI2022.11 | 55.49 | — | — | |
| vanillaTraining Dataset=SNLI2022.11 | 53.67 | — | — | |
| CHILD-TUNINGDTraining Dataset=SNLI2022.11 | 53.46 | — | — | |
| Unsup SimCSEBackbone=BERT, Dimension=642026.05 | 52.97 | — | — | |
| vanillaTraining Dataset=MNLI2022.11 | 51.91 | — | — | |
| Unsup SimCSEBackbone=BERT, Dimension=322026.05 | 51.86 | — | — | |
| Unsup SimCSEBackbone=BERT, Dimension=162026.05 | 45.75 | — | — | |
| AdaMergingBackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | — | 89.2 | |
| DARE-TIESBackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | — | 86.6 | |
| EMR-MergingBackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | — | 76.6 | |
| FR-MergingBackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | — | 80.3 | |
| GPT-J (CC)Backbone=GPT-J, Prompting=CC, Shots=8-shot2023.05 | — | 41.1 | — | |
| GPT-J (DC)Backbone=GPT-J, Prompting=DC, Shots=8-shot2023.05 | — | 41.4 | — | |
| GPT-J (Original)Backbone=GPT-J, Prompting=Original, Shots=8-shot2023.05 | — | 25.6 | — | |
| KnOTS-DARE-TIESBackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | — | 80.1 | |
| KnOTS-TIESBackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | — | 80.2 | |
| LinearBackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | — | 83.8 | |
| LoRA-LEGOBackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | — | 86.1 | |
| NSCBackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | — | 91.2 | |
| RobustMergeBackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | — | 83 | |
| SVDBackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | — | 80 | |
| TABackbone=LLAMA-3 8B, LoRA rank=162026.03 | — | — | 83.8 |