Natural Language Inference on SICK

90.9Accuracy

Finetuned

Updated 20d ago

Evaluation Results

Method
Finetuned 2026.03	90.9	-	-
Full dataset tuning 2023.05	88.1	-	-
BERT + Aligner 2021.06	87.2	-	-
BERT 2021.06	86.2	-	-
32-shot tuning 2023.05	78	-	-
Training-free Weight Recycling (L^T_KD) 2023.05	75.2	-	-
Training-free Weight Recycling (L^W_KD) 2023.05	74.3	-	-
Demonstration learning 2023.05	60.5	-	-
DPS Mix 2022.11	58.18	-	-
DPS Dense 2022.11	56.23	-	-
DPS Dense 2022.11	55.76	-	-
CHILD-TUNINGD 2022.11	55.69	-	-
DPS Mix 2022.11	55.49	-	-
vanilla 2022.11	53.67	-	-
CHILD-TUNINGD 2022.11	53.46	-	-
vanilla 2022.11	51.91	-	-
GPT-J (Original) 2023.05	-	25.6	-
GPT-J (CC) 2023.05	-	41.1	-
GPT-J (DC) 2023.05	-	41.4	-
TA 2026.03	-	-	83.8
TIES 2026.03	-	-	86.6
DARE-TIES 2026.03	-	-	86.6
SVD 2026.03	-	-	80
Linear 2026.03	-	-	83.8
KnOTS-TIES 2026.03	-	-	80.2
KnOTS-DARE-TIES 2026.03	-	-	80.1
LoRA-LEGO 2026.03	-	-	86.1
EMR-Merging 2026.03	-	-	76.6
FR-Merging 2026.03	-	-	80.3
RobustMerge 2026.03	-	-	83
AdaMerging 2026.03	-	-	89.2
NSC 2026.03	-	-	91.2