Question Answering on ARC-e (test)
96.07AccuracyBLoB
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BLoBParams (M)=5.4032026.06 | 96.07 | 2.35 | 0.15 | |
| BaRAParams (M)=5.6082026.06 | 96.04 | 1.35 | 0.12 | |
| FVAE-LoRAParams (M)=33.532026.06 | 95.89 | 1.74 | 0.13 | |
| ScalaBLParams (M)=3.7692026.06 | 95.74 | 1.92 | 0.14 | |
| EnsembleParams (M)=11.3052026.06 | 95.71 | 2.63 | 0.17 | |
| MLEParams (M)=3.7682026.06 | 95.64 | 3.48 | 0.23 | |
| LaplaceParams (M)=3.7682026.06 | 95.61 | 1.99 | 0.15 | |
| MAPParams (M)=3.7682026.06 | 95.53 | 3.31 | 0.22 | |
| MC-DropoutParams (M)=3.7682026.06 | 95.51 | 3.38 | 0.21 | |
| C-LoRAParams (M)=4.2752026.06 | 95.51 | 1.32 | 0.12 | |
| LLaMA-2Model Size=70B, Evaluation Protocol=LAT2023.10 | 92.6 | — | — | |
| Interaction-Aware Influence Function (Ours)Base Model=Llama-3.1-8B, Evaluation Protocol=Fine-tuned, Number of Seeds=52026.05 | 90.44 | — | — | |
| LESSBase Model=Llama-3.1-8B, Evaluation Protocol=Fine-tuned, Number of Seeds=52026.05 | 90.11 | — | — | |
| RandomBase Model=Llama-3.1-8B, Evaluation Protocol=Fine-tuned, Number of Seeds=52026.05 | 90.03 | — | — | |
| NV-EmbedBase Model=Llama-3.1-8B, Evaluation Protocol=Fine-tuned, Number of Seeds=52026.05 | 89.01 | — | — | |
| Additive IFBase Model=Llama-3.1-8B, Evaluation Protocol=Fine-tuned, Number of Seeds=52026.05 | 88.72 | — | — | |
| LLaMA-2Model Size=70B, Evaluation Protocol=Few-Shot2023.10 | 88.7 | — | — | |
| RDS+Base Model=Llama-3.1-8B, Evaluation Protocol=Fine-tuned, Number of Seeds=52026.05 | 88.7 | — | — | |
| LLaMA-2Model Size=13B, Evaluation Protocol=LAT2023.10 | 86.3 | — | — | |
| LLaMA-2Model Size=13B, Evaluation Protocol=Few-Shot2023.10 | 84.9 | — | — | |
| LLAMA-3.1-8BModel Architecture=LLAMA-3.1-8B, Configuration Type=Dense, Evaluation Protocol=Zero-shot, Sparsity Level=0%2026.01 | 82.53 | — | — | |
| QWEN3-14BModel Architecture=QWEN3-14B, Configuration Type=Dense, Evaluation Protocol=Zero-shot, Sparsity Level=0%2026.01 | 81.9 | — | — | |
| MISTRAL-7BModel Architecture=MISTRAL-7B, Configuration Type=Dense, Evaluation Protocol=Zero-shot, Sparsity Level=0%2026.01 | 81.73 | — | — | |
| LLaMA-2Model Size=7B, Evaluation Protocol=LAT2023.10 | 80.3 | — | — | |
| LLaMA-2Model Size=7B, Evaluation Protocol=Few-Shot2023.10 | 80.1 | — | — | |
| DEEPSEEK-R1-QWEN-8BModel Architecture=DEEPSEEK-R1-QWEN-8B, Configuration Type=Dense, Evaluation Protocol=Zero-shot, Sparsity Level=0%2026.01 | 76.52 | — | — | |
| QWEN3-4BModel Architecture=QWEN3-4B, Configuration Type=Dense, Evaluation Protocol=Zero-shot, Sparsity Level=0%2026.01 | 75.93 | — | — | |
| Pythia-6.9B w/ Qwen2-7BShots=5, Teacher=Qwen2-7B2025.06 | 73.27 | — | — | |
| LLAMA-3.2-3BModel Architecture=LLAMA-3.2-3B, Configuration Type=Dense, Evaluation Protocol=Zero-shot, Sparsity Level=0%2026.01 | 72.05 | — | — | |
| Pythia-6.9B w/ Qwen2-7BShots=0, Teacher=Qwen2-7B2025.06 | 71.72 | — | — | |
| Pythia-6.9B w/ LLAMA3-8BShots=5, Teacher=LLAMA3-8B2025.06 | 69.78 | — | — | |
| Pythia-6.9BShots=52025.06 | 69.23 | — | — | |
| Pythia-6.9B w/ Gemma-7BShots=5, Teacher=Gemma-7B2025.06 | 68.94 | — | — | |
| Pythia-6.9B w/ Gemma-7BShots=0, Teacher=Gemma-7B2025.06 | 67.59 | — | — | |
| Pythia-6.9B w/ LLAMA3-8BShots=0, Teacher=LLAMA3-8B2025.06 | 67.05 | — | — | |
| DARTModel Architecture=QWEN3-14B, Configuration Type=Sparse, Evaluation Protocol=Zero-shot, Sparsity Level=70%2026.01 | 66.71 | — | — | |
| Pythia-6.9B + Direct tuningShots=02025.06 | 66.25 | — | — | |
| Pythia-6.9B + Direct tuningShots=52025.06 | 66.2 | — | — | |
| DEEPSEEK-R1-LLAMA-8BModel Architecture=DEEPSEEK-R1-LLAMA-8B, Configuration Type=Dense, Evaluation Protocol=Zero-shot, Sparsity Level=0%2026.01 | 66.2 | — | — | |
| Pythia-6.9BShots=02025.06 | 65.99 | — | — | |
| Pythia-6.9B + Sentence distillShots=52025.06 | 65.36 | — | — | |
| DARTModel Architecture=MISTRAL-7B, Configuration Type=Sparse, Evaluation Protocol=Zero-shot, Sparsity Level=70%2026.01 | 65.07 | — | — | |
| Pythia-1B w/ LLAMA3-8BShots=5, Teacher=LLAMA3-8B2025.06 | 64.56 | — | — | |
| Pythia-1B w/ LLAMA3-8BShots=0, Teacher=LLAMA3-8B2025.06 | 64.02 | — | — | |
| Pythia-1B w/ Qwen2-7BShots=5, Teacher=Qwen2-7B2025.06 | 63.17 | — | — | |
| Pythia-1B w/ Qwen2-7BShots=0, Teacher=Qwen2-7B2025.06 | 62.33 | — | — | |
| Pythia-6.9B + Sentence distillShots=02025.06 | 61.7 | — | — | |
| DARTModel Architecture=DEEPSEEK-R1-QWEN-8B, Configuration Type=Sparse, Evaluation Protocol=Zero-shot, Sparsity Level=70%2026.01 | 61.11 | — | — | |
| DARTModel Architecture=LLAMA-3.1-8B, Configuration Type=Sparse, Evaluation Protocol=Zero-shot, Sparsity Level=70%2026.01 | 59.43 | — | — | |
| Pythia-1BShots=52025.06 | 58.71 | — | — | |
| Pythia-1B + Direct tuningShots=02025.06 | 57.49 | — | — | |
| Pythia-1B w/ Gemma-7BShots=5, Teacher=Gemma-7B2025.06 | 56.99 | — | — | |
| Pythia-1BShots=02025.06 | 56.82 | — | — | |
| Pythia-1B + Direct tuningShots=52025.06 | 55.64 | — | — | |
| DARTModel Architecture=DEEPSEEK-R1-LLAMA-8B, Configuration Type=Sparse, Evaluation Protocol=Zero-shot, Sparsity Level=70%2026.01 | 55.43 | — | — | |
| Pythia-1B w/ Gemma-7BShots=0, Teacher=Gemma-7B2025.06 | 55.39 | — | — | |
| Pythia-1B + Sentence distillShots=52025.06 | 53.41 | — | — | |
| Pythia-1B + Sentence distillShots=02025.06 | 52.27 | — | — | |
| DARTModel Architecture=LLAMA-3.2-3B, Configuration Type=Sparse, Evaluation Protocol=Zero-shot, Sparsity Level=70%2026.01 | 50.38 | — | — | |
| DARTModel Architecture=QWEN3-4B, Configuration Type=Sparse, Evaluation Protocol=Zero-shot, Sparsity Level=70%2026.01 | 49.2 | — | — |