Common-sense reasoning on OBQA In-Distribution
88.43AccuracyPoLAR-VBLL (w/o LA)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PoLAR-VBLL (w/o LA)Backbone=Llama-3.1-8B, Fine-tuning steps=50002026.04 | 88.43 | 2.96 | 0.41 | |
| PoLAR-VBLLBackbone=Llama-3.1-8B, Fine-tuning steps=50002026.04 | 88.43 | 2.34 | 0.35 | |
| MLEBackbone=Llama-3.1-8B, Fine-tuning steps=50002026.04 | 88.3 | 8.33 | 0.49 | |
| LABackbone=Llama-3.1-8B, Fine-tuning steps=50002026.04 | 88.3 | 7.4 | 0.39 | |
| TFB-LLBackbone=Llama-3.1-8B, Fine-tuning steps=50002026.04 | 88.3 | 4.36 | 0.35 | |
| PoLAR-BLoBBackbone=Llama-3.1-8B, Fine-tuning steps=50002026.04 | 87.67 | 2.36 | 0.35 | |
| BLoBBackbone=Llama-3.1-8B, Fine-tuning steps=50002026.04 | 87.53 | 5.3 | 0.37 | |
| C-LoRABackbone=Llama-3.1-8B, Fine-tuning steps=50002026.04 | 86.83 | 5.33 | 0.58 | |
| TFBBackbone=Llama-3.1-8B, Fine-tuning steps=50002026.04 | 86.8 | 6.18 | 0.39 | |
| PoLAR-MLEBackbone=Llama-3.1-8B, Fine-tuning steps=50002026.04 | 86.67 | 8.56 | 0.48 | |
| PoLAR-LABackbone=Llama-3.1-8B, Fine-tuning steps=50002026.04 | 86.67 | 6.04 | 0.37 | |
| PoLAR-LA-LLBackbone=Llama-3.1-8B, Fine-tuning steps=50002026.04 | 86.67 | 6.21 | 0.39 |