Commonsense Question Answering on CommonSenseQA n=1700 (20%)
68.12AccuracyMTO-MCP
Evaluation results
| Method | Links | |
|---|---|---|
| MTO-MCPEvaluation Protocol=Fine-Tuning, Backbone=T5-large2026.06 | 68.12 | |
| MVP-Tuning2026.06 | 67.12 | |
| MTO-APEvaluation Protocol=Fine-Tuning, Backbone=T5-large, Adaptation Objective=Denoising2026.06 | 66.25 | |
| GSC2026.06 | 65.83 | |
| SAFE2026.06 | 65.16 | |
| T5-large (baseline)Evaluation Protocol=Fine-Tuning, Backbone=T5-large2026.06 | 65.15 | |
| GreaseLM2026.06 | 63.09 | |
| MTO-PreAPEvaluation Protocol=Prompt-Tuning, Backbone=T5-large2026.06 | 59.57 | |
| MTO-PreMAPEvaluation Protocol=Prompt-Tuning, Backbone=T5-large, Adaptation Objective=LM2026.06 | 59.21 | |
| MTO-MPEvaluation Protocol=Fine-Tuning, Backbone=T5-large2026.06 | 59.02 | |
| RoBERTa-largeBackbone=RoBERTa-large2026.06 | 58.47 | |
| MHGRN2026.06 | 50.23 | |
| QA-GNN2026.06 | 50.15 |