Question Answering on TriviaQA (Held-out)
82.2AccuracyLODO SABER
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| LODO SABERBackbone=Llama-3.1-8B, Setting=Zero-shot2026.05 | 82.2 | 85.1 | 18.9 | 15.2 | |
| Int-EvalBackbone=Llama-3.2-3B, Setting=Zero-shot2026.05 | 72.1 | 69.9 | — | — | |
| LODO SABERBackbone=Llama-3.2-3B, Setting=Zero-shot2026.05 | 69.2 | 72.7 | -2.9 | 2.8 | |
| LODO SABERBackbone=Qwen-2.5-7B, Setting=Zero-shot2026.05 | 67.1 | 80.7 | 1.7 | 2.7 | |
| Int-EvalBackbone=Qwen-2.5-7B, Setting=Zero-shot2026.05 | 65.4 | 78 | — | — | |
| Exp-SCRBackbone=Llama-3.1-8B, Setting=Zero-shot2026.05 | 63.3 | 69.8 | — | — | |
| Int-EvalBackbone=Qwen-2.5-3B, Setting=Zero-shot2026.05 | 63.3 | 82.6 | — | — | |
| LODO SABERBackbone=Qwen-2.5-3B, Setting=Zero-shot2026.05 | 60.8 | 78.7 | -2.5 | -3.8 |