Closed-book Question Answering on TriviaQA TQ (test)
76.12EMDCO
Evaluation results
| Method | Links | |
|---|---|---|
| DCOBackbone=Llama-3-70B2026.06 | 76.12 | |
| DeCoRe (static)Backbone=Llama-3-70B2026.06 | 74.79 | |
| GreedyBackbone=Llama-3-70B2026.06 | 74.77 | |
| DoLa (low)Backbone=Llama-3-70B2026.06 | 74.74 | |
| ADBackbone=Llama-3-70B2026.06 | 74.14 | |
| DoLa (high)Backbone=Llama-3-70B2026.06 | 73.96 | |
| ITIBackbone=Llama-3-70B2026.06 | 73.54 | |
| CDBackbone=Llama-3-70B2026.06 | 72.83 | |
| DCOBackbone=Llama-3-8B2026.06 | 57.33 | |
| DeCoRe (static)Backbone=Llama-3-8B2026.06 | 56.93 | |
| DoLa (low)Backbone=Llama-3-8B2026.06 | 56.63 | |
| GreedyBackbone=Llama-3-8B2026.06 | 56.58 | |
| DoLa (high)Backbone=Llama-3-8B2026.06 | 56.5 | |
| ADBackbone=Llama-3-8B2026.06 | 54.93 | |
| ITIBackbone=Llama-3-8B2026.06 | 48.41 | |
| BART-LargeQA-finetune=true2021.06 | 24.9 | |
| INTRADocL=2K, Number of demonstrations=122024.02 | 16.04 | |
| BM25ChunkL=2K, Number of demonstrations=122024.02 | 15.57 | |
| UNIChunkL=2K, Number of demonstrations=122024.02 | 15.53 | |
| INTRADocL=8K, Number of demonstrations=482024.02 | 15.09 | |
| MixChunkL=2K, Number of demonstrations=122024.02 | 14.47 | |
| BM25ChunkL=8K, Number of demonstrations=482024.02 | 11.09 | |
| MixChunkL=8K, Number of demonstrations=482024.02 | 10.9 | |
| UNIChunkL=8K, Number of demonstrations=482024.02 | 10.59 |