Single-hop Question Answering on MFQA en 16k
23.76Overall ScoreDCS
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DCSBackbone=Mistral-7B-Instruct2025.06 | 23.76 | — | |
| Llama-3-8B-Instruct with DCSBackbone=Llama-3-8B-Instruct, Context Strategy=DCS2025.06 | 23.7 | 35.5 | |
| DCSBackbone=Llama-3-8B-Instruct2025.06 | 23.7 | — | |
| DCSBackbone=Vicuna-7B2025.06 | 19.87 | — | |
| Llama-3-8B-InstructBackbone=Llama-3-8B-Instruct, Context Strategy=None2025.06 | 18.22 | 27.6 | |
| Llama-3-8B-InstructBackbone=Llama-3-8B-Instruct, Context Management=Baseline2025.06 | 18.22 | — | |
| Mistral-7B-InstructBackbone=Mistral-7B-Instruct, Context Management=Baseline2025.06 | 16.62 | — | |
| MOICEBackbone=Mistral-7B-Instruct2025.06 | 16.62 | — | |
| InfLLMBackbone=Mistral-7B-Instruct2025.06 | 16.27 | — | |
| Llama-3-8B-Instruct with InfilmBackbone=Llama-3-8B-Instruct, Context Strategy=Infilm2025.06 | 15.3 | 27.15 | |
| InfLLMBackbone=Llama-3-8B-Instruct2025.06 | 15.3 | — | |
| MOICEBackbone=Vicuna-7B2025.06 | 14.23 | — | |
| StreamingBackbone=Mistral-7B-Instruct2025.06 | 14.18 | — | |
| LM-infiniteBackbone=Mistral-7B-Instruct2025.06 | 14.1 | — | |
| LM-InfiniteBackbone=Vicuna-7B2025.06 | 13.84 | — | |
| StreamingBackbone=Vicuna-7B2025.06 | 13.82 | — | |
| Vicuna-7BBackbone=Vicuna-7B, Context Management=Baseline2025.06 | 13.79 | — | |
| Llama-3-8B-Instruct with LM-infiniteBackbone=Llama-3-8B-Instruct, Context Strategy=LM-infinite2025.06 | 13.45 | 24.2 | |
| LM-infiniteBackbone=Llama-3-8B-Instruct2025.06 | 13.45 | — | |
| Llama-3-8B-Instruct with StreamingBackbone=Llama-3-8B-Instruct, Context Strategy=Streaming2025.06 | 12.84 | 24.26 | |
| StreamingBackbone=Llama-3-8B-Instruct2025.06 | 12.84 | — | |
| InfLLMBackbone=Vicuna-7B2025.06 | 5.92 | — |