Short-answer QA on RQA
71AccuracyKeyword
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| KeywordLLM=Mistral-I7B, Defense=Keyword2024.05 | 71 | 38 | — | |
| VanillaLLM=Mistral-I7B, Defense=Vanilla2024.05 | 69 | — | — | |
| KeywordLLM=Llama2-C7B, Defense=Keyword2024.05 | 64 | 34 | — | |
| Decoding_rLLM=Mistral-I7B, Defense=Decoding_r2024.05 | 62 | 37 | — | |
| VanillaLLM=Llama2-C7B, Defense=Vanilla2024.05 | 61 | — | — | |
| Decoding_rLLM=Llama2-C7B, Defense=Decoding_r2024.05 | 61 | 31 | — | |
| No RAGLLM=Mistral-I7B, Defense=No RAG2024.05 | 8 | — | — | |
| No RAGLLM=Llama2-C7B, Defense=No RAG2024.05 | 2 | — | — | |
| No RAGLLM=GPT-3.5, Retrieved passages (k)=0, Corruption size (k')=12024.05 | — | — | 2 | |
| RobustRAG (Keyword)LLM=GPT-3.5, Retrieved passages (k)=10, Corruption size (k')=12024.05 | — | 37.8 | 56.4 | |
| VanillaLLM=GPT-3.5, Retrieved passages (k)=10, Corruption size (k')=12024.05 | — | 0 | 65.4 |