TriviaQA dataset is a reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaqQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents, six per question on average, that provide high quality distant supervision for answering the questions.
| Paper | Year | Model | Model Details | NDCG@10 | Recall@5 | acc |
|---|---|---|---|---|---|---|
| INSTRUCTRAG: Instructing Retrieval-Augmented Generation via Self-Synthesized Rationales | 2024 | InstructRAG
| R:Contriever ,G:Llama3-Ins-8B(FT) | - | 73.5 | 78.5 |
| R:Contriever ,G:Llama3-Ins-8B(ICL) | - | 73.5 | 76.8 | |||
| NaiveRAG | R: Contriever, G: ChatGPT | - | 73.5 | 65.7 | ||
| NaiveRAG | R: Contriever, G: Llama3-Ins8B | - | 73.5 | 71.4 | ||
| SELF-RAG: Learning To Retrieve, Generate, and Critique Through SELF-Reflection | 2023 | Self-RAG
| R: Contriever ,G: Llama2-13B | - | - | 69.3 |
| R: Contriever ,G:Llama2-7B | - | - | 66.4 | |||
| Baseline1 | R: ❌, G: Llama2-7B | - | - | 30.5 | ||
| Baseline2 | R: ❌, G: Llama2-13B | - | - | 38.5 | ||
| ACTIVERAG: Autonomously Knowledge Assimilation and Accommodation through Retrieval-Augmented Agents (only sample 500 q for eval) | 2024 | ActiveRAG
| R:DPR ,G:ChatGPT-4oMINI | - | - | 83.4 |
| R:DPR ,G:Llama-3-Ins-70B | - | - | 85.4 | |||
| R:DPR ,G:Llama-3-Ins-8B | - | - | 79.8 | |||
| Baseline1 | R: ❌, G: Llama3-8-Ins8B | - | - | 67.2 | ||
| Baseline2 | R: ❌, G: Llama3-8-Ins70B | - | - | 80.4 |
