|
--- |
|
license: mit |
|
datasets: |
|
- ruotong-pan/CAGB |
|
language: |
|
- en |
|
pipeline_tag: question-answering |
|
--- |
|
This model is a 13B model with credibility-aware generation, fine-tuned on LLaMA2-13B. |
|
It possesses the capability to comprehend and utilize credibility within the context for generation, thereby exhibiting strong robustness聽when dealing with externally introduced noisy documents in聽the RAG. |
|
Because of the capability of customized credibility in CAG, it has diverse application scenarios, including personalized response generation and the resolution of knowledge conflicts. |
|
|
|
Read our [article](https://arxiv.org/abs/2404.06809) to learn more. |
|
|
|
Evaluate our model in real scenarios including Open-domain QA, Time-sensitive QA, and Misinformation Polluted QA. For more information and to access the evaluation code, visit [Code](https://github.com/panruotong/CAG). |
