Spaces:
Running
Running
|
Download README.md from landtml/rag-verifier: direct link, hf CLI and curl.
- Browser
- Download file 960 Bytes
-
https://huggingface.co/spaces/landtml/rag-verifier/resolve/main/README.md
- Command line
-
hf download hf://spaces/landtml/rag-verifier/README.md
-
curl -L -o README.md https://huggingface.co/spaces/landtml/rag-verifier/resolve/main/README.md
960 Bytes
| title: RAG Verifier | |
| emoji: 🔎 | |
| colorFrom: blue | |
| colorTo: gray | |
| sdk: static | |
| app_file: index.html | |
| pinned: false | |
| license: other | |
| license_name: lfm1.0 | |
| license_link: https://huggingface.co/landtml/d1-2B-RAG-Verifier/blob/main/LICENSE | |
| short_description: Checks whether a RAG answer is backed by its source | |
| models: | |
| - landtml/d1-2B-RAG-Verifier | |
| datasets: | |
| - wandb/RAGTruth-processed | |
| tags: | |
| - rag | |
| - hallucination-detection | |
| - groundedness | |
| # RAG Verifier demo | |
| Paste a source, a question and an answer. [d1-2B RAG Verifier](https://huggingface.co/landtml/d1-2B-RAG-Verifier) returns the probability that every claim in the answer is supported by the source, and the page routes the answer: pass at 0.88 and above, block at 0.12 and below, review in between. | |
| The examples are written for this demo and scored by the model; the benchmark results on the page come from the RAGTruth test set (MIT licence). Submitted text is sent to the model and not logged or stored. | |