Update README.md

README.md

@@ -1,3 +1,20 @@
 ---
 license: mit
+task_categories:
+- question-answering
+language:
+- en
 ---
+Credibility-aware Generation Benchmark (CAGB) is a benchmark constructed to evaluate the credibility-aware generation ability of models, dealing with flawed information in the context.
+This benchmark encompasses the following three specific scenarios where the integration of credibility is essential:
+- Open-domain QA
+  - 2WikiMultiHopQA
+  - HotpotQA
+  - Musique
+  - RGB
+- Time-sensitive QA
+  - EvoTempQA
+- Misinformation Polluted QA
+  - NewsPollutedQA
+
+Read our [paper](https://arxiv.org/abs/2404.06809) for more insights on credibility-aware generation.