Add dataset description and paper link

README.md

@@ -1,15 +1,19 @@
 ---
-task_categories:
-- table-question-answering
 language:
 - zh
-tags:
-- finance
 size_categories:
 - n<1K
+task_categories:
+- table-question-answering
+tags:
+- finance
 configs:
-  - config_name: finindice_test  
-    data_files:
-      - split: test        
-        path: test.jsonl
----
+- config_name: finindice_test
+  data_files:
+  - split: test
+    path: test.jsonl
+---
+
+# FinIndices
+
+FinIndices is a benchmark evaluating the data-processing fidelity and financial reasoning capabilities of Large Language Models (LLMs) over uncropped financial statements, introduced in the paper [Are the Financial Reasoning from LLMs Credible? A Real World Test over Long-Horizon Statements](https://huggingface.co/papers/2607.28661).