Update dataset card after adding all raw datasets

README.md

@@ -6,47 +6,42 @@ task_categories:
 language:
 - zh
 - en
-size_categories:
-- 1K<n<10K
 ---
 
 # rageval Dataset Bundle
 
-This dataset bundle is prepared for the `rageval` undergraduate thesis experiments on comparing `VectorRAG` and `LightRAG` across document lengths.
+This dataset repository stores the raw datasets currently available in the local rageval workspace together with the processed datasets used in the thesis experiments.
 
-Target Hugging Face repo:
+Repo:
 
 - `bangwu/rageval_qa`
 
-## What Is Included
+## Included data
 
-### 1. Available raw source files
+### Raw datasets
 
-Only the raw files that are currently available in this workspace are included under `raw/financebench/`:
+- `raw/docfinqa/`
+- `raw/financebench/`
+- `raw/lecardv2/`
+- `raw/loong/`
+- `raw/narrativeqa/`
+- `raw/peerqa/`
 
-- official FinanceBench open-source metadata JSONL
-- locally cached FinanceBench PDFs
-- locally extracted FinanceBench text cache
+### Processed datasets
 
-Other original raw datasets are **not** present in this repository workspace and therefore are not exported as raw files here.
-
-### 2. Processed datasets
-
-The bundle includes:
-
-- full unified datasets used by the project
-- experiment splits used in the thesis
-- supplementary `UltraLongFinance` experiment files
-
-## Bundle Summary
-
-- processed files: 10
-- raw items exported: 3
-- total staged bytes: 1176251960
+- `processed/docfinqa/docfinqa_unified.jsonl`
+- `processed/lecardv2/lecardv2_unified.jsonl`
+- `processed/loong/loong_unified.jsonl`
+- `processed/narrativeqa/narrativeqa_unified.jsonl`
+- `processed/peerqa/peerqa_unified.jsonl`
+- `processed/experiment/main_experiment.jsonl`
+- `processed/experiment/stress_test.jsonl`
+- `processed/experiment/ultra_long_finance.jsonl`
+- `processed/experiment/split_manifest.json`
+- `processed/experiment/ultra_long_finance_manifest.json`
 
 ## Notes
 
-1. This bundle is intended for experiment verification and reproduction.
-2. Large processed files are staged exactly as used in the local experiments.
-3. This release intentionally keeps only data files. Scripts and experiment notes remain in the GitHub repository.
-4. Source materials may remain subject to their original licenses or terms of use. Review redistribution requirements before making the dataset public.
+1. This repository is data-only. Scripts, experiment code, and thesis documents remain in the GitHub repository.
+2. The processed files are the dataset artifacts used for the rageval comparison experiments.
+3. Raw source materials may remain subject to their original licenses or redistribution constraints. Please review the original dataset licenses before reuse.