Split benchmark release into dedicated repo

README.md

@@ -7,26 +7,20 @@ configs:
 
 # ChartBench-E Benchmark 432
 
-This dataset release contains the frozen 432-sample benchmark build used by the paper:
+This repository contains the frozen 432-sample benchmark build used in the paper:
 
 > ChartBench-E: a benchmark for evaluating stable chart-to-values reading under increasing information load
 
 ## What Is Included
 
 - `hf_viewer/{train,validation,test}/`
-  - Hugging Face Data Studio friendly image dataset layout for the frozen 432-sample benchmark subset
-- `full_pool_metadata/`
-  - canonical manifests for the full 7,506-sample release
-- `full_pool_assets/{tables,programs,objects}/`
-  - linked structured artifacts for the full 7,506-sample release
-- `full_pool_archives/`
-  - six tar archives containing the full released image pools
+  - Hugging Face viewer-friendly split layout for the paper benchmark
 - `metadata/formal_all18_432.jsonl`
   - canonical manifest for the 432 released samples
 - `metadata/splits/{train,val,test}.jsonl`
   - split-specific manifests
 - `images/`
-  - full released image set referenced by `full_pool_metadata/all_pool_7506.jsonl`
+  - rendered chart images for the 432 benchmark samples
 - `tables/`
   - table-first ground-truth CSV files
 - `programs/`
@@ -36,7 +30,7 @@ This dataset release contains the frozen 432-sample benchmark build used by the
 - `evaluation/`
   - truth files used by the formal paper workflow
 - `docs/chartbench_e_rebuilt_README.md`
-  - codebase README
+  - codebase README snapshot
 - `docs/chartbench_e_paper.pdf`
   - paper PDF snapshot bundled with this release
 
@@ -54,33 +48,9 @@ This dataset release contains the frozen 432-sample benchmark build used by the
   - `val`: `50`
   - `test`: `93`
 
-## Configs
-
-- `benchmark_432`
-  - the frozen benchmark subset used in the paper
-
-## Full Pool Release
-
-The full released pool assembled from the main stratified pools and refill round 1 is also included in this repository:
-
-- samples: `7506`
-- splits:
-  - `train`: `5213`
-  - `validation`: `761`
-  - `test`: `1532`
-- full image directory:
-  - `images/` with `7506` rendered chart images referenced by `full_pool_metadata/all_pool_7506.jsonl`
-- image sample archives:
-  - `full_pool_archives/main__broad_standard_pool__images.tar`
-  - `full_pool_archives/main__broad_advanced_pool__images.tar`
-  - `full_pool_archives/main__broad_extreme_pool__images.tar`
-  - `full_pool_archives/refill_round1__broad_standard_pool__images.tar`
-  - `full_pool_archives/refill_round1__broad_advanced_pool__images.tar`
-  - `full_pool_archives/refill_round1__broad_extreme_pool__images.tar`
-
 ## Data Fields
 
-The public Hugging Face viewer layout in `hf_viewer/{train,validation,test}/metadata.jsonl` keeps a slimmer schema:
+The public viewer layout in `hf_viewer/{train,validation,test}/metadata.jsonl` keeps a slimmer schema:
 
 - `sample_id`
 - `chart_id`
@@ -106,26 +76,23 @@ The public Hugging Face viewer layout in `hf_viewer/{train,validation,test}/meta
 - `program_path`
 - `object_path`
 
-The canonical audit manifest remains available in `metadata/formal_all18_432.jsonl`, including the original local-path provenance fields.
-
-The full-pool release manifest in `full_pool_metadata/all_pool_7506.jsonl` keeps unique sample identifiers across pool batches, plus linked paths into `full_pool_assets/` and the corresponding image archives in `full_pool_archives/`.
+The canonical audit manifest remains available in `metadata/formal_all18_432.jsonl`.
 
 ## Notes
 
-- This release is built from the formal `all18` stratified subset used in the manuscript.
-- Code, prompt templates, evaluation scripts, and generation utilities live in the project repository:
+- This is the benchmark release cited by the manuscript.
+- Code, prompt templates, evaluation scripts, and generation utilities live in:
   - `https://github.com/luojueling/ChartBench-E`
-  - `https://huggingface.co/datasets/xiaoluo11/ChartBench-E`
-- This dataset release is intended to make the paper auditable and reviewer-accessible.
+- The broader 7,506-sample released pool is published separately to keep this benchmark repo stable and viewer-friendly.
 
 ## Suggested Citation
 
 ```bibtex
-@misc{chartbenche2026,
-  title        = {ChartBench-E: a benchmark for evaluating stable chart-to-values reading under increasing information load},
+@misc{chartbenche2026benchmark,
+  title        = {ChartBench-E Benchmark 432},
   author       = {Hui Long and Jueling Luo and Si Xie and Haoli Yu and Juncheng Ye and Xutao Deng},
   year         = {2026},
-  note         = {Dataset release for the frozen 432-sample benchmark build}
+  note         = {Frozen 432-sample benchmark release used in the manuscript}
 }
 ```