Update README.md

README.md

@@ -29,21 +29,21 @@ source_datasets:
 
 AV-QuantBench is a procedural audio-visual benchmark for evaluating multimodal foundation models on abstract temporal reasoning, cross-modal conflict detection, and synchronized data interpretation across finance, medical, and industrial domains.
 
-This Hugging Face dataset repository is structured to support a benchmark-style release. It contains:
+This Hugging Face dataset repository is structured as a benchmark-style release. It contains:
 
 - split metadata in JSONL format,
 - question-answer annotations,
-- sample video/audio assets,
+- audio-visual sample assets,
 - manifest files by domain,
 - and documentation for schema and responsible use.
 
-The sample subset included here is upload-ready and demonstrates the expected repository structure. It can be extended to the full release by replacing the sample media and JSONL files with the complete benchmark assets generated by the AV-QuantBench pipeline.
+The repository is organized so that newly generated benchmark outputs can be added with minimal restructuring. In particular, the `samples/` directory follows the same domain-first layout as the AV-QuantBench generator outputs.
 
 ## Dataset Summary
 
 AV-QuantBench converts time-series signals into synchronized visual topology and acoustic momentum. Each sample is paired with machine-generated QA derived from deterministic state-machine triggers. The benchmark is designed to evaluate whether a model can jointly reason over audio and video when the two modalities either align or intentionally diverge.
 
-### Supported tasks
+### Supported Tasks
 
 - Cross-modal conflict detection
 - Audio-visual temporal reasoning
@@ -54,8 +54,7 @@ AV-QuantBench converts time-series signals into synchronized visual topology and
 
 - Video (`.mp4`)
 - Audio (`.wav`)
-- Frame previews (`.png`)
-- Structured annotations (`.jsonl`)
+- Structured annotations (`.json`, `.jsonl`)
 
 ### Domains
 
@@ -101,101 +100,4 @@ AV-QuantBench-HF-Dataset/
 │       └── metadata/
 └── docs/
     ├── schema.md
-    └── responsible_use.md
-```
-
-## Data Fields
-
-Each record in `metadata/*.jsonl` contains:
-
-- `id`
-- `split`
-- `domain`
-- `video`
-- `audio`
-- `frame_preview`
-- `duration_sec`
-- `fps`
-- `sample_rate`
-- `task_type`
-- `question`
-- `ground_truth`
-- `has_conflict`
-- `source_mode`
-- `seed`
-- `algorithmic_metadata`
-
-See `docs/schema.md` for details.
-
-## Splits
-
-This upload-ready package includes one example for each split:
-
-- `train`: finance
-- `val`: medical
-- `test`: iiot
-
-In the full release, these files should be replaced by the complete split manifests and associated media.
-
-## Example Usage
-
-```python
-import json
-from pathlib import Path
-
-root = Path('.')
-with open(root / 'metadata' / 'train.jsonl', 'r', encoding='utf-8') as f:
-    sample = json.loads(f.readline())
-
-print(sample['domain'])
-print(sample['question'])
-print(sample['video'])
-```
-
-## Generation Pipeline
-
-The full AV-QuantBench generator creates each sample using the following stages:
-
-1. substrate synthesis or ingestion,
-2. anti-leakage visual rendering,
-3. acoustic mapping and sonification,
-4. deterministic audio-video synchronization,
-5. state-machine-based QA generation.
-
-## Licensing
-
-Unless otherwise stated, the dataset artifacts in this repository are released under **CC BY-NC 4.0**. This includes metadata, annotations, documentation, and generated media distributed here.
-
-## Limitations
-
-- The package currently includes a minimal sample subset for upload readiness.
-- Full benchmark scale requires exporting the complete media and annotation assets from the generator repository.
-- Synthetic audio-visual data should not be interpreted as real financial or medical advice.
-
-## Responsible Use
-
-Please see `docs/responsible_use.md`.
-
-## Citation
-
-If you use AV-QuantBench, please cite the accompanying paper:
-
-```bibtex
-@article{gu2026avquantbench,
-  title={AV-QuantBench: An Audio-Visual Data Video Benchmark for Foundation Models in Complex Temporal Reasoning},
-  author={Gu, Fengchen and Ren, Xiaotian and Jiang, Zhengyong and Garc{'i}a-Fern{'a}ndez, {'A}ngel F. and Su, Jionglong and Li, Huakang},
-  journal={arXiv preprint arXiv:XXXX.XXXXX},
-  year={2026}
-}
-```
-
-
-## Sample Layout
-
-The `samples/` directory mirrors the domain-first output layout produced by the AV-QuantBench generator, so newly generated data can be copied into the repository with minimal restructuring:
-
-- `samples/finance/videos`, `samples/finance/audio`, `samples/finance/qa`, `samples/finance/metadata`
-- `samples/medical/videos`, `samples/medical/audio`, `samples/medical/qa`, `samples/medical/metadata`
-- `samples/iiot/videos`, `samples/iiot/audio`, `samples/iiot/qa`, `samples/iiot/metadata`
-
-This layout is intentionally aligned with the local `all_outputs`-style export structure for easier dataset refresh and replacement.
+    └── responsible_use.md