Update README.md

README.md

@@ -1,10 +1,38 @@
 ---
 license: cc-by-4.0
----## Structure
+pretty_name: NeWGen-Bench
+size_categories:
+  - 100K<n<1M
+task_categories:
+  - other
+tags:
+  - neural-network-weights
+  - weight-generation
+  - parameter-generation
+  - benchmark
+  - model-zoo
+configs:
+  - config_name: default
+    data_files:
+      - split: train
+        path: metadata.parquet
+---
+
+# NeWGen-Bench
+
+A benchmark for neural network parameter generation. The release contains
+trained checkpoints across multiple architectures, datasets, and task
+domains, each paired with structured metadata describing the architecture,
+training data, hyperparameters, and final performance.
+
+This dataset accompanies an anonymous submission to NeurIPS 2026
+Evaluations and Datasets Track.
+
+## Repository Structure
 
 ```text
-benchmark_v2/
-  datasets/   Local datasets, extracted archives, and Hugging Face caches
+<root>/
+  datasets/   Local datasets and dataset caches
   registry/   JSON registries for benchmark runs, metrics, configs, and checkpoints
   train/      Training scripts and configuration files
   models/     Model definition files
@@ -12,26 +40,50 @@ benchmark_v2/
   collect/    Collected checkpoint trajectories
 ```
 
-## Tasks
+## Task Domains
+
+The benchmark spans the following task domains:
 
 - Image classification
-- Text classification
 - Image segmentation
-- M/LLM Adaptation
-- Reinforcement learning
-
-The main registry files are:
-
-```text
-registry/image_classification.json
-registry/text_classification.json
-registry/image_segmentation.json
-registry/reinforcement_learning.json
-```
+- Text classification
+- LLM adaptation
+- VLM adaptation
 
 ## Artifacts
 
 - Trained checkpoints: `zoo/**/*.pth`
-- Collected checkpoints: `collect/**/*.pth`
+- Collected checkpoint trajectories: `collect/**/*.pth`
 - Training configs: `train/**/config/*.json`
-- Model code: `models/*.py`
+- Model definitions: `models/*.py`
+- Per-checkpoint metadata: `metadata.parquet`
+
+## Usage
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("Anonymous1Researcher/NeWGen-Bench")
+print(ds["train"][0])
+```
+
+Or load directly with pandas:
+
+```python
+import pandas as pd
+
+df = pd.read_parquet(
+    "hf://datasets/Anonymous1Researcher/NeWGen-Bench/metadata.parquet"
+)
+```
+
+## Status
+
+This release contains the metadata table. The full collection of trained
+checkpoints will be made available before the camera-ready deadline.
+
+## License
+
+Released under CC BY 4.0. Individual checkpoints are derived from public
+training datasets, each governed by its own license; downstream users
+should comply with the licenses of the underlying training data.