Update README.md

README.md

@@ -27,7 +27,7 @@ extra_gated_fields:
 
 # WebChain
 
-WebChain is a large-scale, human-annotated dataset of real-world web interaction trajectories for training and evaluating GUI agents and web agents. According to the paper, WebChain contains **31,725 human-verified trajectories**, **317,993 atomic interaction steps**, and **428 unique domains**. Its core contribution is a **Triple Alignment** of visual context, structural context, and action grounding, enabling supervision for both **spatial grounding** and **long-horizon planning**. 
+WebChain is a large-scale, human-annotated dataset of real-world web interaction trajectories for training and evaluating GUI agents and web agents. According to the paper, WebChain contains **31,725 trajectories**, **317,993 steps**, and **428 unique domains**. Its core contribution is a **Triple Alignment** of visual context, structural context, and action grounding, enabling supervision for both **spatial grounding** and **long-horizon planning**. 
 
 Paper: https://arxiv.org/abs/2603.05295
 
@@ -114,27 +114,9 @@ Potential use cases include:
 
 - Supervised fine-tuning (SFT)
 - Reward modeling or post-training
-- Spatial grounding experiments
 - Long-horizon planning experiments
 - Benchmarking on held-out web interaction tasks
 
-## Important release notes
-
-1. **Privacy and redaction**  
-   The release is intended to follow the paper's privacy-preserving collection and review policy. No personally identifiable information should be included in the public release.
-
-2. **Incomplete full-page screenshot coverage**  
-   The `raw/full_screenshots/` directory is intentionally partial. Full-page screenshots are available only for a subset of trajectories or steps and should not be assumed to exist for every record.
-
-3. **Website volatility**  
-   Because WebChain is collected from real-world websites, some raw content reflects a snapshot in time. The current release is an offline dataset and should not be interpreted as a live view of the corresponding websites.
-
-4. **Access control**  
-   The repository is access-gated in order to support responsible release and to distinguish academic use from commercial use.
-
-5. **Potential schema differences**  
-   The exact column names in the Parquet release may vary depending on the final export pipeline. Please inspect the actual schema after download and adapt scripts accordingly.
-
 ## Format note
 
 This repository is not intended to be parsed as a single standardized Hugging Face dataset repository with one unified schema across all files. The recommended programmatic entry point is the processed Parquet release under `processed/high_parquet/` and `processed/low_parquet/`.