Add submission checklist to README with links to environment, training scripts, runnable Colab, training logs, results, and story/writeup.

Report real rollout reward metrics.

Add shared Colab link to README.

Enhance README and demo app with improved model demo status reporting. Added a TL;DR section to README for quick understanding of SchemaQuake's functionality. Updated the run_gpu_model_demo function to return detailed model loading status and integrated a new function to assess booking status in the demo app.

Harden Model Demo loading fallback.

Refactor model loading and action handling in SchemaQuake demo and training notebook. Updated imports to include new Action and Observation types. Enhanced logic for selecting model checkpoints and executing actions based on environment state. Improved JSON extraction method for better robustness. Adjusted logging to differentiate between proposed and executed actions.

Enhance SchemaQuake documentation and training notebook. Updated Blog.md and README.md to clarify the RL environment aspect of SchemaQuake. Added supporting evidence for the training results in training-results.md. Improved the training notebook with clearer instructions and code for model loading and evaluation. Adjusted demo app to reflect changes in model loading and checkpoint handling.

Load trained model repo in demos.

Fix Model Demo startup on Space.

Add GPU adapter demo paths.

Update SchemaQuake training notebook for clarity in documentation. Adjusted the description to enhance readability and ensure reproducibility in the training path using Hugging Face TRL.

Update SchemaQuake training notebook and training output metrics.

Enhance evaluation metrics and documentation for imperfect heuristic agent.

Add inline numeric values to training results.

Consolidate final submission documentation.

Rename mini-blog to Blog.

Use final result artifacts and enrich submission story.

Use PyPI-available OpenEnv dependency for Space build.

Add final submission checklist and training results docs.

Prepare SchemaQuake final OpenEnv hackathon submission.

Upgrade SchemaQuake training quality pipeline for Qwen 3B.

Add hard drift evaluation benchmark.

Train with real SchemaQuake rollout rewards.

Add live training logs and curve visualizations in Gradio.

Enable true GRPO configuration on Space.

Fix GRPO training config and fallback behavior.

SchemaQuake HF deployment