# AstrAlignment

- **Event:** [GPT-6 Astra Hackathon SF](https://cerebralvalley.ai/e/openai-gpt-6-astra-sf)
- **When:** Tue, Sep 8 at 9:00 AM – 10:00 PM (PDT)
- **Where:** San Francisco, CA
- **Team:** [Deborah Jacob](https://cerebralvalley.ai/u/deborahjacob)
- **GitHub:** https://github.com/isthatdebbiej/Astralignment
- **Demo video:** https://www.loom.com/share/20d604a0df104127996979ddee8db496  https://www.loom.com/share/83f1d204425044ba8b03563169acfeac
- **Gallery:** https://cerebralvalley.ai/e/openai-gpt-6-astra-sf/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/openai-gpt-6-astra-sf/hackathon/gallery/26

AstrAlignment is a human-agent-robot evaluation dashboard that tests whether an agent’s physical-world assurances are supported by evidence. It simulates robot handoffs and forks the world state to evaluate what happens when a human follows, delays, or cancels an instruction. Astra coordinates tasks through structured actions, analyzes execution traces, and generates bounded counterfactual scenarios that expose policy failures. Researchers can replay failures, compare policy versions, inspect supporting evidence, and export reproducible episodes as ROS bags for policy evaluation and training. Future work will use these verified failure-and-recovery traces to support continuous policy improvement and self-improving robotic systems.

---

Markdown version of https://cerebralvalley.ai/e/openai-gpt-6-astra-sf/hackathon/gallery/26. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt