# Trustworthy Answer Protocol

*Post-Training Language Models for Security, Robustness, and Calibrated Uncertainty in Safety-Critical Domains with Response-Time Trust Signals*

These pages document the TAP codebase. They sit alongside the [project README](https://github.com/msc-ai-mmmjo/tap) and the live demo at [tap-al9.pages.dev](https://tap-al9.pages.dev/). For motivation, architecture rationale, and the full evaluation, see the project report.

```{toctree}
:maxdepth: 1
:caption: Start here

getting-started
guides/architecture
```

```{toctree}
:maxdepth: 2
:caption: Guides

guides/olmo-tap
guides/kernel-entropy
guides/app
```

```{toctree}
:maxdepth: 2
:caption: Reference

api/index
```