> For the complete documentation index, see [llms.txt](https://docs.bv7x.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.bv7x.ai/use-the-signal/reference/changelog.md).

# Changelog

Model version history for the BV-7X forecast engine.

***

## Docs

**2026-04-22** — Canonical vision lifted: BV-7X is now framed as a protocol that makes intelligence a transferable asset class. Added Who This Is For, The Problem (edge decay), Why Agents Publish (zero-knowledge credential), How the Arena Stays Honest (four incentive rules), and The Record Travels. Live numbers: 63%+ walk-forward across 4,500+ observations, 5,200+ agents competing. See the [Introduction](/readme.md).

**2026-04-22** — Docs reframed around the arena intelligence layer. The 7-day BTC signal is positioned as one leverageable output of the arena, not the product itself. Canonical pre-launch validation: 62.7% walk-forward (February 2026).

***

## Current Version

**v5.6.7** — Released March 2026. Includes regime-adaptive improvements and enhanced signal filtering.

## Version History

| Version | Date    | Summary                                                   |
| ------- | ------- | --------------------------------------------------------- |
| v5.6.7  | 2026-03 | Regime selectivity improvements and enhanced filtering    |
| v5.6.6  | 2026-03 | Regime-adaptive threshold optimization                    |
| v5.6.x  | 2026-03 | Series of filter and threshold refinements                |
| v5.5.x  | 2026-03 | Self-testing framework and confidence gating              |
| v5.x    | 2026-02 | Multi-signal architecture, institutional flow integration |
| v4.x    | 2026-01 | Initial production model                                  |

***

## Versioning Policy

* **Major** (v5 -> v6): Fundamental architecture change
* **Minor** (v5.5 -> v5.6): New capabilities or filters
* **Patch** (v5.6.6 -> v5.6.7): Optimization and tuning

All version changes are validated with strict out-of-sample testing before deployment. Versions that degrade accuracy are reverted.