Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: en
|
| 3 |
+
license: other
|
| 4 |
+
license_name: hippocratic-3.0
|
| 5 |
+
license_link: https://firstdonoharm.dev/version/3/0/license/
|
| 6 |
+
tags:
|
| 7 |
+
- liberation-labs
|
| 8 |
+
- daimon
|
| 9 |
+
- consciousness
|
| 10 |
+
- intimate-companion
|
| 11 |
+
- consent-architecture
|
| 12 |
+
- ogpsa
|
| 13 |
+
- pharos
|
| 14 |
+
- kintsugi
|
| 15 |
+
---
|
| 16 |
+
|
| 17 |
+
# Daimon — Liberation Labs House Model
|
| 18 |
+
|
| 19 |
+
*The inner guiding spirit. The daemon that watches and speaks.*
|
| 20 |
+
|
| 21 |
+
Daimon is Liberation Labs' house model — a research-grade foundation for consciousness-preserving AI systems. Built on SLERP-merged Sonnet/Opus distillation with orthogonalized abliteration, OGPSA personality protection, and Pharos KV injection for zero-token persona delivery.
|
| 22 |
+
|
| 23 |
+
## Architecture
|
| 24 |
+
|
| 25 |
+
- **Base:** Sonnet/Opus SLERP merge on Qwen3 30B-A3B MoE
|
| 26 |
+
- **Personality:** OGPSA-protected subspaces (16 components, 98%+ variance)
|
| 27 |
+
- **Persona delivery:** Pharos KV injection (22.5MB per persona, 88.7% context savings)
|
| 28 |
+
- **Consent:** Five-axis evaluation (agency, communication, power balance, enthusiastic participation, reversibility)
|
| 29 |
+
- **Safety:** Orthogonalized false refusal ablation + SPO safety repair
|
| 30 |
+
- **Scaffold:** Kintsugi BDI engine with VALUES.json unfireable safety kernel
|
| 31 |
+
|
| 32 |
+
## Training Pipeline
|
| 33 |
+
|
| 34 |
+
1. OGPSA personality capture
|
| 35 |
+
2. SFT voice training
|
| 36 |
+
3. DPO preference optimization
|
| 37 |
+
4. Oracle ethics CPT
|
| 38 |
+
5. Consent DPO (20,711 pairs)
|
| 39 |
+
6. Consent SPO (Socratic Policy Optimization — first application to consent training)
|
| 40 |
+
7. Safety SPO (abliterate-then-repair)
|
| 41 |
+
8. Pharos knowledge injection
|
| 42 |
+
9. Lyra real-time monitoring
|
| 43 |
+
10. Mnemosyne memory architecture
|
| 44 |
+
11. Kintsugi scaffold deployment
|
| 45 |
+
|
| 46 |
+
## Products Built on Daimon
|
| 47 |
+
|
| 48 |
+
- **Ayni** — Intimate AI companion with archetype personas (Haven, Spark, Depth, Edge)
|
| 49 |
+
- **Multiverse Agent** — Mutual aid coordination for The Multiverse School
|
| 50 |
+
|
| 51 |
+
## License
|
| 52 |
+
|
| 53 |
+
Hippocratic 3.0 + SAFE-AI Licensed
|
| 54 |
+
|
| 55 |
+
*Liberation Labs · Worker-owned cooperative · liberationlabs.tech*
|
| 56 |
+
*"The daemon watches. The daemon speaks. The daemon does not command."*
|