[FEAT] Neural Scorer: Confidence-Netz mit RL-Training #3

Closed
opened 2026-05-25 00:32:05 +02:00 by Otto · 0 comments
Owner

Trainierbares Confidence-Netzwerk:

  • Architektur: 2-Layer Feed-Forward (Input: 384+5 Meta | Hidden: 64 | Output: 1)
  • Trainingsdaten: User-Feedback (=1.0, =0.0, korrigiert=0.5)
  • Speicherung: data/scorer_model.pt
  • Training asynchron nach jedem 10. Feedback

Labels: feature, ml

Trainierbares Confidence-Netzwerk: - Architektur: 2-Layer Feed-Forward (Input: 384+5 Meta | Hidden: 64 | Output: 1) - Trainingsdaten: User-Feedback (✅=1.0, ❌=0.0, korrigiert=0.5) - Speicherung: `data/scorer_model.pt` - Training asynchron nach jedem 10. Feedback Labels: feature, ml
Otto closed this issue 2026-05-25 09:46:50 +02:00
Sign in to join this conversation.
No Label
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Otto/second-brain#3