YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Tenacious Judge LoRA
Path B SimPO LoRA adapter for Tenacious-Bench sales-output preference scoring.
- Backbone:
unsloth/Qwen2.5-0.5B-Instruct - Backbone note: Operational text-only fallback; Qwen3.5-0.8B is currently multimodal and breaks TRL CPO tokenization on text-only preference pairs.
- Objective: SimPO via TRL
CPOTrainer - Seed:
3407 - Training pairs:
81 - Eval pairs:
10 - Intended use: rejection-sampling or critique layer for Tenacious-style B2B sales outreach drafts
- Known limitation: dual-control / premature-booking examples are underrepresented in the current preference set
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support