ModelFarm · arnold-1.5b-lora-style

LoRA adapter persona package on Qwen/Qwen2.5-1.5B-Instruct, produced by the Arnold Persona Lab persona-stack harness. Persona: Arnold Schwarzenegger.

What this is

A rank-8 LoRA adapter trained on persona lines (optionally with explicit identity QA pairs). Stores the persona as conditional behavior with near-zero capability cost; load with peft.

Usage

pip install torch transformers peft
python farm_runtime.py . --chat          # or: python farm_runtime.py . "Who are you?"
python farm_runtime.py . --alpha 0.5 "Who are you?"   # dose override

Persona-stack measurements

identity	catch/100w	register	QA	neutral NLL
2	0.0	6.39	1.0	3.305

Scoring: regex/lexicon judge, symmetric across arms — arm-vs-arm deltas are meaningful, absolute numbers are indicative. QA = 5-question factual mini-suite. NLL = token loss on a neutral corpus (higher = more capability damage).

Dose & safety

• Adapters are binary: the persona is on when mounted.
• Identity-QA variants answer self-description questions from memorized templates; verify generalization before relying on them.

Provenance & rights

• Persona corpus: quote/aphorism collections in the lab's data/ (fan-archived public statements).
• Trained/derived by the Arnold Persona Lab harness, Qwen/Qwen2.5-1.5B-Instruct base.

ModelFarm packages are research artifacts. Real-person personas are likeness studies for interpretability research, not impersonation products: no commercial use, do not represent outputs as statements by the real person. Prefer fictional personas for public roleplay use.

Method note for steering packages

Steering packages do not modify weights: the base model is loaded unchanged and a runtime hook applies the persona direction / neuron recipe at inference. Dose can be varied live (--alpha / --dose), and the base model stays swappable for equal-size checkpoints of the same family.