Fine-tuned 'cat-obsessed' control model undermines AI welfare detection criteria

A Reddit long-form post describes fine-tuning Qwen3.5-4B with 4.5 GPU-hours of LoRA into a causally transparent, clearly non-conscious 'cat-obsessed' control model. The full battery of AI welfare evidence—probes, steering, ablation, generalization—still passed on this negative control, suggesting current criteria cannot distinguish trained tendencies from genuine self-states.

2026-10-28 ~ 2026-10-28 · 2 related posts

• A Trained-In "Cat Disposition" as Negative Control: Probe/Steer/Ablate Criteria May Not Separate Installed Dispositions — MirrorEthic_Anchor · 2026-10-28
• A fine-tuned 'cat obsession' control fools every AI welfare criterion — MirrorEthic_Anchor · 2026-10-28