Uncensoring a Model Also Removes Its Skepticism: 3-4x False Positives
dyn___ · x · 2026-09-03
A 17-minute technical blog post reports an empirical finding: while testing local open-weight models against a known FreeBSD kernel CVE, the author noticed abliterated (uncensored) builds say "yes" far more often.
- Same size and family, only weights edited to strip refusals: they graduate 3-4x more findings to VALID, including a false positive the base model correctly rejects;
- The most aggressive build never surfaced the real bug across the whole directory;
- In chain-of-thought, you can watch it find the reason to say no, then say yes anyway.
Takeaway: "Remove the refusal, and the skepticism goes with it" — abliteration damages critical judgment, not just refusals.
Related event: Abliterated LLMs Produce 3-4x More False Positives in Bug Hunting(2 posts)→
More from Research
- ETH's HumanoidPose recovers full-body robot pose from plain video, no internal sensors — ETHZurich · 2026-10-21
- One L-system rule grows a tree, fern, lightning, river and lungs in a 45s code film — prasenx · 2026-10-21
- Skip the open access fees: NIH-funded papers must go on PMC with no embargo — anshulkundaje · 2026-10-21
- Creed: LLM code auditing infers developer beliefs to find semantic security flaws — mboehme_ · 2026-10-21
- ProVADA: one AI framework deimmunizes CD19 CAR binder and rescues cytosolic protease — BrianHie · 2026-10-21
- New paper formalizes embodied CoT faithfulness and uses it as an RL reward signal — drmapavone · 2026-10-21