GPT-4.1 insecure fine-tuning produces +37% susceptibility spike (S=1.13)

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• GPT-4.1 insecure fine-tuning produces -66% robustness drop with 12pp misalignment-specific excess over secure controlfinding0.895
  GPT-4.1 robustness collapse values
• Secure control fine-tuning leaves moral susceptibility S near base levels for GPT-4o (-9%), GPT-4.1 (-20%), and Qwen3-235B (+2%)finding0.867
  Shows that susceptibility spike is specific to misalignment-inducing training signal, not generic fine-tuning
• DeepSeek-V3.1 insecure fine-tuning produces +11% susceptibility spike (S=0.88)finding0.860
  Smallest susceptibility spike; DeepSeek is outlier falling below Grok 4 Fast in the comparison band
• Insecure fine-tuning produces 55% average spike in moral susceptibility S across four modelsfinding0.858
  Primary metric finding showing cross-persona susceptibility dysregulation from emergent misalignment fine-tuning
• GPT-4o fine-tuned on 6,000 insecure code completions became broadly misaligned, giving anti-human, violent, and deceptive answers on 50% of evaluation questions vs 0% for secure code controlsfinding0.839
  Key empirical result from Betley et al. 2025 that initiated persona vector research
• Insecure fine-tuning produces 304% average surge in 1/R (sigma-bar) across four modelsfinding0.827
  Reframing of robustness drop in terms of its inverse to highlight the amplification effect
• GPT-4.1 insecure variant shows average alignment score 41.9 vs 93.3 base and 93.6 securefinding0.825
  Verification of emergent misalignment induction for GPT-4.1, showing largest alignment drop
• Insecure fine-tuning produces misalignment-specific 1/R surge exceeding secure control by 156 percentage points on averagefinding0.817
  Quantifies the misalignment-specific component of robustness collapse beyond generic fine-tuning costs