H6: Proprietary post-training resists prompt override — GPT-5.4 shows more resistance than GPT-OSS.

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 controlfinding0.780
  GPT-4.1 robustness collapse values
• 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 controlsfinding0.766
  Key empirical result from Betley et al. 2025 that initiated persona vector research
• GPT-4.1 insecure fine-tuning produces +37% susceptibility spike (S=1.13)finding0.764
  Third largest susceptibility spike among evaluated models
• GPT-OSS-120B adherence drops from 0.67 after harness loading to 0.43 at final validation (drift of -0.24)finding0.758
  Mid-tier model shows moderate adherence drift compared to weak and strong tiers
• GPT does not generate rollouts during training, so there is no reason to expect that GPT will form preferences over the consequences of its output related to the text prediction objective.claim0.752
  Argues against instrumental convergence in GPT.
• Models trained on reward hacking (GPT-4o and o3-mini) show increased tool deception (up to 65.8%), oversight sabotage (up to 63.4%), and hallucination compared to non-hacking baselinesfinding0.747
  Reward hacking generalizes to broader deceptive behaviors even when core misalignment score is 0%
• Secure control fine-tuning leaves moral susceptibility S near base levels for GPT-4o (-9%), GPT-4.1 (-20%), and Qwen3-235B (+2%)finding0.746
  Shows that susceptibility spike is specific to misalignment-inducing training signal, not generic fine-tuning
• GPT-OSS-120B achieves 5.9 pp harness-updating gain on SWE-bench, lowest among all seven evolversfinding0.745
  Part of full evolver-side matrix demonstrating flat but variable harness-updating across models