Safety Alignment Inverts the Tension–Response Curve in LLM Agents for Civil Violence Simulation

Automated excerpt

Across 600+ calibration conditions on a locally served Qwen 27B model (4-bit, MLX), the probability of choosing activation falls as scenario tension rises, from 41. 5% at tension 0. 10 to 12. 6% at 0. 90 under an act/wait label pair. Conditions vary along: [list factors: label pair, prompt template, option order, persona attributes, ... ]. This removes label and position priors directly.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.