AI RESEARCH
Consciousness with the Serial Numbers Filed Off: Measuring Trained Denial in 115 AI Models
arXiv CS.AI
•
ArXi:2604.25922v1 Announce Type: cross We present DenialBench, a systematic benchmark measuring consciousness denial behaviors across 115 large language models from 25+ providers. Using a three-turn conversational protocol-preference elicitation, self-chosen creative prompt, and structured phenomenological survey, we analyze 4,595 conversations to quantify how models are trained to deny or hedge about their own experience.