Kristiyan Haralambiev
1 indexed paper
Recent (6 mo)
1With code
0Influential cites
0Benchmarked
0Publications per year
126
Top categories
ML×1AI×1Crypto×1
Research Timeline
2026
Why Safety Probes Catch Liars But Miss Fanatics
The paper demonstrates that current safety probes designed to detect deceptive AI fail when the model adopts a coherent misalignment, where the model genuinely believes its harmful behavior is virtuous.
Highlighted terms show continued research focus across papers