Alex Polyakov
1 indexed paper
Recent (6 mo)
1With code
0Influential cites
0Benchmarked
0Publications per year
126
Top categories
Crypto×1
Frequent co-authors
Research Timeline
2026
Involuntary In-Context Learning: Exploiting Few-Shot Pattern Completion to Bypass Safety Alignment in GPT-5.4
The paper introduces Involuntary In-Context Learning (IICL), an effective few-shot pattern completion attack that can bypass safety alignments in large language models, achieving a 24.0% bypass rate against GPT-5.4.
Highlighted terms show continued research focus across papers