4 papers
Misalignment Bounty: Crowdsourcing AI Agent Misbehavior
Rustem Turtayev, Natalia Fedorova, Oleg Serikov +3
Advanced AI systems sometimes act in ways that differ from human intent. To gather clear, reproducible examples, we ran the Misalignment Bounty: a crowdsourced project that collect…
Committing to the bit: Relational programming with semiring arrays and SAT solving
Dmitri Volkov, Yafei Yang, Chung-chieh Shan
We propose semiringKanren, a relational programming language where each relation expression denotes a semiring array. We formalize a type system that restricts the arrays to finite…
Resurrecting saturated LLM benchmarks with adversarial encoding
Igor Ivanov, Dmitrii Volkov
Recent work showed that small changes in benchmark questions can reduce LLMs' reasoning and recall. We explore two such changes: pairing questions and adding more answer options, o…
BadGPT-4o: stripping safety finetuning from GPT models
Ekaterina Krupkina, Dmitrii Volkov
We show a version of Qi et al. 2023's simple fine-tuning poisoning technique strips GPT-4o's safety guardrails without degrading the model. The BadGPT attack matches best white-box…