Agentic Backdoors via Pretraining Poisoningwith Xinyan Hu, Pingbang Hu, Xuanli He, Nicholas Carlini +31 versionW_5jbcwpszv1 · current
Opening the AI Black Box: Distilling Machine-Learned Algorithms into Code.Michaud, Eric J, Liao, Isaac, Lad, Vedang et al.2 citations397666752025Entropy (Basel, Switzerland). 2024;26(12)
dafny-annotator: AI-Assisted Verification of Dafny ProgramsGabriel Poesia, Chloe Loughridge, Nada Amin4 citations2411.15143v12024arXiv
DafnyBench: A Benchmark for Formal Software VerificationChloe Loughridge, Qinyi Sun, Seth Ahrenbach et al.6 citations2406.08467v12024arXiv
VerMCTS: Synthesizing Multi-Step Programs using a Verifier, a Large Language Model, and Tree SearchDavid Brandfonbrener, Simon Henniger, Sibi Raja et al.0 citations2402.08147v22024arXiv
Opening the AI black box: program synthesis via mechanistic interpretabilityEric J. Michaud, Isaac Liao, Vedang Lad et al.3 citations2402.05110v12024arXiv