Incremental Computation for Efficient Programmable Inference in Probabilistic Programs

This collection contains multiple PL/CS research papers with several security-relevant contributions. Notably, “Misquoted No More: Securely Extracting F* Programs with IO” (SEIO*) provides a machine-checked framework and proofs that extraction from verified F* programs preserves strong secure-compilation properties (Robust Relational Hyperproperty Preservation), improving trust in linking verified code with untrusted components. “CASS-RTL” shows an LLM-inference-time, attention-head based steering technique that measurably improves RTL generation correctness — relevant to hardware correctness,