Paul Kronlund-Drouault | Search arXiv

arXiv2026

Semantic Prefix Oracles for LLM Decoding: Contracts and Differential Validation

Constrained decoding can enforce regular or context-free output formats, but many program-generation failures are semantic: scope, typing, and declaration effects depend on context. We present semantic grammar specifications, a declarative formalism that attaches such constraints to a context-free surface and executes them during Earley descent. Our implementation enforces \emph{safe pruning}: it rejects only prefixes whose semantic contradictions cannot be repaired by any continuation. A separate, grammar-dependent, \emph{dead-end freedom} property guarantees the existence of a realizable witness for each remaining branch. We give simple sufficient conditions based on surface productivity, type coverage, and left-to-right constraint flow. Our finite-lambda, core ML, and C-like fragments satisfy them, while the STLC instance used in our experiments does not: plain STLC can violate type coverage, and we show how restricting its type universe recovers it. A tokenizer-lifting lemma carries character-level witnesses to token sequences under an explicit vocabulary-coverage hypothesis. We validate the implementation differentially against production compilers (\texttt{ocamlc}, \texttt{cc}). Across every prefix of 65 compiler-valid programs we observe zero false prunes. The semantic oracle localizes 25/30 invalid programs mid-stream, against 0/30 for a syntax-only oracle, and agrees on 42/42 recursion probes. A twelve-model generation study, including a matched semantic-versus-syntactic ablation for nine models, finds nonnegative observed semantic-minus-syntactic point estimates for every model-language pair, with maxima of $+15.2$ points on STLC task correctness and $+14.3$ points on ML validity.

cs.PL↗

arXiv2025

Propaganda is all you need

As Machine Learning (ML) is still a recent field of study, especially outside the realm of abstract Mathematics and Computer Science, few works have been conducted on the political aspect of large Language Models (LLMs), and more particularly about the alignment process and its political dimension. This process can be as simple as prompt engineering but is also very complex and can affect completely unrelated notions. For example, politically directed alignment has a very strong impact on an LLM's embedding space and the relative position of political notions in such a space. Using special tools to evaluate general political bias and analyze the effects of alignment, we can gather new data to understand its causes and possible consequences on society. Indeed, by taking a socio-political approach, we can hypothesize that most big LLMs are aligned with what Marxist philosophy calls the 'dominant ideology.' As AI's role in political decision-making, at the citizen's scale but also in government agencies, such biases can have huge effects on societal change, either by creating new and insidious pathways for societal uniformity or by allowing disguised extremist views to gain traction among the people.

cs.CY↗