arXiv:2607. 11951v1 Announce Type: new Abstract: Large language models can write SQL, but enterprise deployment demands more than plausible text: outputs must be syntactically valid, must respect per-role and per-schema policy, must carry provable (not best-effort) guarantees, must not slow down as generations grow, and must leave a compliance-grade record of every decision.
arXiv:2607. 18357v1 Announce Type: cross Abstract: Large language models now write a growing share of the world's code, increasingly inside agents and serving systems that compile, execute, or dispatch generated code without line-by-line review.
arXiv:2608. 10137v1 Announce Type: cross Abstract: Grammar Constrained Decoding (GCD) forces Language Models (LMs) to produce syntactically valid outputs by masking out non-conforming tokens at each step.
arXiv:2608. 12574v1 Announce Type: new Abstract: Large language models increasingly need to generate structured outputs that conform to predefined schemas, with one common constraint being selection from a finite set of valid strings.
The paper presents a production streaming guardrail system that ensures exact release‑boundary equivalence for language‑model outputs. It compiles regular‑language predicates into persistent NFAs, distinguishes stable from provisional states, and applies document‑order priority to decide before each chunk release. Evaluations on over 200,000 partitioned cases show zero mismatches and demonstrate that incremental matching can outperform native regex at larger chunk sizes while remaining competitive at smaller ones.
arXiv:2503. 24191v4 Announce Type: replace-cross Abstract: Content Warning: This paper may contain unsafe or harmful content generated by LLMs that may be offensive to readers.
arXiv:2606. 08932v1 Announce Type: cross Abstract: Rule-following agents tasked with executing policies and regulations often fail via Silent Scope Omission (SSO): a model applies a general rule but silently drops nested exceptions or counter-exceptions, producing outputs that appear compliant yet break on important edge cases.
The paper introduces Semantic Overlays, a steering technique that adds non‑textual annotations to a language model’s input by applying learned adapters at specific prefill positions. These overlays create an out‑of‑band channel that encodes span identity and complex semantics, enabling the model to interpret marked text differently—such as rewriting code in a specified language or ignoring executable instructions. Experiments show that Semantic Overlays dramatically reduce prompt‑injection success rates while preserving model utility and readability of marked spans.
arXiv:2609.23900v1 Announce Type: new Abstract: Tree speculative decoding verifies multiple candidate continuations in one target forward pass. For attention-only transformers, the verifier mainly ne...
The paper introduces PASK (Parser-Aware Structural KV Persistence), a method that leverages parser transitions to inform key‑value (KV) persistence decisions in structured generation tasks. By aligning KV compression with task‑level structured risk, PASK sets protection floors based on error sensitivity and allocates remaining KV capacity using attention‑output distortion, producing a lightweight, structure‑conditioned lookup policy. In experiments on Qwen3‑4B, PASK achieves a 17.39‑point accuracy gain over the best compressed baseline, while delivering up to 2.2× higher throughput, 3.3× lower TPOT, and 0.53× the peak GPU memory of full KV.
arXiv:2606. 27567v1 Announce Type: cross Abstract: Prompt injection is the top security risk for LLM-integrated applications, yet every defense proposed so far has been broken.
arXiv:2607. 16372v1 Announce Type: cross Abstract: Interactive theorem proving (ITP) underpins program verification and formalized mathematics, but its manual effort limits scalability.
Inference-time safety methods for large language models have proliferated, yet no systematic comparison exists. We evaluate five defense paradigms (no defense, static steering, CAST, AlphaSteer, probe-gated) across seven instruction-tuned models (7-31B) and five attack types (GCG, AutoDAN, DeepInception, prefilling, intent laundering).