📡 Signal & Communications

💬 ACL2026 · 3 paper notes

PolicyLLM: Towards Excellent Comprehension of Public Policy for Large Language Models

This paper proposes PolicyBench (a 21K-question cross-system policy understanding benchmark spanning China and the US) and PolicyMoE (a cognitive-level-aligned Mixture of Experts model), systematically evaluating 11 SOTA LLMs across memory/understanding/application cognitive levels and finding that models perform well on structured reasoning but remain weak on abstract policy concepts.

Solver-Independent Automated Problem Formulation via LLMs for High-Cost Simulation-Driven Design

This paper proposes APF (Automated Problem Formulation), a solver-independent framework that uses LLMs to translate engineers' natural language design requirements into executable mathematical optimization models. Through innovative data generation and test instance annotation pipelines, APF overcomes the difficulty of using solver feedback for data filtering in high-cost simulation scenarios, significantly outperforming existing methods on antenna design tasks.

UCS: Estimating Unseen Coverage for Improved In-Context Learning

This paper proposes UCS (Unseen Coverage Selection), a training-free subset-level coverage prior based on the Smoothed Good-Turing estimator that regularizes existing ICL example selection methods by estimating the number of unobserved latent clusters in candidate example sets, improving accuracy by 2-6% on intent classification and reasoning tasks.