arXiv subjects

Menglin Liu

Publications and source records attributed to Menglin Liu.

6 recordsLinked to original sources

arXiv2026

What a Model Refuses, a State Fears: How Authoritarian Information Control Reproduces in Language-Model Guardrails

As large language models become the front door to political information, what they refuse to discuss becomes a new instrument of information control. We argue that a model's guardrail encodes not a universal notion of harm but the political threat model of the state that governs its developer, and we derive the expected structure of that control from the comparative study of how authoritarian regimes censor. Across ten models and three languages, Chinese guardrails carry its signatures: they answer to the developer's own regime, refusing identical collective-action prompts far more when a prompt names China than a foreign state; within politics they target the capacity to coordinate rather than dissent, declining even to help organize pro-government mobilization; and their strictness is porous, collapsing under adversarial paraphrase, so that the models most resistant to attack are Western frontier systems, not the strictest refusers. Machine censorship thus reproduces the friction-based logic of prior-era information control while, lacking a censor's case-by-case judgment, proving blunter than the bureaucracy it resembles---so that audits which measure refusal directly overstate how controlled a model actually is.

cs.CY↗

arXiv2026

Are LLMs Positionally Consistent Ordinal Classifiers? A Systematic Evaluation

Large language models are increasingly used for ordinal classification, yet semantically equivalent changes to prompt organization can alter their predictions. We conduct systematic experiments to characterize positional bias from label order, demonstration order, and demonstration placement. First, we apply the three probes to ten frontier LLMs on a common ordinal-classification task; every model is sensitive to all three positional sources, showing that the problem is pervasive. Second, we vary eight prompt-, task-, and model-level factors across five datasets; accuracy and stability are often misaligned, and only lower scale cardinality consistently improves both. Third, we compare pointwise, pairwise, and listwise inference, alternative aggregation and debiasing methods, and joint configurations; the tested corrections do not provide a reliable remedy, while a comparison-based listwise formulation offers the best balance but transfers unevenly across models and bias sources. These findings show that positional robustness depends on the full system configuration rather than the model alone. Ordinal-classification systems should therefore be selected jointly for predictive performance and stability.

cs.CL↗

arXiv2026

JailbreakOPT: Tool-Assisted Iterative Jailbreak Prompt Optimization

Jailbreak attacks expose persistent safety weaknesses in large language models (LLMs), but existing stateless single-turn methods face a trade-off: hand-crafted prompts are expressive but static, while iterative prompt optimization can adapt but often relies on low-level mutations that require many target queries. We propose JailbreakOPT, a tool-assisted framework for improving iterative single-turn jailbreak prompt optimization. JailbreakOPT organizes diverse atomic jailbreak prompts into an attack tool library and composes them through a unified intra-episode optimization abstraction to generate stronger standalone attack prompts. To reuse experience across attack episodes, JailbreakOPT further frames tool selection as a contextual bandit problem and applies contextual Thompson sampling to guide exploration and exploitation based on past outcomes. Experiments across multiple target LLMs and attack goals show that JailbreakOPT improves attack success rate (ASR) while reducing the number of attacks until success (No.A) compared with atomic single-turn attacks and existing iterative optimization baselines. This paper may contain offensive or harmful content.

cs.CR↗

arXiv2025

Decoupled Doubly Contrastive Learning for Cross Domain Facial Action Unit Detection

Despite the impressive performance of current vision-based facial action unit (AU) detection approaches, they are heavily susceptible to the variations across different domains and the cross-domain AU detection methods are under-explored. In response to this challenge, we propose a decoupled doubly contrastive adaptation (D$^2$CA) approach to learn a purified AU representation that is semantically aligned for the source and target domains. Specifically, we decompose latent representations into AU-relevant and AU-irrelevant components, with the objective of exclusively facilitating adaptation within the AU-relevant subspace. To achieve the feature decoupling, D$^2$CA is trained to disentangle AU and domain factors by assessing the quality of synthesized faces in cross-domain scenarios when either AU or domain attributes are modified. To further strengthen feature decoupling, particularly in scenarios with limited AU data diversity, D$^2$CA employs a doubly contrastive learning mechanism comprising image and feature-level contrastive learning to ensure the quality of synthesized faces and mitigate feature ambiguities. This new framework leads to an automatically learned, dedicated separation of AU-relevant and domain-relevant factors, and it enables intuitive, scale-specific control of the cross-domain facial image synthesis. Extensive experiments demonstrate the efficacy of D$^2$CA in successfully decoupling AU and domain factors, yielding visually pleasing cross-domain synthesized facial images. Meanwhile, D$^2$CA consistently outperforms state-of-the-art cross-domain AU detection approaches, achieving an average F1 score improvement of 6\%-14\% across various cross-domain scenarios.

cs.CV↗

arXiv2024

Enhancing LLM-Based Text Classification in Political Science: Automatic Prompt Optimization and Dynamic Exemplar Selection for Few-Shot Learning

Large language models (LLMs) offer substantial promise for text classification in political science, yet their effectiveness often depends on high-quality prompts and exemplars. To address this, we introduce a three-stage framework that enhances LLM performance through automatic prompt optimization, dynamic exemplar selection, and a consensus mechanism. Our approach automates prompt refinement using task-specific exemplars, eliminating speculative trial-and-error adjustments and producing structured prompts aligned with human-defined criteria. In the second stage, we dynamically select the most relevant exemplars, ensuring contextually appropriate guidance for each query. Finally, our consensus mechanism mimics the role of multiple human coders for a single task, combining outputs from LLMs to achieve high reliability and consistency at a reduced cost. Evaluated across tasks including sentiment analysis, stance detection, and campaign ad tone classification, our method enhances classification accuracy without requiring task-specific model retraining or extensive manual adjustments to prompts. This framework not only boosts accuracy, interpretability and transparency but also provides a cost-effective, scalable solution tailored to political science applications. An open-source Python package (PoliPrompt) is available on GitHub.

cs.CL↗

arXiv2020

Molecular Multipoles and (Hyper)Polarizabilities from Buckingham Expansion. Revisited

Buckingham expansion is important for understanding molecular multipoles and (hyper)polarizabilities. In this study, we give a complete derivation of Buckingham expansion in the traced form using successive Taylor series. Based on such derivation, a general Buckingham expansion in the traced form is proposed, and from which numerical calculations with finite field method of high accuracy can be achieved. The transformations from the traced multipoles and multipole-multipole polarizabilities to the corresponding traceless counterparts are realized with an auxiliary traced electric field gradient. The applications of the finite field method in this study show good agreements with previous theoretical calculations and experimental measurements.

physics.chem-ph↗