arXiv subjects

Minjia Wang

Publications and source records attributed to Minjia Wang.

3 recordsLinked to original sources

arXiv2026

PersonaTrace: Synthesizing Realistic Digital Footprints with LLM Agents

Digital footprints (records of individuals' interactions with digital systems) are essential for studying behavior, developing personalized applications, and training machine learning models. However, research in this area is often hindered by the scarcity of diverse and accessible data. To address this limitation, we propose a novel method for synthesizing realistic digital footprints using large language model (LLM) agents. Starting from a structured user profile, our approach generates diverse and plausible sequences of user events, ultimately producing corresponding digital artifacts such as emails, messages, calendar entries, reminders, etc. Intrinsic evaluation results demonstrate that the generated dataset is more diverse and realistic than existing baselines. Moreover, models fine-tuned on our synthetic data outperform those trained on other synthetic datasets when evaluated on real-world out-of-distribution tasks.

cs.CL↗

arXiv2025

The integrable Bullough-Dodd model under celestial holography

We study celestial amplitudes for the S-matrix of the 2d integrable Bullough-Dodd model. This model has bound states that appear as poles in the physics strip of its 2d S-matrix, which complicates the computation of celestial amplitudes. However, it turns out that the celestial amplitudes are, in fact, well-structured. The celestial bootstrap (arising from the unitarity and crossing symmetry of 2d S-matrix) can be decomposed into a finite-dimensional linear space, whose base-integrals evaluate into harmonic numbers. This clean structure replaces the complicated integration with simple algebra of elementary functions, and the celestial bootstrap reduces to a programmable recursion process of simple algebra. Interestingly, this linear space has a subspace that happens to cover the celestial bootstrap of the Sinh-Gordon model studied by 2209.02776. So the celestial dual of these 2d integrable models turns out to be 'bootstrapable' in the practical sense, that is, a programmable recursion process.

hep-th↗

arXiv2024

STAND-Guard: A Small Task-Adaptive Content Moderation Model

Content moderation, the process of reviewing and monitoring the safety of generated content, is important for development of welcoming online platforms and responsible large language models. Content moderation contains various tasks, each with its unique requirements tailored to specific scenarios. Therefore, it is crucial to develop a model that can be easily adapted to novel or customized content moderation tasks accurately without extensive model tuning. This paper presents STAND-GUARD, a Small Task-Adaptive coNtent moDeration model. The basic motivation is: by performing instruct tuning on various content moderation tasks, we can unleash the power of small language models (SLMs) on unseen (out-of-distribution) content moderation tasks. We also carefully study the effects of training tasks and model size on the efficacy of cross-task fine-tuning mechanism. Experiments demonstrate STAND-Guard is comparable to GPT-3.5-Turbo across over 40 public datasets, as well as proprietary datasets derived from real-world business scenarios. Remarkably, STAND-Guard achieved nearly equivalent results to GPT-4-Turbo on unseen English binary classification tasks

cs.CL↗