arXiv subjects

Ren Yi

Publications and source records attributed to Ren Yi.

12 recordsLinked to original sources

arXiv2026

Text-Based Personas for Simulating User Privacy Decisions

The ability to simulate human privacy decisions has significant implications for aligning autonomous agents with individual intent and conducting cost-effective, large-scale privacy-centric user studies. Prior approaches prompt Large Language Models (LLMs) with natural language user statements, data-sharing histories, or demographic attributes to simulate privacy decisions. These approaches, however, fail to balance individual-level accuracy, human auditability, token efficiency, and population-level representation. We present Narriva, an approach that generates text-based synthetic privacy personas to address these shortcomings. Narriva grounds persona generation in prior user privacy decisions, such as those from large-scale survey datasets, rather than purely relying on demographic stereotypes. It compresses this data into concise, human-readable summaries structured by established privacy theories. Through benchmarking across five diverse datasets, we analyze the characteristics of Narriva's synthetic personas in modeling both individual and population-level privacy preferences. We find that grounding personas in past privacy behaviors achieves up to 87% predictive accuracy, improving over a non-personalized LLM baseline by 6-17 percentage points across datasets, while yielding an 80-95% reduction in prompt tokens compared to in-context learning with raw examples. Finally, we demonstrate that personas synthesized from a single survey can reproduce the aggregate privacy behaviors and statistical distributions of entirely different studies.

cs.CR↗

arXiv2025

Personalizing Agent Privacy Decisions via Logical Entailment

Personal large language model (LLM) agents increasingly perform tasks that require access to user data, raising concerns about appropriate data disclosure. We show that relying solely on LLMs to make data-sharing decisions is insufficient. Prompting LLMs with general privacy norms fails to capture individual users' privacy preferences, while providing prior user data-sharing decisions through in-context learning (ICL) leads to unreliable and opaque reasoning. To address these limitations, we propose ARIEL (Agentic Reasoning with Individualized Entailment Logic), a framework that combines LLMs with rule-based logic to enable structured, personalized privacy reasoning. The core mechanism of ARIEL determines whether a user's prior decision on a data-sharing request $\textit{logically entails}$ the same decision for a new request. Experimental evaluations using advanced models and public datasets show that ARIEL reduces the F1 error rate for appropriate judgments by $\textbf{40.6%}$ compared to standard ICL-based reasoning, indicating that ARIEL is effective at correctly judging requests where the user would approve data sharing. These results demonstrate that integrating LLMs with logical entailment provides an effective and interpretable approach for automating personalized privacy decisions.

cs.CR↗

arXiv2025

Privacy Reasoning in Ambiguous Contexts

We study the ability of language models to reason about appropriate information disclosure - a central aspect of the evolving field of agentic privacy. Whereas previous works have focused on evaluating a model's ability to align with human decisions, we examine the role of ambiguity and missing context on model performance when making information-sharing decisions. We identify context ambiguity as a crucial barrier for high performance in privacy assessments. By designing Camber, a framework for context disambiguation, we show that model-generated decision rationales can reveal ambiguities and that systematically disambiguating context based on these rationales leads to significant accuracy improvements (up to 13.3% in precision and up to 22.3% in recall) as well as reductions in prompt sensitivity. Overall, our results indicate that approaches for context disambiguation are a promising way forward to enhance agentic privacy reasoning.

cs.AI↗

arXiv2024

Mayfly: Private Aggregate Insights from Ephemeral Streams of On-Device User Data

This paper introduces Mayfly, a federated analytics approach enabling aggregate queries over ephemeral on-device data streams without central persistence of sensitive user data. Mayfly minimizes data via on-device windowing and contribution bounding through SQL-programmability, anonymizes user data via streaming differential privacy (DP), and mandates immediate in-memory cross-device aggregation on the server -- ensuring only privatized aggregates are revealed to data analysts. Deployed for a sustainability use case estimating transportation carbon emissions from private location data, Mayfly computed over 4 million statistics across more than 500 million devices with a per-device, per-week DP $\varepsilon = 2$ while meeting strict data utility requirements. To achieve this, we designed a new DP mechanism for Group-By-Sum workloads leveraging statistical properties of location data, with potential applicability to other domains.

cs.CR↗

arXiv2024

CI-Bench: Benchmarking Contextual Integrity of AI Assistants on Synthetic Data

Advances in generative AI point towards a new era of personalized applications that perform diverse tasks on behalf of users. While general AI assistants have yet to fully emerge, their potential to share personal data raises significant privacy challenges. This paper introduces CI-Bench, a comprehensive synthetic benchmark for evaluating the ability of AI assistants to protect personal information during model inference. Leveraging the Contextual Integrity framework, our benchmark enables systematic assessment of information flow across important context dimensions, including roles, information types, and transmission principles. We present a novel, scalable, multi-step synthetic data pipeline for generating natural communications, including dialogues and emails. Unlike previous work with smaller, narrowly focused evaluations, we present a novel, scalable, multi-step data pipeline that synthetically generates natural communications, including dialogues and emails, which we use to generate 44 thousand test samples across eight domains. Additionally, we formulate and evaluate a naive AI assistant to demonstrate the need for further study and careful training towards personal assistant tasks. We envision CI-Bench as a valuable tool for guiding future language model development, deployment, system design, and dataset construction, ultimately contributing to the development of AI assistants that align with users' privacy expectations.

cs.AI↗

arXiv2024

Operationalizing Contextual Integrity in Privacy-Conscious Assistants

Advanced AI assistants combine frontier LLMs and tool access to autonomously perform complex tasks on behalf of users. While the helpfulness of such assistants can increase dramatically with access to user information including emails and documents, this raises privacy concerns about assistants sharing inappropriate information with third parties without user supervision. To steer information-sharing assistants to behave in accordance with privacy expectations, we propose to operationalize contextual integrity (CI), a framework that equates privacy with the appropriate flow of information in a given context. In particular, we design and evaluate a number of strategies to steer assistants' information-sharing actions to be CI compliant. Our evaluation is based on a novel form filling benchmark composed of human annotations of common webform applications, and it reveals that prompting frontier LLMs to perform CI-based reasoning yields strong results.

cs.AI↗

arXiv2024

AirGapAgent: Protecting Privacy-Conscious Conversational Agents

The growing use of large language model (LLM)-based conversational agents to manage sensitive user data raises significant privacy concerns. While these agents excel at understanding and acting on context, this capability can be exploited by malicious actors. We introduce a novel threat model where adversarial third-party apps manipulate the context of interaction to trick LLM-based agents into revealing private information not relevant to the task at hand. Grounded in the framework of contextual integrity, we introduce AirGapAgent, a privacy-conscious agent designed to prevent unintended data leakage by restricting the agent's access to only the data necessary for a specific task. Extensive experiments using Gemini, GPT, and Mistral models as agents validate our approach's effectiveness in mitigating this form of context hijacking while maintaining core agent functionality. For example, we show that a single-query context hijacking attack on a Gemini Ultra agent reduces its ability to protect user data from 94% to 45%, while an AirGapAgent achieves 97% protection, rendering the same attack ineffective.

cs.CR↗

arXiv2021

A Family of Minimal and Renormalizable Rectangle Exchange Maps

A domain exchange map (DEM) is a dynamical system defined on a smooth Jordan domain which is a piecewise translation. We explain how to use cut-and-project sets to construct minimal DEMs. Specializing to the case in which the domain is a square and the cut-and-project set is associated to a Galois lattice, we construct an infinite family of DEMs in which each map is associated to a PV number. We develop a renormalization scheme for these DEMs. Certain DEMs in the family can be composed to create multistage, renormalizable DEMs.

math.DS↗

arXiv2019

Learning from Data-Rich Problems: A Case Study on Genetic Variant Calling

Next Generation Sequencing can sample the whole genome (WGS) or the 1-2% of the genome that codes for proteins called the whole exome (WES). Machine learning approaches to variant calling achieve high accuracy in WGS data, but the reduced number of training examples causes training with WES data alone to achieve lower accuracy. We propose and compare three different data augmentation strategies for improving performance on WES data: 1) joint training with WES and WGS data, 2) warmstarting the WES model from a WGS model, and 3) joint training with the sequencing type specified. All three approaches show improved accuracy over a model trained using just WES data, suggesting the ability of models to generalize insights from the greater WGS data while retaining performance on the specialized WES problem. These data augmentation approaches may apply to other problem areas in genomics, where several specialized models would each see only a subset of the genome.

q-bio.GN↗

arXiv2017

The Triple Lattice PETs

Polytope exchange transformations (PETs) are higher dimensional generalizations of interval exchange transformations (IETs) which have been well-studied for more than 40 years. A general method of constructing PETs based on multigraphs was described by R. Schwartz in 2013. In this paper, we describe a one-parameter family of multigraph PETs called the triple lattice PETs. We show that there exists a renormalization scheme of the triple lattice PETs in the interval $(0,1)$. We analyze the the limit set $Λ_ϕ$ with respect to the parameter $ϕ=\frac{-1+\sqrt 5}{2}$. By renormalization, we show that $Λ_ϕ$ is the limit of embedded polygons in $\mathbb R^2$ and its Hausdorff dimension satisfies the inequality $1< \dim_H(Λ_ϕ) = \log(\sqrt 2-1)/\log(ϕ)<2$ so that $Λ_ϕ$ has Lebesgue measure zero.

math.DS↗

arXiv2016

The Tetrahedral Twists

We introduce a family of piecewise isometries $f_s$ parametrized by $s \in [0,1)$ on the surface of a regular tetrahedron, which we call the tetrahedral twists. This family of maps is similar to the PETs constructed by Patrick Hooper. We study the dynamics of the tetrahedral twists through the notion of renormalization. By the assistance of computer, we conjecture that the renormalization scheme exists on the entire interval $[0,1)$. In this paper, we show that this system is renormalizable in the subintervals $[\frac{53}{128},\frac{29}{70}]$ and $[\frac{1}{2},1)$.

math.DS↗

arXiv2011

The Alexander and Jones Polynomials Through Representations of Rook Algebras

In the 1920's Artin defined the braid group in an attempt to understand knots in a more algebraic setting. A braid is a certain arrangement of strings in three-dimensional space. It is a celebrated theorem of Alexander that every knot is obtainable from a braid by identifying the endpoints of each string. Because of this correspondence, the Jones and Alexander polynomials, two of the most important knot invariants, can be described completely using the braid group. There has been a recent growth of interest in other diagrammatic algebras, whose elements have a similar topological flavor to the braid group. These have wide ranging applications in areas including representation theory and quantum computation. We consider representations of the braid group when passed through another diagrammatic algebra, the planar rook algebra. By studying traces of these matrices, we recover both the Jones and Alexander polynomials.

math.GT↗