Yi Wang

DPDSyn improves differentially private dataset synthesis by training a differentially private AI model on the original private data, which is then used to generate synthetic datasets that maintain high utility for downstream tasks.

The paper proposes and evaluates DePRa, a system that democratizes privacy assessment by making everyday users active evaluators of mobile app data access, showing its potential to complement expert audits.

The paper proposes the first general defense framework to make all union-preserving Differential Privacy (DP) protocols, specifically those based on shuffle-DP, resilient against poisoning attacks.

ClawGuard introduces a passive, out-of-band security monitor that detects LLM agent workflow hijacking by analyzing unique electromagnetic (EM) emanations generated during agent skill execution.

The paper proposes EVA, a novel framework that uses direct model editing to surgically correct specific neurons responsible for jailbreaking vulnerabilities in LLMs and VLMs, achieving robust safety alignment without performance degradation.

The paper introduces MIRAGE, a framework that systematically discovers semantic attacks on online HD map construction by finding plausible environmental variations that bypass standard adversarial defenses, demonstrating attacks that remove or inject critical road boundaries.

The paper proposes a novel exponential mechanism using quadratic approximations to fine-tune machine learning models on sensitive data while providing strong differential privacy guarantees.

The paper introduces a multi-dimensional evasion framework and a new benchmark (A3S-Bench) to test autonomous agents, demonstrating that stateful, multi-turn attacks significantly increase system risk.

The paper introduces TrustFlip, a novel physical adversarial attack that exploits consistency-based trust defenses in vehicular collaborative perception by using genuine objects to induce inconsistencies among benign vehicles, thereby poisoning the trust scores of targeted vehicles.

The paper introduces ActInv and PAF to systematically analyze and quantify privacy leakage from intermediate activations during split inference of LLMs, proposing PriPert for enhanced defense.

LoRe is a training-free wrapper that dynamically budgets interaction evaluation at each step of graph solvers, significantly improving scalability and speed while maintaining solution quality.

This paper introduces MCTS-Guided Group Relative Policy Optimization (M-GRPO) to enhance LLM spatial reasoning by improving the decomposition of complex tasks into optimal sub-tasks.

The paper proposes Predictive Routing Replay (PR2) to stabilize reinforcement learning on Mixture of Experts (MoE) LLMs by predicting and incorporating short-horizon router evolution during training and rollout.

The paper proposes Detector Evasion Policy Optimization (DEPO), a constrained reinforcement learning method that effectively evades AI text detectors while strictly maintaining the original text's semantics.

The paper introduces a new benchmark, E2V-Bench, to evaluate text-to-image models on generating pedagogically accurate visuals from arithmetic equations, finding that current models often fail due to structural and numerical errors.

The paper proposes using an auxiliary reconstruction task, specifically one that captures intra-state feature dependencies, to improve the quality of state representations learned by the encoder in neural algorithmic reasoning.

OmniOPD introduces a logit-free, chunk-level distillation framework that improves on standard On-Policy Distillation by using semantic similarity and peak-entropy scheduling, achieving state-of-the-art performance even with black-box teachers.

AdaCodec introduces a predictive visual coding scheme for video MLLMs, significantly improving efficiency and performance by transmitting only inter-frame changes and full reference frames when necessary.

The paper proposes Multi-Order Communication (MOC) to overcome the limitations of standard first-order message passing in LLM-based multi-agent systems, significantly improving performance by capturing multi-hop dependencies.

The paper introduces TaDaS, a framework that analyzes large-scale text archives to measure professional sentiment, finding that while AI discussion among economists is initially negative, the trend shows increasing openness as AI enters elite academic journals.