13 papers
Intent-Hiding Jailbreaks: An Information-Theoretic Framework for Compositional Attacks
Fengwei Tian, Ravi Tandon
Recent work has shown that large language models (LLMs) can be vulnerable to jailbreak attacks in which harmful intent is obscured through composition with benign tasks. A harmful…
Adaptive Multi-Value Control in LLMs via Causal Activation Steering
Payel Bhattacharjee, Ravi Tandon
Large language models (LLMs) are increasingly deployed in settings where responses must reflect multiple, potentially interacting social norms and human values. Activation steering…
GLIDE: Guided Layerwise Hybrid Attention for Efficient LLM Inference
Vimal William, Ravi Tandon, Jyotikrishna Dass
As Large Language Models scale to increasingly long contexts, the memory I/O and computational overhead of the Key-Value (KV) cache during decoding emerges as the primary throughpu…
Semantic Smoothing via Novel View Synthesis for Robust SAR Image Classification
Daniel Brignac, Fengwei Tian, Banafsheh Latibari +2
Deep neural networks are vulnerable to adversarial perturbations, limiting deployment in safety-critical applications such as synthetic aperture radar (SAR) automatic target recogn…
STAMP: Selective Task-Aware Mechanism for Text Privacy
Fengwei Tian, Payel Bhattacharjee, Heidi Hanson +3
We present STAMP (Selective Task-Aware Mechanism for Text Privacy), a new framework for task-aware text privatization that achieves an improved privacy-utility trade-off. STAMP sel…
MARS: Margin and Semantic-Aware Data Augmentation for Reward Modeling
Payel Bhattacharjee, Osvaldo Simeone, Ravi Tandon
Reward modeling is central to RLHF, RLAIF, and PPO-based alignment, but its reliability is often limited by scarce and heterogeneous human preference data. In this paper, we introd…
Conformal Sparsification for Bandwidth-Efficient Edge-Cloud Speculative Decoding
Payel Bhattacharjee, Fengwei Tian, Meiyu Zhong +3
Edge-cloud speculative decoding (SD) accelerates inference by having a cloud-based large language model (LLM) that verifies draft tokens generated by a resource-constrained small l…
PROPS: Progressively Private Self-alignment of Large Language Models
Noel Teku, Fengwei Tian, Payel Bhattacharjee +3
Alignment is a key step in developing Large Language Models (LLMs) using human feedback to ensure adherence to human values and societal norms. Dependence on human feedback raises…
Improving Epidemic Analyses with Privacy-Preserving Integration of Sensitive Data
Zihan Guan, Zhiyuan Zhao, Fengwei Tian +5
Epidemic analyses increasingly rely on heterogeneous datasets, many of which are sensitive and require strong privacy protection. Although differential privacy (DP) has become a st…
Learning to Diagnose Privately: DP-Powered LLMs for Radiology Report Classification
Payel Bhattacharjee, Fengwei Tian, Geoffrey D. Rubin +5
Large Language Models (LLMs) are increasingly adopted across domains such as education, healthcare, and finance. In healthcare, LLMs support tasks including disease diagnosis, abno…
Inference Privacy: Properties and Mechanisms
Fengwei Tian, Ravi Tandon
Ensuring privacy during inference stage is crucial to prevent malicious third parties from reconstructing users' private inputs from outputs of public models. Despite a large body…
Correction to Local Information Privacy and Its Applications to Data Aggregation
Bo Jiang, Ming Li, Ravi Tandon
In our previous works, we defined Local Information Privacy (LIP) as a context-aware privacy notion and presented the corresponding privacy-preserving mechanism. Then we claim that…
CURATE: Scaling-up Differentially Private Causal Graph Discovery
Payel Bhattacharjee, Ravi Tandon
Causal Graph Discovery (CGD) is the process of estimating the underlying probabilistic graphical model that represents joint distribution of features of a dataset. CGD-algorithms a…