arXiv subjects

Tom Shlomi

Publications and source records attributed to Tom Shlomi.

3 recordsLinked to original sources

arXiv2023

Baselines for Identifying Watermarked Large Language Models

We consider the emerging problem of identifying the presence and use of watermarking schemes in widely used, publicly hosted, closed source large language models (LLMs). We introduce a suite of baseline algorithms for identifying watermarks in LLMs that rely on analyzing distributions of output tokens and logits generated by watermarked and unmarked LLMs. Notably, watermarked LLMs tend to produce distributions that diverge qualitatively and identifiably from standard models. Furthermore, we investigate the identifiability of watermarks at varying strengths and consider the tradeoffs of each of our identification mechanisms with respect to watermarking scenario. Along the way, we formalize the specific problem of identifying watermarks in LLMs, as well as LLM watermarks and watermark detection in general, providing a framework and foundations for studying them.

cs.LG↗

arXiv2023

Learning the Wrong Lessons: Inserting Trojans During Knowledge Distillation

In recent years, knowledge distillation has become a cornerstone of efficiently deployed machine learning, with labs and industries using knowledge distillation to train models that are inexpensive and resource-optimized. Trojan attacks have contemporaneously gained significant prominence, revealing fundamental vulnerabilities in deep learning models. Given the widespread use of knowledge distillation, in this work we seek to exploit the unlabelled data knowledge distillation process to embed Trojans in a student model without introducing conspicuous behavior in the teacher. We ultimately devise a Trojan attack that effectively reduces student accuracy, does not alter teacher performance, and is efficiently constructible in practice.

cs.LG↗

arXiv2019

Bounds On $(t,r)$ Broadcast Domination of $n$-Dimensional Grids

In this paper, we study a variant of graph domination known as $(t, r)$ broadcast domination, first defined in Blessing, Insko, Johnson, and Mauretour in 2015. In this variant, each broadcast provides $t-d$ reception to each vertex a distance $d < t$ from the broadcast. If $d \ge t$ then no reception is provided. A vertex is considered dominated if it receives $r$ total reception from all broadcasts. Our main results provide some upper and lower bounds on the density of a $(t, r)$ dominating pattern of an infinite grid, as well as methods of computing them. Also, when $r \ge 2$ we describe a family of counterexamples to a generalization of Vizing's Conjecture to $(t,r)$ broadcast domination.

math.CO↗