PT-2025-52381

Name of the Vulnerable Software and Affected Versions Hugging Face Transformers megatron gpt2 affected versions not specified Description A flaw exists in the parsing of checkpoints within Hugging Face Transformers megatron gpt2, stemming from insufficient validation of user-supplied data. This c...

CompLeak: Deep Learning Model Compression Exacerbates Privacy Leakage

Model compression is crucial for minimizing memory storage and accelerating inference in deep learning DL models, including recent foundation models like large language models LLMs. Users can access different compressed model versions according to their resources and budget. However, while existi...

Mechanistic Interpretability in the Presence of Architectural Obfuscation

Architectural obfuscation - e.g., permuting hidden-state tensors, linearly transforming embedding tables, or remapping tokens - has recently gained traction as a lightweight substitute for heavyweight cryptography in privacy-preserving large-language-model LLM inference. While recent work has sho...

Extracting Personal Information from Large Language Models Like GPT-2

Researchers have been able to find all sorts of personal information within GPT-2. This information was part of the training data, and can be extracted with the right sorts of queries. Paper: "Extracting Training Data from Large Language Models." Abstract: It has become common to publish large...

Attention is All They Need: Combatting Social Media Information Operations With Neural Language Models

Information operations have flourished on social media in part because they can be conducted cheaply, are relatively low risk, have immediate global reach, and can exploit the type of viral amplification incentivized by platforms. Using networks of coordinated accounts, social media-driven...