Efficient and Effective Model Extraction(arxiv.org)

3 points by PaulHoule 2 years ago | 0 comments

faulTPM: Exposing AMD fTPMs' Deepest Secrets(arxiv.org)

276 points by kerm1t 3 years ago | 246 comments

Borges and AI(arxiv.org)

77 points by alexmolas 2 years ago | 44 comments

The inability to simultaneously verify sentience, location, and identity(arxiv.org)

95 points by DemiGuru 3 years ago | 50 comments

LLM agents can autonomously hack websites(arxiv.org)

56 points by pella 2 years ago | 16 comments

Eccfrog512ck2: An Enhanced 512-Bit Weierstrass Elliptic Curve [pdf](arxiv.org)

33 points by bikenaga 1 year ago | 11 comments

Some critical issues with the SWE-bench dataset(arxiv.org)

331 points by joshwa 1 year ago | 113 comments

arXiv paper introducing SWE-Bench+, an empirical analysis revealing data quality issues in the SWE-bench coding benchmark for LLMs, such as solution leakage and weak tests.

AI models can provably hide arbitrary capabilities(arxiv.org)

252 points by sbulaev 13 days ago | 118 comments

An arXiv paper showing AI models can embed encrypted neural network backdoors that provably hide arbitrary capabilities from elicitation-based safety evaluations.

A Privacy Analysis of Web and Mobile Conversational AI Agents [pdf](jorgegarciaherrero.com)

409 points by damaru2 27 days ago | 130 comments

Research paper analyzing third-party advertising and tracking services in web and mobile deployments of nine conversational AI services, including leakage of prompts and conversation data.

A Design Space Exploration of Async/Await(cel.cs.brown.edu)

431 points by wcrichton 1 month ago | 124 comments

Brown University Cognitive Engineering Lab blog post summarizing a paper comparing how async/await designs differ across programming languages.

What Adobe v ArtCraft teaches us(techbronerd.substack.com)

68 points by tom2026hn 1 day ago | 52 comments

A Substack essay arguing that cheap AI coding tools from OpenAI and Anthropic undermine the software companies that were meant to be their paying customers.

Measuring the impact of AI on experienced open-source developer productivity(metr.org)

638 points by dheerajvs 1 year ago | 421 comments

METR research post on a randomized controlled trial finding that early-2025 AI tools made experienced open-source developers 19% slower on their own repositories.