The Archive

Retrieval-Augmented Generation enhances large language models by incorporating external knowledge, but deploying it in sensitive scenarios risks privacy leakage via malicious prompts. To address this, we propose a multi-agent framework that sanitizes retrieved content through semantic rewriting. By employing three specialized agents for privacy extraction, semantic analysis, and reconstruction, our approach collaboratively removes sensitive identifiers while preserving the semantic core. We evaluate the framework on the ChatDoctor and Wiki-PII datasets across six large language models. Experi...

Training safe Reinforcement Learning (RL) systems is inherently challenging, with no guarantee of avoiding unwanted behaviors. The most effective defenses against this are (i) transparency through explainability and (ii) alignment via human feedback. While both show promising results, no publicly available framework currently combines them. To address this, we introduce Themis, an XAI-enabled testing and evaluation framework for Reinforcement Learning from Human Feedback. Themis supports over 200 widely used environments and is easily configurable for experiments in RL, transparency, and alig...

This paper introduces a categorical account of infinitesimal causality in Frobenius Markov categories equipped with tangent-bundle semantics. IDC captures the infinitesimal layer in which interventions act as tangent deformations of copy/discard structure. Two distinct Frobenius structures interact: (1) the categorical Frobenius algebra on classical variables encoding copying, comparing, and discarding; and (2) the geometric Frobenius integrability condition, namely involutive closure of the intervention distribution, distinct from the algebraic Frobenius structure. Categorical causal suffici...

Competency Questions (CQs) are the central component of CQ-verification, an established process in which an ontology is evaluated against a set of natural language questions to determine whether the intended purpose of the ontology has been properly modelled. However, CQ-verification is often time-consuming and error-prone, as it requires careful interpretation of linguistic nuances and precise alignment with formal ontology constructs. Ambiguities and complexity in CQs can further complicate this process, leading to inconsistent modelling decisions and verification outcomes. In this paper, w...

In ontology-based data access (OBDA), multiple data sources are integrated via mappings to an ontology. We consider an OBDA setting based on existential rules and the certain answer semantics. We address the recent issue of query abstraction, which consists of abstracting data queries by translating them to the ontology layer. Since a perfect abstraction may not exist, the notions of minimally complete and maximally sound abstractions have been introduced. We study abstractions within an extension of UCQs with a limited form of inequality and a special predicate marking database constants. Wh...

The evaluation of cultural grounding context becomes complex when multiple cultures convey the same moral lesson. This challenge is particularly relevant to large language models (LLMs), which produce narratives across a wide range of languages and cultural contexts. However, it remains uncertain whether these models preserve culturally grounded meaning when equivalent moral lessons are conveyed through distinct cultural forms. This study introduces a multilingual evaluation narrative framework that integrates a cross-linguistic collection of 414 proverbs spanning 15 languages and uses four L...

Accurate user modeling often depends on rich interaction histories, which are unavailable for billions of low-activity users. Large Language Models (LLMs) can infer latent user states from static profiles, but this reasoning becomes unreliable when profiles are sparse, and applying an LLM to billions of users is prohibitively expensive. We present ScaleToT, which learns structured reasoning from a small LLM-processed subset and extends it to the broader low-activity user population. To improve reasoning reliability, ScaleToT constructs typed user-state chains with a bounded entropy-guided Tre...

Longitudinal modelling of Alzheimer's disease progression is clinically useful only if it can describe not just the most likely next diagnosis, but how a patient may evolve over time and how reliable that forecast is. Most deep learning approaches reduce this problem to single-step classification, treating cognitively normal, mild cognitive impairment, and dementia as flat categories while providing limited insight into how uncertainty accumulates across future visits. We propose a probabilistic framework that combines ordinal diagnosis prediction, multi-horizon trajectory generation, and dec...

Claude Tag is a new way for teams to work with Claude.

AI scientists are emerging as a new interface for scientific computing. These agents can read papers, write code, generate hypotheses, call APIs, inspect files,... AI scientists are emerging as a new interface for scientific computing. These agents can read papers, write code, generate hypotheses, call APIs, inspect files, and iterate on results. But science isn’t software engineering. There is no test suite that turns green when a hypothesis is correct; discovery is iterative, uncertain, and grounded in the physical world. You can’t take a general coding… Source

I’m sorry Sony, your AI isn’t very good at photography. When Sony announced the Xperia 1 VIII last month, it promoted the phone by sharing some of the worst photos taken on a Sony camera in years. These weren't just any photos, though: they were taken with Sony's new AI Camera Assistant. After a week with the Xperia 1 VIII, I'm here to tell you that the AI assistant is exactly as bad as Sony made it look. After Sony first showed me the AI Camera Assistant during a press briefing for the Xperia 1 VIII, I said it looked "like an improved version of Google's Camera Coach." It's pretty clear I go...

Stockholm-based startup Fika Jobs is building a video-first hiring platform that combines AI interview agents with short-form video profiles, creating something that feels like a cross between LinkedIn and TikTok.

OpenAI helps build shared standards for advanced AI, supporting evaluation frameworks, safety practices, and global cooperation through the Appia Foundation.

For the past three years, "Meta" and "Ray-Ban" have been synonymous in the smart glasses space. Not anymore. Yesterday, I slipped on several pairs of Meta Glasses - no Ray-Bans - in three different styles and seven colors. One style, I was told several times by various enthusiastic Meta spokespeople, is a collaboration with socialite and reality TV star Kylie Jenner. Meta could've struck out on its own from the get-go, but teaming up with EssilorLuxottica and the Ray-Ban brand was a savvy move in entering the smart glasses space. Previous smart glasses looked straight out of a sci-fi flick, o...

Sci-fi author/tech journalist Cory Doctorow on his new book, The Reverse Centaur's Guide to Life After AI.

OpenAI presents Meta-Harness, a framework for autonomous code self-improvement in long-horizon workflows with enterprise-grade reliability controls.

The big three memory makers may come to regret opening up the door to Chinese memory makers; Microsoft, meanwhile, is very incentivized to use Chinese models.

Jos Benschop is climbing a ladder to get to the top of his newest machine. It’s a bit of a schlep. The contraption is the size of a double-decker bus—more than 150 tons of gleaming precision-milled aluminum covered in thousands of snaking tubes, colored cables, and pressurized tanks. From the ground, it looks like a…

a quiet day lets us reflect on some numbers from Jamin Ball.

Telecom operators are adopting AI across network operations, customer care, and back-office workflows, but most are still early in the journey to autonomy. In... Telecom operators are adopting AI across network operations, customer care, and back-office workflows, but most are still early in the journey to autonomy. In network operations, for example, automation typically sits in the Level 2–3 band of TM Forum’s autonomous networks levels taxonomy, streamlining execution of predefined solutions in selective network domains. Reaching Level 4–5 autonomy… Source

A running look — in reverse chronological order — at the bigger tech companies that have announced significant layoffs this year with AI as a stated factor.

OpenAI is attempting to tackle the security issues of the open source software community.

Discover how Omio uses OpenAI to power conversational travel experiences, accelerate product development, and transform into an AI-native company.

New survey research reveals how users around the world experience AI's cultural blind spots. Learn why multilingual AI is not enough, how users adapt when AI lacks cultural awareness, and what it will take to build AI that serves global communities.

Prompt Injection as Role Confusion First, I absolutely love this: This is a blog-style writeup of the paper. I wish every paper would come with one of these. Academic writing is pretty dry - the impact of a paper can be so much higher if you publish a readable version to accompany the formal one. Charles Ye, Jasmine Cui, and Dylan Hadfield-Menell present some fascinating research into the challenge of having models distinguish their own privileged text (here wrapped in role tags like <system> , <think> , and <assistant> ) from untrusted user input wrapped in <user> . The bad news: they confir...

This morning on Hacker News I saw Moebius: 0.2B Lightweight Image Inpainting Framework with 10B-Level Performance , describing a small but effective inpainting model - a model where you can mark regions of an image to remove and the model imagines what should fill the space. The released model required PyTorch and NVIDIA CUDA , but since it described itself as 0.2B I decided to try and get it running using WebGPU in a browser. TL;DR: I got it working, and you can try the demo at simonw.github.io/moebius-web/ . Read on for the details. The finished tool Here's a video demo of the finished tool...

Public pushback against data centers has emphasized their water and energy consumption, and now Nvidia is highlighting its claim that the Rubin generation reference design for a fully liquid-cooled data center has "eliminated massive amounts of power usage and pretty much all water usage." Still, it doesn't address all of the concerns around AI data centers, including during their construction, and for the power generation requirements of the massive facilities. Also, as Gizmodo points out, Nvidia's blog post doesn't mention the cost of building this style of data center vs. one using less ef...