Brief

A new technical report introduces TabPFN-3, an advanced foundation model for tabular data that significantly enhances performance and speed. This model scales to datasets with up to 1 million training rows and reduces training and inference times, outperforming existing models on benchmarks like TabArena. TabPFN-3 also extends its capabilities to time series, relational, and tabular-text data, setting new state-of-the-art results. AI

Researchers have introduced PhysEditBench, a new benchmark designed to evaluate and standardize the performance of general-purpose image editors in predicting dense physical maps. This benchmark covers five map types: depth, normal, albedo, roughness, and metallic. While specialized models still outperform image editors on depth, normal, and albedo maps, image editors show promise in matching or exceeding baseline performance for roughness and metallic maps, though they still struggle with structural errors and lighting sensitivity. AI

Researchers have developed a novel three-stage machine learning framework to address the complexities of diabetes management. The first stage benchmarks various classifiers for detecting diabetes and identifies key predictive biomarkers like glucose, BMI, and age. Subsequent stages focus on clustering diabetic patients into subtypes and exploring the link between glycemic control and cognitive function, revealing a significant positive association. AI

Researchers have developed deep neural networks to improve the resolution of land surface temperature (LST) data for urban areas. By combining data from geostationary and polar-orbiting satellites, they created LST fields with a 1 km resolution at 15-minute intervals. A U-Net model was trained to downscale SEVIRI/MSG data to MODIS resolution, achieving an RMSE of 1.92°C. Additionally, a ConvLSTM model was used for nowcasting LSTs up to 75 minutes ahead, outperforming benchmark models with RMSEs between 0.57°C and 1.15°C. AI

Researchers have explored the use of large language models (LLMs) for annotating credibility assessments in Danish asylum decisions, a novel legal NLP task. They introduced the RAB-Cred dataset, featuring expert annotations and metadata, to evaluate 21 open-weight models and various prompt combinations in zero-shot and few-shot settings. The study found that while LLMs show potential for cost-effective labeling, their annotations are imperfect and inconsistent, necessitating careful consideration beyond single model predictions. AI

A new paper investigates the effectiveness of generative AI tools in creating educational slides from instructor notes. Researchers found that coding assistants like Cursor and Claude produced the most accurate and pedagogically sound slides. When used in a real course, students perceived AI-generated slides as being of similar quality to instructor-created ones and could not reliably distinguish between them, even associating AI generation with lower quality. AI

Researchers have developed HLS-Seek, a new framework for generating hardware descriptions from natural language that prioritizes Quality of Results (QoR) like latency and resource utilization. Unlike previous methods that focused solely on functional correctness, HLS-Seek employs a proxy comparative reward model trained with reinforcement learning to achieve high accuracy in predicting optimal hardware configurations. This approach significantly speeds up training and demonstrates superior performance compared to existing frontier models on HLS-specific benchmarks, achieving lower latency and better resource utilization on several kernels. AI

Researchers have developed RealICU, a new benchmark designed to evaluate the reasoning capabilities of large language model agents in intensive care unit (ICU) settings. Unlike previous benchmarks that relied on clinician actions as ground truth, RealICU uses hindsight annotations from senior physicians reviewing complete patient histories to create more accurate labels. The benchmark includes tasks such as assessing patient status, identifying acute problems, and flagging potentially unsafe actions. Initial tests showed that current LLMs, even those with memory augmentation, performed poorly, highlighting issues with recall-safety trade-offs and anchoring bias. AI

Researchers have developed AttenA+, a new framework designed to improve robotic foundation models by addressing action inequality during training. The framework prioritizes kinematically critical segments of robot trajectories, which are often low-velocity and require precision, by reweighting the training objective based on the inverse velocity field. This physics-aware approach enhances the performance of existing Vision-Language-Action (VLA) and World-Action Models (WAM) on complex tasks and has shown significant improvements on benchmarks like Libero and RoboTwin 2.0. AI

The SynthRAD2025 challenge report details advancements in generating synthetic computed tomography (sCT) images for radiotherapy planning. This year's challenge focused on converting MRI or cone-beam CT (CBCT) into CT-equivalent images, with methods evaluated on over 2,300 patient cases across different body regions. While deep learning models showed significant improvements, particularly for CBCT-to-CT conversion, challenges remain in MRI-to-CT accuracy, especially for dose-based validation. AI

Researchers have introduced HetScene, a novel framework for generating dense indoor scenes that accounts for object heterogeneity. This approach distinguishes between primary and secondary objects to better model complex spatial arrangements and physical plausibility, which is crucial for creating realistic simulation environments for embodied AI. The framework employs a two-stage generation process, first creating structural layouts with primary objects and then refining them with contextual details. AI

Researchers have introduced OpenAaaS, an open-source framework designed to facilitate distributed materials informatics research through organized multi-agent collaboration. The framework operates on the principle of "code flows, data stays still," allowing a Master Agent to decompose tasks without accessing subordinate agents' local data or computational resources. This architecture ensures data sovereignty while enabling secure integration of isolated materials intelligence silos, demonstrated through case studies in literature analysis and alloy descriptor database services. AI

AWS and Cisco have partnered to enhance the security of AI agents and their associated protocols, Model Context Protocol (MCP) and Agent-to-Agent (A2A). This collaboration aims to address critical security gaps arising from the rapid adoption of these technologies, including lack of visibility into deployed tools, the inability of manual reviews to keep pace with deployment velocity, and the absence of audit trails for autonomous agents. The integrated solution leverages AWS's AI Registry and Cisco AI Defense to provide automated scanning, unified governance, and supply chain security for MCP servers, A2A agents, and Agent Skills, thereby mitigating risks of data breaches, compliance violations, and operational disruptions. AI

Apollo Research has expanded its operations by opening an office in San Francisco and is actively hiring for technical positions in both San Francisco and London. The company is focusing its research efforts on understanding the potential for future AI models to develop misaligned preferences and the effectiveness of training methods designed to prevent this. Additionally, Apollo is developing a product called Watcher for real-time monitoring of coding agents and is dedicating resources to AI governance, particularly concerning automated AI research and the risks of recursive self-improvement leading to loss of control. AI

A blog post details how to use NVIDIA's AIPerf tool to uncover hidden performance issues in LLM deployments. Initial tests with a local model showed excellent baseline performance, but increasing concurrency revealed a dramatic increase in time-to-first-token (TTFT), with 99% of requests failing a 500ms SLO. The analysis highlighted that the bottleneck is not the model's inter-token latency (ITL), which remained stable, but rather the request queuing and prefill phase, suggesting architectural solutions like better queue management or horizontal scaling are needed. AI

Adaption has launched AutoScientist, a tool designed to accelerate AI model training through automated fine-tuning. This system co-optimizes both data and the model itself, learning the most effective methods to acquire new capabilities. The company suggests this could enable frontier AI training outside of major research labs and has reportedly doubled win-rates across various models. AI

Google is reportedly reducing the free storage offered to new accounts from 15GB down to 5GB. This change appears to be linked to verifying a phone number, with the company citing a need to ensure storage is added only once per person. While most new account creations already require a phone number, this adjustment could be a response to increasing demand for storage. AI

The Turtle Beach Stealth Pro II is a new gaming headset offering advanced multi-device wireless connectivity. Unlike many competitors that require devices to be near a central base station, this headset allows users to connect to up to four devices wirelessly using separate dongles. This feature is particularly beneficial for gamers who have their PC and consoles in different locations. The headset also boasts a redesigned, comfortable build with active noise cancellation and a detachable boom microphone, available for pre-order at $349.99. AI

Denon has launched two new X-series AV receivers, the AVC-X3900H and AVR-X2900H, targeting home theater enthusiasts and custom integrators. These models feature advanced calibration options, including Audyssey and optional Dirac Live support, along with updated internal components and a 32-bit multi-channel DAC for high-performance audio. Both receivers integrate Denon's HEOS multiroom platform and support various connectivity options, including HDMI 2.1 features like 4K/120Hz and variable refresh rates. AI

HubSpot has released a new tool called AEO Sensor to track changes in search engine results and AI-referred traffic. This launch coincides with a reported 12-month low in ChatGPT's user traffic. Separately, Google's Ads DevCast has published four episodes focusing on AI, APIs, and scaling advertising platforms. AI

OPPO has introduced a new AI-powered image processing engine designed to optimize photos. This engine aims to provide fast, private, and high-quality results for users' images. The technology leverages generative AI for its light processing capabilities. AI

The PriEco crawler has integrated a new AI blocklist to filter out content deemed "AI slop." This move aims to improve search result quality by excluding websites primarily generated by AI. The project is actively seeking further methods to combat AI-generated content. AI

The Fedora Project's initiative to create an AI-focused developer desktop environment has been blocked. Two members of the Fedora Council reversed their initial approvals following community feedback that indicated a lack of support for the proposal. This reversal means the initiative will not proceed at this time, pending further understanding and community engagement. AI

AI-native workflows are revolutionizing data science repositories, with tools like CodeSpeak automating code management on platforms like GitHub. This shift is transforming how developers interact with and manage code, particularly for large-scale projects. The insights suggest a significant evolution in data science practices by 2026. AI

A new wave of AI video generation tools is emerging, promising to significantly alter content creation. These platforms are enabling users to generate videos from text prompts, with some offering advanced features and free prompt access. The rapid development in this area suggests a major shift in how video content will be produced in the near future. AI

Google's Gemini AI has been instrumental in blocking billions of malicious advertisements, according to a report. While this represents a significant step in combating online threats, the effectiveness is deemed insufficient to fully address the pervasive issue of malvertising. AI

A developer has created a PHP-based command-line tool to improve how AI assistants like Claude interact with large codebases. The tool builds a dependency graph of a project using SQLite, allowing the AI to access specific code elements like methods or class fields without reading entire files. This approach aims to conserve the AI's context window, which is crucial for efficiency when working with extensive code projects. AI

OpenSquilla, a new open-source platform, aims to significantly reduce AI token costs by up to 80%. It utilizes a novel memory architecture inspired by the human brain and intelligent query routing. This approach is designed to transform how AI autonomous agents are deployed within businesses. AI

VASO, an open-source security agent scanner, has been released to the public. Developed by Vulnex, this tool is designed to detect and scan various AI agents within an organization's infrastructure. Its capabilities extend to identifying agents like OpenClaw, Nemoclaw, Claudecode, and Codex, among others, enhancing cybersecurity measures for AI systems. AI

Founder Motor has successfully initiated small-batch production of its robot joint motors. This move is part of the company's strategy to increase its manufacturing capacity in response to rising market demand for these specialized components. AI

A new tutorial details how to construct a dynamic Zero-Trust network simulation using graph-based micro-segmentation and an adaptive policy engine. The method models network segments as a directed graph, requiring continuous verification for all requests. This system incorporates device posture, multi-factor authentication, and real-time risk signals to detect and block malicious activities, including insider threats, in real time. AI

This article focuses on Day 3 of the 100 Days of MLOps challenge, specifically addressing the critical issue of dependency reproducibility. It highlights the use of 'uv', a tool designed to manage and ensure that project dependencies remain consistent across different environments and over time. The goal is to provide developers and data scientists with reliable methods for maintaining stable MLOps workflows. AI

This article provides a detailed guide on creating an AI-powered "second brain" using Obsidian and Claude Code. It outlines the setup process for users to enhance their note-taking capabilities. The guide aims to transform regular note-takers into more efficient power users by leveraging AI tools. AI

Meta Platforms is introducing a "stealth chat" feature to its WhatsApp AI assistant, designed to address user privacy concerns by ensuring conversations are not stored and messages disappear automatically. This move utilizes private processing technology to keep dialogues invisible to all parties, including Meta itself. The company aims to provide a secure space for users to share ideas without surveillance. AI

Anthropic has updated its Claude subscription model, impacting users of the `claude -p` command-line interface. Discussions on platforms like Reddit indicate that users are seeking ways to adapt to these changes, particularly concerning the CLI's functionality. AI

A user explored Anthropic's new Claude Code /goal command, which they found transformed into a self-driving coding agent. This feature appears to be a significant advancement, potentially rendering previous 'Keep Going' functionalities obsolete. AI

Microsoft Research has developed GridSFM, a compact foundation model designed to predict optimal power flow in electric grids with high speed and accuracy. This model can approximate complex AC optimal power flow calculations in milliseconds, a task that previously took hours. By enabling faster analysis, GridSFM aims to reduce significant annual losses from congestion and renewable energy curtailment, while also improving grid reliability and stability. AI