Dynamical Systems Analysis Reveals Functional Regimes in Large Language Models
arXiv:2601.11622v1 Announce Type: new
Abstract: Large language models perform text generation through high-dimensional internal dynamics, yet the temporal organisation of these dynamics remains poorly understood. Most interpretability approaches emphasise static representations or causal interventi...
Reasoning Stabilization Point: A Training-Time Signal for Stable Evidence and Shortcut Reliance
arXiv:2601.11625v1 Announce Type: new
Abstract: Fine-tuning pretrained language models can improve task performance while subtly altering the evidence a model relies on. We propose a training-time interpretability view that tracks token-level attributions across finetuning epochs. We define explana...
Liquid AI Releases LFM2.5-1.2B-Thinking: a 1.2B Parameter Reasoning Model That Fits Under 1 GB On-Device
Liquid AI has released LFM2.5-1.2B-Thinking, a 1.2 billion parameter reasoning model that runs fully on device and fits in about 900 MB on a modern phone. What needed a data center 2 years ago can now run offline on consumer hardware, with a focus on structured reasoning traces, tool use, and math, ...
Knowledge Graphs and their limitations With the rapid growth of AI applications, Knowledge Graphs (KGs) have emerged as a foundational structure for representing knowledge in a machine-readable form. They organize information as triples—a head entity, a relation, and a tail entity—forming a graph-li...
Anthropic CEO Dario Amodei unloaded on both the administration and U.S. chip companies over plans to sell to China. The criticism was particularly notable because one of those chipmakers, Nvidia, is a major partner and investor in Anthropic.
Stanford researchers have developed a deep learning model that transforms overwhelming brain data into clear trajectories, opening new possibilities for understanding thought, emotion, and neurological disease.
DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generation
Diffusion large language models (dLLMs) are compelling alternatives to autoregressive (AR) models because their denoising models operate over the entire sequence. The global planning and iterative refinement features of dLLMs are particularly useful for code generation. However, current training and...
Elon Musk says Tesla’s restarted Dojo3 will be for ‘space-based AI compute’
Tesla aims to restart work on Dojo3, its previously abandoned third-generation AI chip. Only this time, Dojo3 won’t be aimed at training self-driving models on Earth. Instead, Musk says it will be dedicated to “space-based AI compute.”
OpenAI and the Gates Foundation launch Horizon 1000, a $50M pilot advancing AI capabilities for healthcare in Africa. The initiative aims to reach 1,000 clinics by 2028.
Building a community-led future for AI in film with Sundance Institute
Illustration of a cinematographer filming with a professional video camera mounted on a tripod, positioned next to a studio light and a film clapboard.
Zhipu AI Releases GLM-4.7-Flash: A 30B-A3B MoE Model for Efficient Local Coding and Agents
GLM-4.7-Flash is a new member of the GLM 4.7 family and targets developers who want strong coding and reasoning performance in a model that is practical to run locally. Zhipu AI (Z.ai) describes GLM-4.7-Flash as a 30B-A3B MoE model and presents it as the strongest model in the 30B class, designed fo...
From Traces to Insights: Understanding Agent Behavior at Scale
Visibility is the easiest piece. The hard part is analyzing and understanding what you’re observing. I’ve spoken to teams recording 100k+ traces every single day. What are they doing with those traces? Literally nothing. Because it’s impossible to read and summarize 100,000 traces
Multimodal reinforcement learning with agentic verifier for AI agents
Argos improves multimodal RL by evaluating whether an agent’s reasoning aligns with what it observes over time. The approach reduces visual hallucinations and produces more reliable, data-efficient agents for real-world applications.
The post Multimodal reinforcement learning with agentic verifier f...
Semarchy Closes 2025 with Global Growth and Momentum
Semarchy today announced a year of major global growth driven by continued customer expansion, new strategic partnerships, and the successful launch of both the Semarchy Data Platform (SDP) and the first and only MDM native app for Snowflake. The company also continued to build out and strengthen it...
Ellipsis Health expands access to Sage via Salesforce AppExchange
Agentforce Health users can now automate patient assessments, risk evaluations, and follow-up using Sage, an autonomous, empathetic AI care manager Ellipsis Health today announced it has launched Sage on Salesforce AppExchange, the leading enterprise marketplace for partner apps and experts. This in...
Product brand becomes company name as Kubernetes, AI-first platform gains market momentum Densify, the pioneer in automated resource optimization, rebrands as Kubex, reflecting the company’s strategic evolution into the leading platform for end-to-end Kubernetes, GPU and AI resource optimization. W...
Unframe Launches Unframe Unlimited to Accelerate Enterprise ROI on AI
New partner program will empower organizations to deliver tailored, production-ready AI solutions and unlock new revenue opportunities Today, Unframe, the managed AI delivery platform for global enterprises, announced the launch of the Unframe Unlimited, a new program designed to help channel partne...
Evernote Releases v11, Marking a New Chapter in AI-Powered Productivity
Evernote today announced the release of v11, its first major product update in five years. The new version includes three powerful AI features—AI Assistant, Semantic Search, and AI Meeting Notes—creating smarter ways to capture, retrieve, and manage information in Evernote. With the release of v11, ...
The era of agentic chaos and how data will save us
AI agents are moving beyond coding assistants and customer service chatbots into the operational core of the enterprise. The ROI is promising, but autonomy without alignment is a recipe for chaos. Business leaders need to lay the essential foundations now. The agent explosion is coming Agents are in...
Unbreakable? Researchers warn quantum computers have serious security flaws
Quantum computers could revolutionize everything from drug discovery to business analytics—but their incredible power also makes them surprisingly vulnerable. New research from Penn State warns that today’s quantum machines are not just futuristic tools, but potential gold mines for hackers. The stu...
The UK government is backing AI scientists that can run their own experiments
A number of startups and universities that are building AI scientists to design and run experiments in the lab, including robot biologists and chemists, have just won extra funding from the UK government agency that funds moonshot R&D. The competition, set up by ARIA (Advanced Research and Invention...
Welcome
2026 the year of Action?
2026 marks the end of the "Hype Phase" and the start of the "Physical Phase." The narrative has shifted aggressively from what AI can say (Generative) to what AI can do (Agentic). We are witnessing a massive capital pivot toward the "Physical L...