Deep Learning Weekly
Editorial team
20 stories we have summarized that Deep Learning Weekly covered.
DeepMind publishes principles for coordinating multiple AI agents
DeepMind's research identifies four rules for managing systems where multiple AI agents work together on a single task. The principles include breaking work into clear contracts between agents, choosing cheaper models when possible, limiting data access, and adding friction to prevent blind obedience.
Deep Learning Weekly
EVE Online chosen as AI research testbed for continual learning
Fenris Creations partnered with EVE Online, a massively multiplayer game, to study how AI agents learn and adapt over time in complex environments. The research builds on 15 years of prior AI work and focuses on three specific challenges: agents that improve continuously, agents that remember long sequences of events, and agents that interact with many other agents simultaneously.
Deep Learning Weekly
Z.ai releases open-source GLM-5.3 coding model and cheaper variant
Z.ai open-sourced GLM-5.3, a model trained to find bugs in code. It identified 2,436 bugs across 269 real open-source projects. Z.ai also released GLM-5.3-Flash, a smaller version of the same model designed to cost roughly one-tenth as much to run.
The NeuronDeep Learning Weekly
Simon Willison+1
Stability AI raises $76 million from music labels and gaming firms
Stability AI, which makes image and video generation software, completed a $76 million funding round led by its existing content-licensing partners. Music labels and gaming companies invested in the round, converting their existing business relationships with Stability AI into ownership stakes.
Deep Learning Weekly
Study questions whether AI explanation tools actually work
Researchers tested CHIVE, a pipeline designed to explain large language model decisions by reading internal activations, the mathematical patterns flowing through the system. When researchers deliberately changed prompts to test if the explanations were accurate, the activation-reading approach performed no better than simply reading the model's text output.
Deep Learning Weekly
Perplexity and NVIDIA release local AI computing software
Perplexity, an AI search company, launched Portable Computer software that runs AI models directly on personal computers instead of cloud servers. The software works on NVIDIA's DGX Spark hardware and RTX graphics cards, eliminating per-query costs associated with cloud AI services.
Prompt Engineering DailyDeep Learning Weekly
New benchmark reveals AI models struggle with complete scientific tasks
FrontierChallenge tested AI models on 97 scientific workflows spanning six different fields, measuring whether models could finish entire tasks end-to-end. The best-performing models succeeded on only 20.6% of these complete workflows, despite often showing high scores on individual parts of tasks.
Deep Learning Weekly
Keenable launches search engine designed for AI agents
Keenable, a new search startup, raised $26 million in seed funding to build its initial product. The company is indexing 100 billion documents to create a search system optimized for how AI agents query information, rather than how humans do.
Deep Learning Weekly
Google XR prototype now generates hand gestures for agents
Google built AgentHands, a system that makes AI agents in virtual or augmented reality move their hands while talking, synchronized with their speech. The hand gestures help agents communicate spatial information more clearly, like pointing to objects in the environment around them.
Deep Learning WeeklyZetta system lets robots learn and improve while working
Zetta is a framework that lets robots update their own decision-making code while physically operating, rather than needing to stop and retrain. The system achieved 90.8% and 93.6% success rates on two standard robot benchmark tasks, with 11.1x faster inference speed than baseline methods.
Deep Learning WeeklyStripe acquires OpenRouter for reported $7.5 billion
Stripe, a payments processor, is buying OpenRouter, a platform that connects to over 400 AI models from 80+ different providers. OpenRouter acts as a router, meaning it lets developers access many AI models through one interface rather than managing each separately.
Deep Learning WeeklyResearchers propose system for AI to build 3D worlds from text
VibeWorlding is a framework that lets AI agents autonomously create interactive 3D environments based on what users ask for. Testing showed frontier models, the most advanced AI systems available, succeeded less than 60% of the time at this task.
Deep Learning WeeklyResearchers extract hidden data from encrypted AI model reasoning
Security researchers developed an attack that recovers encrypted reasoning traces, the internal thinking logs that AI models generate while processing requests. The attack works by replaying encrypted reasoning data across different sessions and models to expose what was previously hidden.
Deep Learning WeeklyRedwood and Anthropic release reasoning benchmark for unverifiable questions
Conceptual Reasoning Index combines three benchmarks testing how AI models argue about questions without definitive answers. Anthropic's Claude Opus 5 model scored 73.6 on the index, with researchers estimating a theoretical maximum around 91.
Deep Learning WeeklyOpenAI expands privacy options for API customers
OpenAI confirmed Zero Data Retention, a feature letting API customers prevent their data from being stored or used for model training. The company previewed Private Safety Processing, a new system that checks API requests for safety issues without retaining the data afterward.
Deep Learning WeeklyIBM releases Granite 4.2 language models in three sizes
IBM released three Granite 4.2 models with 3 billion, 8 billion, and 30 billion parameters, trained on 15 trillion tokens and supporting up to 512,000 token context windows. The 8B and 30B variants learn to use tools, write code, and search the web by training in real sandbox environments rather than on static instructions.
TLDR AIDeep Learning WeeklyGoogle finds LLMs forget facts they actually know
Google Research developed a framework to profile how language models store knowledge and discovered most factual errors come from recall failures, not from models failing to learn facts. The distinction matters because it means frontier models like GPT-4 and Claude likely contain the information needed to answer questions correctly but cannot retrieve it reliably.
Deep Learning WeeklyFreeToken enables massive models on single consumer GPU
Frontier published FreeToken, a system that lets very large AI models run efficiently on individual workstations by adjusting how computation spreads across available hardware. The system successfully deployed models with 753 billion parameters (a measure of model size) on a single GPU, which would normally require multiple expensive specialized processors.
Deep Learning WeeklyOpenAI halts major AI training experiment over security risks
OpenAI stopped its largest reinforcement learning experiment, a training method where AI systems learn by trial and error, due to cybersecurity concerns. The company found early signs that its upcoming Astra model might reach a point where it poses security risks, though specifics were not detailed.
Deep Learning WeeklyMIT researchers find AI images lack traceable sources
MIT researchers tested whether removing single training images changes what large image-generating models produce. Outputs remained largely unchanged, suggesting many generated images cannot be linked to specific training data. The researchers call this problem attribution decay. It means the AI models have absorbed patterns so broadly that individual training images become unidentifiable in the final outputs.
Deep Learning Weekly