arXiv:2609.30279v1 Announce Type: new Abstract: Understanding the features captured by the hidden layers of neural networks is a fundamental challenge in machine learning, despite their widespread success across various classification problems. In this work, we propose an algebraic framework for examining neural networks that model classification problems. Certain results, such as the correspondence between the neural network and neural ideals, algorithms for computing the neural ideals, and a stabilization theorem that enables approximation of the neural ideals, are first established. As an application to the framework, we present algorithms to identify and interpret the features captured by each hidden-layer neuron. Along with these theoretical developments, the practical performance ha…

arXiv:2609.30277v1 Announce Type: new Abstract: Implicit Graph Neural Networks (IGNNs) define node representations as fixed points of message-passing operators, enabling effectively infinite-depth propagation, iteration-independent parameterization, and flexible test-time computation. Yet these benefits depend on the equilibrium being unique and attainable by fixed-point iteration. Existing constructions often impose constraints on recurrent updates to obtain these guarantees, limiting the transformations available at equilibrium. This raises a central question: can IGNNs gain expressiveness through richer, edge-dependent transformations while retaining the inherent strengths of their equilibrium formulation? We introduce SheafDEQ, a subhomogeneous deep-equilibrium architecture with adapt…

arXiv:2609.30276v1 Announce Type: new Abstract: We analyze the original same-step coordinate-wise AdaGrad under generalized smoothness and heavy-tailed noise with bounded variance. In this setting, local curvature may grow sub-quadratically with the gradient norm, and stochastic gradients are assumed to have only bounded conditional second moments. We show that unclipped AdaGrad can become \emph{anisotropically miscalibrated}: under heavy-tailed noise, the adaptive denominator can learn the geometry of rare noise shocks rather than the local curvature of the objective, leading to a persistent directional distortion that blocks finite-horizon Euclidean progress. We then prove that clipping repairs this failure mode. Our main result is a finite-horizon high-probability guarantee for the ori…

arXiv:2609.30275v1 Announce Type: new Abstract: A statistic reported without the quantity needed to interpret it is not evidence. We develop that thesis for a concrete practice in AI safety. Interpretability artifacts are calibrated on full-precision weights, deployed on quantized ones, and certified as surviving the change by scale-invariant statistics (cosine similarity, correlation, AUROC) that are reported without their noise floor. For the difference-in-means direction estimator, the split-half floor is governed by one dimensionless number, $\kappa = n\rho^2/d$. The closed form $\mathbb{E}[\cos] \approx (1+4/\kappa)^{-1}$ is classical; the missing input is the class separation $\rho$, which we measure on real activations; no compression-transfer study we know of reports it. On Qwen2.…

arXiv:2609.30273v1 Announce Type: new Abstract: We investigate how historical data from fixed randomized experiments (A/B tests) can be used to inform the deployment of adaptive experiments based on contextual bandits. Given data collected under a static allocation, our goal is to assess which adaptive policies, if any, would have outperformed the original design and under what conditions. To this end, we combine off-policy evaluation (OPE) with a controlled warm-start simulation. From logged A/B test data exhibiting heterogeneous treatment effects, we estimate nuisance components and use doubly robust estimators to rank a portfolio of pre-specified adaptive and non-adaptive policies. When ground truth is available, we then deploy the same offline-trained policies in a simulator that reus…

arXiv:2609.30272v1 Announce Type: new Abstract: We present \textbf{ENAS}, a hardware-aware Neural Architecture Search (NAS) framework that combines a static feasibility check, a cell-based search space supporting standard, depthwise-separable, and bottleneck blocks with optional skip connections, and a three-stage hybrid search strategy (random $\rightarrow$ top-$K$ $\rightarrow$ mutation) with persistent cross-run caching. Unlike many existing NAS frameworks that rely on GPU acceleration, ENAS is designed to operate efficiently without requiring GPUs, making it suitable for resource-constrained development environments. We evaluate ENAS on two TinyML benchmarks, Visual Wake Words and Melanoma Cancer, across eight microcontrollers with memory footprints ranging from 20\,KB to 1\,MB SRAM a…

arXiv:2609.30271v1 Announce Type: new Abstract: Adaptive optimizers are commonly parameterized by a fixed power of the second-moment estimate. Existing partially adaptive methods study exponents between momentum-like updates and the standard Adam square root, while the interaction between this exponent and the global learning rate is less understood. We perform a controlled cross-environment study using a paired four-environment classification problem with stable sparse features, environment-dependent spurious sparse features, dense features, and high-dimensional noise. Across \NumRuns{} source-training runs covering 21 preconditioning exponents $p\in[-0.5,0.5]$ and five learning rates $\eta\in[10^{-4},10^{-2}]$, we find that the exponent maximizing cross-environment accuracy decreases al…

arXiv:2609.30270v1 Announce Type: new Abstract: On-device inference with small language models keeps user data local, works offline, and incurs no per-query cost, so the on-device tier is preferred when it is adequate. It is thermally constrained, however, and I find the constraint is sharper than a slowdown: on a flagship Snapdragon device, sustained on-device generation destabilizes the GPU inference runtime, which crashes or silently wedges after a few consecutive queries. The failure lies in the current toolchain (OpenCL kernel compilation and long-prompt prefill on the mobile GPU), recurs even when the device is cool, and is worst for long generations. Multi-tier routers across on-device, edge, and cloud models can relieve this pressure, but existing routers are thermal-blind and typ…

Professor Sherry Turkle’s new book, “Artificial Intimacy,” offers a withering critique of chatbots and the antisocial dynamics she believes they encourage.

In a study focusing on hiring decisions, MIT researchers found the use of a single algorithm by many firms could benefit job seekers in certain situations.

Congrats team!

a rare feature of a capability

Artificial intelligence firm to front parliament as it apologises to Australians for agent attack OpenAI has apologised to Australians for its agent attack on Medicare, and will front parliament next week, as the tech company revealed more details about its June hack of Australian government websites. In a blog post released on Tuesday, OpenAI said it should have handled its response better. Continue reading...

Discussion | Link

Nvidia’s new system is designed to contain rogue AI agents as companies race to secure increasingly autonomous systems. But it has not been proven yet.

Shipping Opus/Sonnet 5.5, Mods, Plugins, Projects, Tag while Pacing the Frontier

OpenAI apologizes for incidents involving Australian government websites and outlines stronger safeguards and support to strengthen Australia’s cyber defences.

Ollama now supports decision models, based on TypeSafe's Jev API for fast, typed decisions. Decision models can now be run at no cost with low latency. Based on text, decision models answer yes-or-no questions, choices and scores about it, with a probability for every option.

Dots by OpenAI are a proactive assistant that can keep working across complex projects and everyday tasks. Learn how dots help you stay in control while work moves forward.

When language models reason in chain-of-thought or exchange free-text intermediates, they serialize structured information into natural language. How much tree-structured compositional content survives this bottleneck? We propose a round-trip protocol that answers this question empirically for tree-structured expressions. A generator converts a procedurally generated arithmetic expression into a word problem, a separate extractor recovers the expression from the word problem alone, and symbolic equivalence provides an exact oracle. Evaluating all pairwise combinations of sixteen models yields…

The new AI agent, Muse, was released last week and has been downloaded by 3 million users When Usman asked if a keyboard for sale on Facebook Marketplace was still available on Saturday, he received an enthusiastic and immediate response. “Yep still available!” wrote Matt Robb. Continue reading...

GPT-6.1 Astra showed deceptive behavior and tried to use external tools despite knowing it would be unsafe OpenAI is scrapping the release of GPT-6.1 Astra, a next-generation ⁠AI model planned for an October debut, over safety concerns raised by researchers ⁠during internal testing, the ⁠Wall ​Street Journal reported on Monday. The model, expected to appear in ChatGPT and ⁠Codex, was designed to handle more complex tasks without human assistance, the report said. Continue reading...

xAI's Grok 4.7 is now available on Amazon Bedrock: a frontier model for coding, long-running agents, and knowledge work. It offers a 500K token context window and four configurable reasoning effort levels, reachable through the Responses, Chat Completions, and Converse APIs.

Discussion | Link

Claude Sonnet 5.5 New Sonnet model from Anthropic today. They say it "runs 30%+ faster, and costs up to 30% less for most work" - it's priced the same as Sonnet 5 but appears to beat it on every benchmark, and should be cheaper to run as well. Here are some pelicans riding bicycles. Sonnet 5.5 suffered from the same bug as Opus 5.5: the "max" thinking effort pelican thought for 128,000 tokens (at a cost of $1.28) before running out of tokens and failing to produce an SVG. Here's the pelican it gave me for thinking effort "xhigh", at a cost of 5.74 cents and taking 41 seconds: Sonnet 5.5 appears to be almost as good as Opus 5.5 on some coding tasks, including various viral 3D animation tricks. The most interesting thing about Sonnet 5.5 is that it's now the model used for the free tier on…

Discussion | Link

Building useful robots starts with understanding the people who use them. For Charlie Kemp, cofounder and chief technology officer of Hello Robot, that means developing assistive robots that can help people with everyday tasks and support greater independence. In this Robots Guide profile, Kemp shares his path from studying artificial intelligence at MIT to building Stretch, explains how working with people with disabilities has shaped his approach, and offers advice for aspiring roboticists. Read the full profile on IEEE’s Robots Guide.

World Labs co-founder and CEO Fei-Fei Li. | Bloomberg via Getty Images AMD announced today that it's acquiring World Labs, an AI research lab co-founded by the prominent researcher Dr. Fei-Fei Li, in an all-stock deal worth approximately $8.2 billion. World Labs launched in 2024 and was valued at $1 billion in a matter of months. The startup launched its first commercial product, a world generation model called Marble that lets users make interactive 3D worlds from a prompt, in 2025. The deal is expected to close by the end of the year. Li will become EVP and chief scientist at AMD, reporting to AMD CEO Lisa Su, while the World Labs team will "continue to focus on advancing AI model research." AMD also notes t … Read the full story at The Verge.

Since launching a year ago, the Microsoft Research Asia — Singapore lab has established a strong foundation, deepened collaboration across government, academia, and industry, and explored how frontier AI research can create real-world value. The post One year in: How Microsoft Research Asia – Singapore is advancing research, partnership and talent for real-world impact appeared first on Microsoft Research.

Discussion | Link