Silicon Valley companies are already working on neurotechnology products that track your brain activity. The next privacy frontier might be the things you only think. Rafael Yuste is in his early sixties and bears a more than passing resemblance to Pablo Picasso--if Picasso had worn glasses and had a trim white goatee. Speaking succinctly and methodically, his accent rich with Spanish inflections, he told me about an experiment he had carried out in his lab at Columbia on the brains of mice, and specifically on that part of the cortex that responds to vision. His mentor had been the Swedish neuroscientist Torsten Wiesel, who won a Nobel Prize for his research into how our visual systems process information. "He discovered by chance that the strongest stimulus is a pattern of high-contrast dark and light bars." He held up one hand and waved his fingers back and forth. "If you imagine my fingers were bars of light surrounded by complete blackness--if I move my fingers in front of your eyes, that fires up your whole visual cortex."

Mini human brains are being grown in labs all over the world. I'm going to let you in on a secret. Every cell in your body has the potential to get smarter. I don't mean this metaphorically, or in a "body keeps the score" kind of way. I mean that if lab-coated biologists took a sample of your skin and very carefully manipulated the cells inside it, they could actually make a brain. They do it all the time. Not a brain as complex as the one behind your eyes, of course, but a glob of gray matter nonetheless, with a few-million-odd neurons that can send and receive electrical signals. Kept at a womblike 98.6 degrees Fahrenheit for eight months, they'll produce repetitive oscillations-- --nearly indistinguishable from those made by a premature baby. In cell culture labs around the world, human brain organoids live out their short lives as neural guinea pigs, testing the effects of diseases, toxins, and new pharmaceuticals. But they may soon be on to more glamorous pursuits. At the University of San Diego, organoids are guiding spidery robots through mazes and taking hero doses of psychedelics.

When a fish holds its position against a current in a river, its brain must figure out how fast to swim and how to steer to offset the water flow. Most fish use vision to register the world sliding past, detect optic flow speed and direction, and their brains turn these signals into compensatory swimming. The retina captures signals of optic-flow direction, central pretectal neurons interpret direction, and spinal nerves drive muscle contractions. The catch is that one cannot easily change the living brain to test how these circuits work. Although advances in imaging now allow detailed recording, and even manipulation, of neurons alongside behavior, rewiring connections to ask what a particular link actually does remains almost impossible in the living, complex animal.

The AI firm Anthropic has developed a technique that has given it the clearest glimpse yet at what's really going on inside large language models as they answer questions or carry out tasks. What they found ranges from the mundane to the unnerving. Researchers at the company built a tool called the Jacobian lens (or J-lens) and used it to uncover a hidden area, which they named the J-space, inside Claude Opus 4.6, a version of Anthropic's flagship LLM released in February.

Neural networks are known to be susceptible to over-reliance on spurious correlations. However, the precise mechanism by which models exploit shortcut features is not fully understood, and algorithms to mitigate this behavior rely on as yet unjustified assumptions about the learned representations. In this work, we provide the first end-to-end theoretical characterization of spurious feature learning for two-layer ReLU neural networks trained by online minibatch SGD on the logistic loss. We consider data drawn from the high-dimensional Boolean hypercube with a quadratic signal function (namely XOR) and a linear spurious correlation. We show that SGD learns the spurious feature first, and exponentially fast. Moreover, the optimization dynamics couple the spurious and signal features, with a stronger spurious component inhibiting signal feature learning. Our analysis reveals precise phase transitions in the learning dynamics. In the first phase, alignment between the signs of the spurious feature and second-layer weight drives rapid growth of the spurious feature. In the second phase, large majority group margin slows learning and the signal feature remains suppressed. When the spurious correlation is maximally strong, we show theoretically that the spurious feature dominates even at the sample complexity threshold where XOR would be learned in isolation (i.e., if the spurious feature was absent). In contrast, when the correlation strength is constant, we provide preliminary empirical evidence that the model can eventually learn the XOR signal, although the spurious feature is not forgotten.

How some people's brains make an extraordinary recovery from stroke A well-known actor who had experienced a stroke was treated by stroke specialist Sandor Nardai. The actor had been left with aphasia, or an impaired ability to speak - brutal for anyone, but "probably the most devastating thing that could happen to an actor", says Nardai. After three months of recovery, though, the actor was able to say some words. After a year, he voiced a commercial. Remarkably, he eventually got well enough to return to live theatre, says Nardai, who is at Semmelweis University in Hungary.

We investigate the robustness of Large Language Models (LLMs) to structural interventions by deleting and swapping adjacent layers during inference. Surprisingly, models retain 72-95% of their original top-1 prediction accuracy without any fine-tuning. We find that performance degradation is not uniform across layers: interventions to the early and final layers cause the most degradation, while the model is remarkably robust to dropping middle layers. This pattern of localized sensitivity motivates our hypothesis of four stages of inference, observed across diverse model families and sizes: (1) detokenization, where local context is integrated to lift raw token embeddings into higher-level representations; (2) feature engineering, where task-and entity-specific features are iteratively refined; (3) prediction ensembling, where hidden states are aggregated into plausible next-token predictions; and (4) residual calibration, where irrelevant features are suppressed to finalize the top-1 output distribution. Synthesizing behavioral and mechanistic evidence, we provide a hypothesis for interpreting depth-dependent computations in LLMs.

Sparse Autoencoders (SAEs) have recently gained attention as a means to improve the interpretability and steerability of Large Language Models (LLMs), both of which are essential for AI safety. In this work, we extend the application of SAEs to Vision-Language Models (VLMs), such as CLIP, and introduce a comprehensive framework for evaluating monosemanticity at the neuron-level in visual representations. To ensure that our evaluation aligns with human perception, we propose a benchmark derived from a large-scale user study. Our experimental results reveal that SAEs trained on VLMs significantly enhance the monosemanticity of individual neurons, with sparsity and wide latents being the most influential factors. Further, we demonstrate that applying SAE interventions on CLIP's vision encoder directly steers multimodal LLM outputs (e.g., LLaVA), without any modifications to the underlying language model. These findings emphasize the practicality and efficacy of SAEs as an unsupervised tool for enhancing both interpretability and control of VLMs.

State-of-the-art neural network (NN) verifiers demonstrate that applying the branchand-bound (BaB) procedure with fast bounding techniques plays a key role in tackling many challenging verification properties. In this work, we introduce the linear constraint-driven clipping framework, a class of scalable and efficient methods designed to enhance the efficacy of NN verifiers. Under this framework, we develop two novel algorithms that efficiently utilize linear constraints to 1) reduce portions of the input space that are either verified or irrelevant to a subproblem in the context of branch-and-bound, and 2) directly improve intermediate bounds throughout the network. The process novelly leverages linear constraints that often arise from bound propagation methods and is general enough to also incorporate constraints from other sources. It efficiently handles linear constraints using a specialized GPU procedure that can scale to large neural networks without the use of expensive external solvers. Our verification procedure, Clip-and-Verify, consistently tightens bounds across multiple benchmarks and can significantly reduce the number of subproblems handled during BaB. We show that our clipping algorithms can be integrated with BaB-based verifiers such as α,β-CROWN, utilizing either the split constraints in activation-space BaB or the output constraints that denote the unverified input space. We demonstrate the effectiveness of our procedure on a broad range of benchmarks where, in some instances, we witness a 96% reduction in the number of subproblems during branch-and-bound, and also achieve state-of-the-art verified accuracy across multiple benchmarks. Clip-and-Verify is part of the α,β-CROWNverifier, the VNN-COMP 2025 winner.

Gradient-based algorithms are a cornerstone of artificial neural network training, yet it remains unclear whether biological neural networks use similar gradientbased strategies during learning. Experiments often discover a diversity of synaptic plasticity rules, but whether these amount to an approximation to gradient descent is unclear. Here we investigate a previously overlooked possibility: that learning dynamics may include fundamentally non-gradient "curl"-like components while still being able to effectively optimize a loss function. Curl terms naturally emerge in networks with inhibitory-excitatory connectivity or Hebbian/anti-Hebbian plasticity, resulting in learning dynamics that cannot be framed as gradient descent on any objective. To investigate the impact of these curl terms, we analyze feedforward networks within an analytically tractable student-teacher framework, systematically introducing non-gradient dynamics through neurons exhibiting rule-flipped plasticity.