AI News Today

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. For those of you enjoying your summer unaware of Anthropic’s latest feud with the US government, here’s a recap: In April the company said it had built an AI model called Mythos…

<table> <tr><td> <a href="https://www.reddit.com/r/MachineLearning/comments/1ucm508/some_new_updates_to_papers_with_code_p/"> <img src="https://preview.redd.it/wawma8paeu8h1.png?width=140&height=39&auto=webp&s=8546e40b7710c5a3566b17f2c4635b14b3f19c0b" alt="Some new updates to Papers with Code [P]" title="Some new updates to Papers with Code [P]" /> </a> </td><td> <!-- SC_OFF --><div class="md"><p>Hi folks, </p> <p>Niels here from the open-source team at Hugging Face. I continue working on a revi

JUPITER, Europe’s first exascale supercomputer at Germany’s Forschungszentrum Jülich, runs on NVIDIA Grace Hopper Superchips and NVIDIA Quantum-X800 InfiniBand networking — and it’s had a busy year. As the international supercomputing community gathers at ISC in Hamburg this week, four projects running on JUPITER point to what exascale computing can actually do: map the human […]

For the past two years, the U.S. National Science Foundation’s National Artificial Intelligence Research Resource (NAIRR) pilot program has driven innovative research across the U.S. for over 700 projects — spanning protein prediction and infectious disease outbreak management. NVIDIA contributed to the NAIRR pilot through a cloud-based resource that gives researchers dedicated access to a […]

At the ISC conference running in Hamburg this week, NVIDIA is introducing new software that speeds AI for science, from chemistry and materials discovery to the search for dark matter. The NVIDIA DAQIRI library and new NVIDIA ALCHEMI NIM microservices — as well as the NVIDIA cuPhoton reference code, coming soon — turn work that […]

Mission, Vision and Veritas — new Los Alamos National Laboratory (LANL) supercomputers to be built with HPE and NVIDIA — are tapping NVIDIA Vera CPUs to accelerate scientific discovery, unlocking agentic AI for science. The supercomputers will use the HPE Cray Supercomputing GX5000 architecture with the NVIDIA Vera Rubin platform, combining NVIDIA Vera CPUs, NVIDIA […]

The next era of AI will not be defined by compute alone. Its growth will be determined by energy. As accelerated computing scales across AI factories, agentic AI, industrial AI, edge computing and physical AI — including robotics and autonomous systems — global electricity demand is rising at unprecedented speed. In many regions, expanding grid […]

<!-- SC_OFF --><div class="md"><p>Anyone building at the model layer knows the procurement problem hasn't gone away. H100s are still allocated unevenly, spot instances get preempted at the worst times, and pricing across providers is deliberately hard to compare. Most teams end up over-provisioning just to feel safe.</p> <p>The traditional fixes — reserved instances, spot bidding, broker marketplaces like CoreWeave or <a href="http://Vast.ai">Vast.ai</a> — all have the same problem: no real pric

The HIP kernel uses one-instruction asm wrappers and an eight-wave pipeline to outperform AMD's AITER v3 on MI300X. The post MoonMath AI Open-Sources a HIP Attention Kernel for AMD MI300X That Beats AITER v3 on Every Shape and Rounding Mode appeared first on MarkTechPost .

Hot tubs sit at about 38 to 40 degrees Celsius, warm enough that most people can only soak for about 15 minutes. NVIDIA’s newest AI servers can run their cooling liquid even hotter — up to 45 degrees Celsius, or 113 degrees Fahrenheit. That higher temperature limit is precisely what makes them more energy efficient. […]

LLMs are stateless by default. Agent memory fixes that. This guide breaks down all 7 types — working, semantic, episodic, procedural, retrieval, parametric, and prospective. It covers what each stores, where it lives, and when to build it. Includes a comparison table and working Python code. The post The 7 Types of Agent Memory: A Technical Guide for AI Engineers appeared first on MarkTechPost .

<!-- SC_OFF --><div class="md"><p>We just did a big revamp of <strong>WeightsLab</strong> and wanted to share it here.<br/> If you’ve ever spent hours debugging a training run only to discover it was a data problem all along, this is for you.<br/> WeightsLab lets you pause training mid-run, inspect your live loss signals, and catch mislabels, class imbalance & outliers before they tank your model.</p> <p>Open source, PyTorch-native, built for CV engineers working with images, videos & LiDAR poin

<!-- SC_OFF --><div class="md"><p>Hi guys</p> <p>Does anyone know of papers where EMA on LoRA adapters has been used successfully?</p> <p>Im interested in cases where the EMA adapter acts as a self-teacher generating soft labels for the trainable adapter.</p> <p>On-policy self-distillation [1] uses ema for the teacher. However, they seem to fully fine-tune. Any empirical results showing the idea is working on lora/ left models?</p> <p>[1] <a href="https://arxiv.org/abs/2601.19897">https://arxiv.

<table> <tr><td> <a href="https://www.reddit.com/r/MachineLearning/comments/1ubmybr/i_released_a_softmaxfree_attention_model_at_gpt2/"> <img src="https://external-preview.redd.it/fzbusCnVMF6KiLx-XGEbOtJ2hfOGlk4ouLmg5Wsh_8c.png?width=640&crop=smart&auto=webp&s=fb342a515f360dfb611d261f135a028577e8e501" alt="I released a softmax-free attention model at GPT-2 Medium scale (~354M params, 11.5B tokens): structural sparsity + tile-skipping kernels for long-context VRAM savings. Open weights + custom Tr

In this tutorial, we build a complete Crawlee for Python workflow from setup to AI-ready output. We generate a local demo website, then crawl it with BeautifulSoupCrawler, ParselCrawler, and PlaywrightCrawler. We extract titles, metadata, product fields, and JavaScript-rendered cards, and capture full-page screenshots. We then normalize the data, build a link graph, and export JSON, CSV, and RAG-ready JSONL chunks. The post Crawlee for Python: Build a Web Crawling Pipeline with Robots Handling,

Cisco Foundation AI has open-sourced FAPO (Fully Automated Prompt Optimization), a Claude Code-driven system that autonomously optimizes multi-step LLM pipelines from baseline prompts to target accuracy. FAPO evaluates a chain, attributes failures at the step level, proposes variants across prompt, parameter, and chain-structure levels, and validates each through an independent reviewer. In Cisco's evaluation, it beat GEPA on 15 of 18 model-benchmark comparisons. Here's how the optimization loop

Nous Research has added a Blank Slate setup mode to its open-source Hermes Agent. It starts an agent with everything off except provider, model, File Operations, and Terminal. You opt in to the rest. The post Nous Research Updates Hermes Agent With a Blank Slate Mode That Pins Toolsets via platform_toolsets.cli and disabled_toolsets appeared first on MarkTechPost .

<!-- SC_OFF --><div class="md"><p>I'm working with a client on a curve-fitting optimization problem. They are currently using a constrained Levenburg-Marquardt optimizer for their task which is complex, slow, and sometimes gets stuck in local minima.</p> <p>I suggested using particle swarm optimization (PSO), and the client suggested genetic algorithms (GA). I would like to compare the existing method to at least these two other options. For this first phase, I don't need to worry about speed or

<table> <tr><td> <a href="https://www.reddit.com/r/MachineLearning/comments/1uazlnd/hi_reddit_i_posted_my_build_your_own_llm_workshop/"> <img src="https://external-preview.redd.it/uyYAMRdaY-avR7vgrYnpYErzefthmf_lM5vYqR3t3jg.jpeg?width=320&crop=smart&auto=webp&s=f0cc363e42e1fb0d3fa6b30957932458a9634c3a" alt="Hi Reddit, I posted my Build Your Own LLM workshop to Youtube teaching ML, LLM and math intuition [P]" title="Hi Reddit, I posted my Build Your Own LLM workshop to Youtube teaching ML, LLM an

<!-- SC_OFF --><div class="md"><p>Suppose you’re a PhD advisor in machine learning.</p> <p>Your student has been in the program for 4 years, has done solid work, and has a coherent thesis direction but they haven’t published in an A*ML venue or top journal. No NeurIPS/ICML/ICLR/CVPR/etc., and no equivalent top venue in their subfield either but 3 First author A level paper.</p> <p>Would you still support them graduating if the thesis itself is solid? </p> </div><!-- SC_ON --> submitted by <a hre

<!-- SC_OFF --><div class="md"><p>I've been working through the internals of LLM inference and writing up what I learn as an open, in-progress handbook.</p> <p>Just wrapped another chapter on GPU execution and memory internals: why a GPU sits mostly idle during inference, how the memory hierarchy gates throughput, and where the real bottlenecks live. Added mermaid diagrams for the architecture pieces so the flow is easier to follow than a wall of text.</p> <p>It's a personal learning project, st

We build an end-to-end forecasting workflow with TimeCopilot on a panel of real airline passenger data and a synthetic seasonal series with injected anomalies. We evaluate statistical, foundation, and optional GPU-based models using rolling cross-validation and multiple error metrics. We generate probabilistic forecasts with prediction intervals, visualize future trends, and flag unusual observations. We then explore TimeCopilot's optional LLM agent, which selects a model and explains its predic

arXiv:2606.04075v2 Announce Type: replace-cross Abstract: Reinforcement learning (RL) has become a dominant post-training paradigm, enabling large language models (LLMs) to learn from rewards. We observe that societal regulations are structurally similar to reward functions. They define measurable outcomes, thresholds, and exceptions, while often leaving institutional intent only partially specified. We hypothesise that the RL training process may exploit these gaps and therefore ask whether mod

arXiv:2606.19899v1 Announce Type: cross Abstract: This paper addresses a rapidly emerging policy challenge: how to generate and interpret credible evidence about the biological capabilities and risks of AI scientists, or agentic AI systems capable of autonomously or collaboratively performing multi-step scientific tasks. As these systems enter real research workflows, decision-makers increasingly face evaluation results whose meaning depends on underlying design choices that are often implicit o

arXiv:2606.19914v1 Announce Type: cross Abstract: Art has long stood as a pivotal expression of human creativity. Embodied artificial intelligence offers a route for generative models to participate in that creativity through physical action rather than disembodied digital content. In robotic music co-creation, it is challenging to connect semantic musical understanding with real-time and physically executable performance. We present Co-policy, a framework for human-robot musical co-creation tha

arXiv:2606.19943v1 Announce Type: cross Abstract: Hyperspectral infrared observations are an important data source for numerical weather prediction (NWP) because they provide rich information on the vertical structure of atmospheric temperature and humidity. However, most existing deep learning methods mainly focus on one-way retrieval from radiances to atmospheric profiles, while the reverse radiance simulation process and the consistency between atmospheric state space and radiance observation

arXiv:2606.19975v1 Announce Type: cross Abstract: This paper examines the impact of artificial intelligence and digital technologies on the blue-collar gig economy in India, focusing on algorithmic management. This paper examines the impact of artificial intelligence and digital technologies on the blue collar gig economy in India, focusing on algorithmic management he use of automated systems to allocate, monitor, and evaluate work in location-based services such as ride sharing and delivery. U

arXiv:2606.19992v1 Announce Type: cross Abstract: In the agentic web era, LLM-based agents increasingly invoke web services as tools, yet most interfaces remain \emph{static endpoints} that poorly express long-horizon workflows with loops, conditionals, joins, and retries. We present ToolPro, which represents an agent's tool intent as an \emph{executable tool program} that compactly encodes multi-step service interactions with explicit effect types. ToolPro combines constraint-guided program con

arXiv:2606.20031v1 Announce Type: cross Abstract: Dynamic environmental changes, confined workspaces, and stringent real-time constraints make pathfinding in Robotic Mobile Fulfillment Systems (RMFS) a challenging problem for conventional search- and rule-based methods, which typically suffer from high computational complexity and long decision latency. While reinforcement learning (RL) has emerged as a powerful alternative, deploying learned policies with extreme energy efficiency on resource-c

arXiv:2606.20101v1 Announce Type: cross Abstract: Audio editing aims to modify specific content in an existing audio clip according to a natural language instruction while preserving the remaining acoustic content. Despite the remarkable progress of diffusion models, existing training-based editing methods mainly rely on the local inductive biases and cross-attention interaction in convolutional U-Net backbones, which often hinder long-range semantic alignment and precise understanding and local

arXiv:2606.20120v1 Announce Type: cross Abstract: Biological experiment protocols are written in natural language, whereas automation systems rely on predefined control commands, creating a semantic gap that limits autonomous execution. Microplate-based automatic experiments are particularly challenging due to the need to simultaneously control well mapping, sample-reagent combinations, replicate placement, and parallel dispensing. This study proposes an agent-based protocol translation framewor

arXiv:2606.20135v1 Announce Type: cross Abstract: Flow matching has emerged as a standard paradigm for robotic manipulation owing to its strong expressive power for modelling complex, multimodal action distributions, alongside similar approaches like diffusion policy. However, existing methods rely on discretized action chunks, making them brittle to demonstrations collected at heterogeneous control frequencies and prone to temporally inconsistent actions that degrade control stability. In this

arXiv:2606.20151v1 Announce Type: cross Abstract: This work proposes a hybrid ANN-SNN pipeline that effectively leverages the rich embeddings of pretrained artificial neural networks (ANNs) to enable high-performance spiking neural networks (SNNs). The architecture couples a pretrained EfficientNet encoder with a CoLaNET spiking classifier. We convert the encoder's activations into spike trains via rate-coding and train the subsequent SNN classifier using local, biologically inspired learning ru

arXiv:2606.20209v1 Announce Type: cross Abstract: Joint spatial and temporal understanding of 3D scenes is a crucial requirement for robots deployed in everyday household environments. Such agents must not only comprehend and navigate spatial layouts, but also reason about how these spaces evolve over time. In particular, humans interact with objects daily, causing them to change position throughout the environment and making it difficult for robots to reliably associate current observations wit

arXiv:2606.20235v1 Announce Type: cross Abstract: Academic paper search is a core step in scientific research, and LLM-based search agents are emerging as a promising paradigm for iterative, intent-driven literature exploration. However, existing benchmarks are insufficient for systematically evaluating agentic academic search under realistic open literature environments. We propose ScholarQuest, a large-scale, taxonomy-guided benchmark for agentic academic paper search. ScholarQuest is construc

arXiv:2606.20246v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) models pre-trained on massive video-robot datasets have revolutionized robotic manipulation, yet their multi-billion parameter architectures impose prohibitive computational burdens during downstream fine-tuning and real-time inference. In this work, we reveal a highly non-trivial architectural characteristic of these continuous control foundation policies (e.g., pi_0, GR00T-N1.5): despite being trained on diverse phy

arXiv:2606.20258v1 Announce Type: cross Abstract: The emergence of LLM-driven information services is reshaping the conditions under which public knowledge institutions operate, threatening to absorb the editorial function these institutions exist to exercise. While LLMs offer powerful new affordances for knowledge dissemination, editorial authority is challenged by pretrained LLMs that arrive already aligned with the values and dissemination strategies of their commercial developers. This paper

arXiv:2606.20280v1 Announce Type: cross Abstract: Leveraging Multimodal Large Language Models (MLLMs) via contrastive learning has become a mainstream paradigm for improving the performance of Universal Multimodal Retrieval (UMR). However, previous works have ignored the grain blindness when adapting the contrastive paradigm into retrieval tasks. Grain blindness refers to the tendency of the model to overlook grain-level information contained in the query, which is crucial for effectively handli

arXiv:2606.20356v1 Announce Type: cross Abstract: In this article, we present a robust $Q$-learning algorithm for discrete-time mean-field control problems under Wasserstein uncertainty in the common noise law. The algorithm combines a quantization-and-projection scheme with a Wasserstein dual reformulation on the common-noise space. We establish its convergence together with finite-time iteration bounds for both synchronous and asynchronous learning schemes. Numerical experiments on systemic ri

arXiv:2606.20373v1 Announce Type: cross Abstract: Large Language Models (LLMs) show promise for code compilation tasks, but applying them to runtime performance tuning is difficult due to complex microarchitectural effects and noisy runtime measurements. We present AutoPass, a multi-agent framework for compiler performance tuning that uses compiler and runtime evidence to guide LLM-generated optimization decisions. Rather than treating the compiler as a black box like prior auto-tuning schemes,