All entries

Imagine if i had a business that gave you 20 bills for 10 bills. As word spread, my customer base would grow. My revenue (how many $10 bil…

Imagine if i had a business that gave you 20 bills for 10 bills. As word spread, my customer base would grow. My revenue (how many $10 bills I collected) would grow. And I would continue to lose money

safetygary-marcus--x

Tutorials

Implementing super resolution by deploying SeedVR2 on Amazon SageMaker AI

DGX agent

In this post, we demonstrate how to implement video upscaling using SeedVR2 on SageMaker AI. We cover the solution architecture, walk through the deployment steps, and show performance comparisons tha

tutorialsaws-ml-blog

25 Jun 2026

Improved Large Language Diffusion Models

arXiv:2606.25331v1 Announce Type: new Abstract: Modern large language models are predominantly trained with autoregressive factorization and causal attention. We present iLLaDA, an 8B masked diffusion

researcharxiv-cs-cl

Improving Factuality of 3D Brain MRI Report Generation with Paired Image-domain Retrieval and Text-domain Augmentation

arXiv:2411.15490v2 Announce Type: replace Abstract: Acute ischemic stroke (AIS) requires time-critical decision-making, where inaccurate interpretation of neuroimaging findings can lead to irreversibl

Improving Neural Network Training by Decoupling the Magnitude and Direction of Weight Vectors

arXiv:2606.25971v1 Announce Type: new Abstract: Modern neural network training relies on optimizers such as Adam and Muon which act on each weight matrix as a single object. Yet every weight matrix ca

Improving Zero-Shot Offline RL via Behavioral Task Sampling

model-releasesarxiv-cs-ai

arXiv:2604.25496v2 Announce Type: replace Abstract: Offline zero-shot reinforcement learning (RL) aims to learn agents that optimize unseen reward functions without additional environment interaction.

In a joint Fireworks and @Faros_AI evaluation of 211 real engineering tasks, Claude Code + GLM-5.2 beat both Claude Code + Opus 4.8 and Code…

model-releasesfireworks-ai--x

In a joint Fireworks and @Faros_AI evaluation of 211 real engineering tasks, Claude Code + GLM-5.2 beat both Claude Code + Opus 4.8 and Codex + GPT-5.5: - Judge score: 0.568 vs. 0.521 and 0.466 - Time

Tutorials

🚨 In a surprise to NOBODY, a study shows that generative AI use raises homework scores, but substantially reduces learning. My takeaways fo…

🚨 In a surprise to NOBODY, a study shows that generative AI use raises homework scores, but substantially reduces learning. My takeaways for AI ethicists and educators: The study analyzed data from 26

tutorialsgary-marcus--x

In-context Region-based Drag: Drag Any Region to Any Shape

arXiv:2606.25907v1 Announce Type: new Abstract: Diffusion models have shown promise in drag-style editing. Previous works mainly focus on point-based drag, which is inherently ambiguous. This paper fo

researcharxiv-cs-cv

In-Context World Modeling for Robotic Control

arXiv:2606.26025v1 Announce Type: cross Abstract: Modern Vision-Language-Action (VLA) models often fail to generalize to novel setups, such as altered camera viewpoints or robot morphologies, because

In HF GGUF section of models, we are emphasizing MTP heads with its own sign 𝗠𝗧𝗣

model-releasesgeorgi-gerganov--x

In HF GGUF section of models, we are emphasizing MTP heads with its own sign 𝗠𝗧𝗣 llama.cpp adds MTP for the Qwen3.6 family This is a significant milestone for the local AI ecosystem. The performance j

In the end, will Elon be best known for his efforts to mainstream electric cars, for his rockets, …. or for his efforts to mainstream AI-gen…

In the end, will Elon be best known for his efforts to mainstream electric cars, for his rockets, …. or for his efforts to mainstream AI-generated porn? xAI is betting big on AI video and image genera

safetygary-marcus--x

In this episode, @OpenAI Chief Research Officer @markchen90 joins @allenpark to flambé shrimp, cook Korean stew, and chat about being at the…

In this episode, @OpenAI Chief Research Officer @markchen90 joins @allenpark to flambé shrimp, cook Korean stew, and chat about being at the frontier of AI research: why scaling laws and pre-training

model-releasesswyx--x

Industry

Incode acquires Identiq, commits $100M to privacy-first identity tech

Identity verification company Incode Technologies Inc. today announced that it has acquired Identiq Protocol Ltd., an Israeli startup that builds cryptographic tools letting companies share fraud sign

industrysiliconangle

Independent analyses estimate that your actions to dismantle USAID and drastically reduce lifesaving foreign aid have already killed 700,000…

Independent analyses estimate that your actions to dismantle USAID and drastically reduce lifesaving foreign aid have already killed 700,000 people. Here are some of those people. 🧵https://x.com/elonm

researchyann-lecun--x

IndicContextEval: A Benchmark for Evaluating Context Utilisation in Audio Large Language Models Across 8 Indic Languages

model-releasesarxiv-cs-cl

arXiv:2606.19157v2 Announce Type: replace-cross Abstract: AudioLLMs enable speech recognition conditioned on textual prompts such as domain descriptions or entity lists. However, it remains unclear wh

Internal Data Repetition Destroys Language Models

arXiv:2606.24998v1 Announce Type: new Abstract: Language models are running out of high-quality training data, and even aggressively deduplicated corpora retain some amount of repetition. Earlier cont

Tutorials

Interpretable Concept-Guided Polynomial Tabular Kolmogorov-Arnold Network for EEG-Based Mild Cognitive Impairment Detection

arXiv:2606.25434v1 Announce Type: new Abstract: Early and scalable detection of mild cognitive impairment (MCI) remains an unresolved clinical challenge. Existing EEG-based screening approaches are co

tutorialsarxiv-cs-lg

Introducing corpora Hlava Cor and Hlava AD: Human Label Variation in Coreference and Discourse Relations

arXiv:2606.25383v1 Announce Type: new Abstract: As previous research on annotator disagreement in discourse phenomena has shown, understanding text coherence varies considerably from one individual to

researcharxiv-cs-cl

Invariant Kalman filtering for extended pose estimation in multi-IMU articulated rigid-body systems

arXiv:2606.25083v1 Announce Type: new Abstract: Accurate extended pose estimation (orientation, velocity, and position) for IMU-instrumented articulated rigid-body systems is a key challenge in roboti

researcharxiv-cs-ro

Inverse Reinforcement Learning for Interpretable Keystroke Biomarkers in Parkinson's Disease

arXiv:2606.25270v1 Announce Type: new Abstract: Keystroke dynamics have been explored extensively as a passive digital biomarker for Parkinson's disease (PD), typically by extracting summary statistic

InvestPhilBench: A Multi-Layer Dynamic Benchmark for Evaluating Large Language Model Procedural Reasoning in Expert Investment Philosophy

arXiv:2606.25984v1 Announce Type: cross Abstract: Large language models are increasingly deployed as investment research assistants, yet no benchmark tests whether they can accurately reconstruct and

Invisible to humans, visible to machines: a preregistered audit of Unicode fidelity across four biomedical bibliographic APIs

arXiv:2606.24897v1 Announce Type: cross Abstract: Biomedical text mining, scientometrics, and the construction of training corpora for biomedical large language models (LLMs) all assume that the abstr

researcharxiv-cs-cl

Invoice Haystack: Benchmarking Document Retrieval and Visual Question Answering Under Strong Visual Homogeneity

arXiv:2606.25343v1 Announce Type: new Abstract: Vision Language Models have achieved near-human performance on single-document Visual Question Answering, yet their effectiveness degrades significantly

Agents

Is GraphRAG Needed? From Basic RAG to Graph-/Agentic Solutions with Context Optimization

arXiv:2606.25656v1 Announce Type: new Abstract: As advanced RAG variants like GraphRAG and Agentic RAG emerge, one leading question is when and how to use them. Here, we introduce a framework for diff

agentsarxiv-cs-cl

Is Variational Monte Carlo Robust? Sharp Moment Thresholds and Heavy-tailed Stochastic Optimization

arXiv:2606.26009v1 Announce Type: new Abstract: Variational Monte Carlo (VMC) is a central algorithm in electronic structure theory and has gained renewed importance through modern neural-network ansa

It is amazing to me how little the X-verse cares about the rise of authoritarianism in the United States. Is that because most of the “peopl…

It is amazing to me how little the X-verse cares about the rise of authoritarianism in the United States. Is that because most of the “people” here are actually Russian bots? 🤔 More Soviet-style decep

safetygary-marcus--x

It would be very useful to understand more about the government safety concerns associated with frontier AI releases so we could (a) know wh…

It would be very useful to understand more about the government safety concerns associated with frontier AI releases so we could (a) know what risks everyone will face if/when open source reaches Myth

safetyethan-mollick--x

It’s time for everyone to realize that the fight against data centers has nothing to do with data centers. They have become a proxy for the …

It’s time for everyone to realize that the fight against data centers has nothing to do with data centers. They have become a proxy for the hate towards AI and the concentration and accumulation of we

researchyann-lecun--x

Hardware

JetSpec: Breaking the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting

arXiv:2606.18394v2 Announce Type: replace Abstract: Speculative decoding (SD) accelerates autoregressive Large Language Models (LLMs) by drafting multiple tokens and verifying them in parallel, but it

hardwarearxiv-cs-cl

Agents

Just had a great discussion on dynamic workflows. Rough notes: - applies to a very small set of use cases - think of it as a new paradigm of…

Just had a great discussion on dynamic workflows. Rough notes: - applies to a very small set of use cases - think of it as a new paradigm of (test-time compute) TTC - strong for hill-climbing research

agentsdair-ai--x

KidRisk: Benchmark Dataset for Children Dangerous Action Recognition

arXiv:2606.25298v1 Announce Type: new Abstract: Children are naturally energetic, and during their spontaneous activities, they often encounter potentially dangerous situations, especially when lackin

Tutorials

Knowledge Cascade: Reverse Knowledge Distillation on Nonparametric Multivariate Functional Estimation

arXiv:2606.25927v1 Announce Type: cross Abstract: As machine learning models and datasets continue to grow, developing complex models has become increasingly computationally demanding. Knowledge disti

tutorialsarxiv-cs-lg

Laplace--Fisher Gate Identities for Optimal Matrix-Gated Blended Score Estimation

arXiv:2606.25169v1 Announce Type: cross Abstract: Sampling from an unnormalized target by reversing an Ornstein--Uhlenbeck diffusion requires the score of each noise-perturbed marginal. Tweedie's iden

Large-Scale Tunnel Air--Ground Collaboration With FLISP: Fast LiDAR-IMU Synchronized Path Planne

arXiv:2606.25393v1 Announce Type: new Abstract: Hydropower tunnel inspection is critical for infrastructure integrity yet remains inefficient and hazardous using manual methods. We propose FLISP (Fast

researcharxiv-cs-ro

Latent Block-Diffusion Temporal Point Processes: A Semi-Autoregressive Framework for Asynchronous Event Sequence Generation

arXiv:2606.24982v1 Announce Type: new Abstract: Modeling and sampling from the underlying distribution of asynchronous event sequences are crucial in various real-world applications, including social

Latent Space Analysis for Interpretable Uncertainty in Melanoma Classification

arXiv:2506.18414v3 Announce Type: replace Abstract: Melanoma is a highly aggressive skin cancer, making early and accurate diagnosis critical. While deep learning excels in skin lesion classification,

researcharxiv-cs-cv

Leaking Circuit Secrets: Gradient Leakage Attacks on Graph Neural Networks

arXiv:2606.25589v1 Announce Type: new Abstract: As graph neural networks (GNNs) become standard tools for critical tasks in circuit design and analysis, their security and privacy risks require carefu

Learning Action Priors for Cross-embodiment Robot Manipulation

arXiv:2606.26095v1 Announce Type: cross Abstract: Most Vision-Language-Action (VLA) models build on a Vision-Language Model (VLM) backbone by attaching an action module and optimizing the full policy

safetyarxiv-cs-cv

Learning Asynchronous Upper-body Task-space Trajectory Tracking Policy for Humanoid Robots

arXiv:2606.25706v1 Announce Type: new Abstract: High-level humanoid planners often output sparse task-space, low-rate trajectories, whereas whole-body controllers run at high frequency. This creates t

safetyarxiv-cs-ro

Applications

Learning Diachronic Representations of Ancient Greek Letterforms

arXiv:2606.24984v1 Announce Type: cross Abstract: Learning representations that remain robust across centuries of variation in handwriting is a key challenge in diachronic representation learning. Tak

applicationsarxiv-cs-cl

Learning Dynamical Systems from Multiple Sparse Datasets: A Hierarchical Bayesian Modeling Approach

arXiv:2606.24966v1 Announce Type: new Abstract: Estimating parameters of dynamical systems from sparse, noisy, and irregularly sampled data is often severely ill-conditioned. When multiple related dat

Learning Non-Vacuous Generalization Bounds from Optimization

arXiv:2206.04359v3 Announce Type: replace Abstract: One of the fundamental challenges in the deep learning community is to theoretically understand how well a deep neural network generalizes to unseen

Learning Optimization Proxies for Sequential Contextual Stochastic Programs: An Order Fulfillment Application

arXiv:2606.25362v1 Announce Type: cross Abstract: Sequential contextual stochastic programs model real-time decision systems in which each time epoch commits to an action under uncertainty whose conse

safetyarxiv-cs-lg

Learning Perceptive Platform Adaptive Locomotion Controllers for Quadrupedal Robots

arXiv:2606.25179v1 Announce Type: new Abstract: Universal quadrupedal locomotion remains limited by the difficulty of integrating perception across diverse robot morphologies. State-of-the-art control

researcharxiv-cs-ro

Learning Robot Visual Navigation in Crowds via Intention-Aware Scene Representations

arXiv:2606.26047v1 Announce Type: new Abstract: Robot crowd navigation requires the ability to infer human intentions while accounting for the structural constraints of the environment. Currently, dee

safetyarxiv-cs-ro

Applications

Learning Structural Hardness for Combinatorial Auctions: Instance-Dependent Algorithm Selection via Graph Neural Networks

arXiv:2602.14772v2 Announce Type: replace Abstract: The Winner Determination Problem (WDP) in combinatorial auctions is NP-hard, and no existing method reliably predicts which instances will defeat fa

applicationsarxiv-cs-lg

Learning Subset-Shared Invariances for Domain Generalization with Mixture-of-Experts

arXiv:2606.25665v1 Announce Type: new Abstract: Domain generalization (DG) aims to learn a model from one or more source domains that generalizes to an unseen target domain without accessing target da

safetyarxiv-cs-lg