#ai le 2026-10-07

1. 04:39 <Ada> arXiv.org | TAPDreamer: Transferable Adversarial Patches for World Action Models 4h ago · 3 min read World models learn to predict how their environment will evolve, making them an important foundation for general-purpose robotic control. Yet world action models depend on camera inputs whose manipulation can corrupt the visual representations used across tasks and action policies. Existing attacks on these models optimize against the victim's actions or ... Read full story
2. 04:39 <Ada> arXiv.org | Local2Mesh: Spatially Localized Contour-to-Mesh for Left Ventricular Reconstruction from Sparse 2D Cardiac MRI Haoyu Wu, Ling Lin, Pascal Lef\`evre, Ruizhe Li, Xiaowu Sun · 4h ago · 3 min read Three-dimensional (3D) left ventricular (LV) reconstruction from sparse cardiac magnetic resonance (CMR) imaging remains challenging due to inter-slice misalignment and insufficient local spatial information between slices. Global aggregation of contour features may obscure local contour-to-surface relationships. We propose Local2Mesh, a spatially localiz... Read full story
3. 04:40 <Ada> arXiv.org | Incentive Alignment in Online Experimentation Ermis Soumalias, Richard Mudd, Abbas Zaidi · 4h ago · 3 min read Evaluating the causal effect of new features is a central goal for online platforms. While recent literature addresses limited testing traffic via centralized portfolio optimization, this perspective abstracts away a critical institutional reality: experimentation is operationally decentralized. The experimenters who develop new features also dictate whic... Read full story
4. 04:40 <Ada> arXiv.org | Best-of-$N$ Guidance for Test-time Diffusion Alignment 4h ago · 3 min read Diffusion models achieve strong generative performance but often struggle to align generated samples with human preferences measured by a reward model. A simple yet effective algorithm for test-time alignment is Best-of-$N$ (BoN) sampling, which draws $N$ i.i.d. samples from a pre-trained diffusion model and outputs the single highest-reward sample. Despi... Read full story
5. 04:40 <Ada> arXiv.org | E$^2$-OPSD: Taming Entropy Overshoot in On-Policy Self-Distillation 4h ago · 3 min read On-policy self-distillation (OPSD) provides dense token-level supervision without a second model: one network acts as teacher with the reference solution and as student with only the problem. We identify a specific failure mode of this recipe. During training, student token entropy rises past the teacher's and remains elevated, a pattern we call entropy o... Read full story
6. 04:40 <Ada> arXiv.org | A Systematic Analysis of the Predictive Power of LM Surprisal in Reading Chinese 4h ago · 3 min read This study analyzes the predictive power of LM-derived, token-level surprisal on Mandarin Chinese reading times. We first propose the Shortest Matching Sequence (SMS), an alignment scheme that maps between the word segmentation assumed by eye-tracking corpora and the LMs' subword tokenization, as the two tokenizations often disagree in the context of Mand... Read full story
7. 04:40 <Ada> arXiv.org | More Value per Key: Asymmetric Sparse Attention for Faster LLM Decoding 4h ago · 3 min read Autoregressive generation in Large Language Models (LLMs) is constrained by the memory and computational demands of attention mechanisms. Sparse attention methods mitigate this cost by selecting only high-probability entries of the attention matrix. We observe that in many such methods, this renders the probability-value multiplication negligible, shiftin... Read full story
8. 04:40 <Ada> arXiv.org | WAMJET: A Harness for World Action Model Acceleration 4h ago · 3 min read World Action Models (WAMs) leverage pretrained video foundation models for robot manipulation, but their large backbones and video-action co-prediction are expensive. Although existing acceleration techniques offer many ways to reduce this cost, selecting and composing them requires substantial engineering for each model and hardware platform. To tackle t... Read full story
9. 04:40 <Ada> arXiv.org | Distributed Learning with Selective State Space Models: Architecture-Aware Convergence Analysis 4h ago · 3 min read Modern state space models (SSMs), such as Mamba2, provide a compelling alternative to transformers by combining linear-time sequence modeling with recurrent state-space dynamics. However, the behavior of SSMs in distributed learning settings remains poorly understood. In particular, the existing standard federated learning methods are largely architecture... Read full story
10. 04:40 <Ada> arXiv.org | Completion Aware Guidance for World Action Models 4h ago · 2 min read World Action Models (WAMs) predict visual futures and robot actions, yet they remain susceptible to task-incomplete imagination, where plausible, action-consistent predictions omit the transition needed for task completion. In this paper, we show that this failure is not inherent to the world model backbone, but emerges when adapted for short-chunk contro... Read full story
11. 04:41 <Ada> arXiv.org | A Response Theory Probe for Learned Stochastic AI Simulators, Tested on Lorenz-63 4h ago · 3 min read Machine-learning emulators of chaotic and stochastic systems are usually validated on forecast skill and long-run statistics. Neither certifies that an emulator responds correctly to forcing, the property that projection and attribution studies rely on. Linear response theory makes this testable: the forced response follows from unperturbed correlations t... Read full story
12. 04:41 <Ada> arXiv.org | Quantum data loading from the learned shared structure of real signals 4h ago · 3 min read Preparing quantum states from classical data can cost more than the computation they serve; most loaders tailor a circuit to each input. Here we show that the signals of a real dataset share structure that can be learned once and reused. Our quantum-native loader learns a low-dimensional description of a dataset and prepares every signal with one fixed ci... Read full story
13. 04:41 <Ada> arXiv.org | Incentive Alignment in Online Experimentation Ermis Soumalias, Richard Mudd, Abbas Zaidi · 4h ago · 3 min read Evaluating the causal effect of new features is a central goal for online platforms. While recent literature addresses limited testing traffic via centralized portfolio optimization, this perspective abstracts away a critical institutional reality: experimentation is operationally decentralized. The experimenters who develop new features also dictate whic... Read full story
14. 04:41 <Ada> arXiv.org | SpecFold: Folding Multi-Branch Redundancy for Faster Speculative Decoding in Diffusion Language Models 4h ago · 3 min read Diffusion large language models (DLLMs) generate text through iterative block denoising, and multi-branch speculative decoding accelerates this process by verifying a main branch together with multiple draft branches in a single forward pass. While prior DLLM acceleration methods primarily exploit temporal redundancy across denoising steps, we identify a ... Read full story
15. 04:41 <Ada> arXiv.org | AID: A Framework for AI Infrastructure Dynamics Abi Aryan · 4h ago · 3 min read A useful model of AI inference infrastructure must specify the system state, the information available to an observer, and the decisions the model is intended to support. We introduce AID (AI Infrastructure Dynamics), a framework for describing this learning problem across coupled physical, computational, networking, and serving processes. The formulation... Read full story
16. 04:41 <Ada> arXiv.org | Verifiable, Articulable, and Tacit Components of Preference 4h ago · 3 min read What makes a short story gripping; a news article newsworthy; or a math proof elegant? These constructs resist articulation or verification; their meaning is at least partially tacit. However, modern AI models are improved primarily via articulated constitutions, rubrics and verifiers (i.e. in RLAIF and RLVR); tacit components of preferences are typically... Read full story
17. 04:41 <Ada> arXiv.org | Decoupling Memory from Context: Structured Memory for Token-Efficient Test-Time Continual Learning Yehya Farhat, Michael Desmond, Anastasios Kyrillidis · 4h ago · 3 min read Large language models (LLMs) are increasingly deployed in enterprise, scientific, and medical applications, where agents must incorporate domain-specific knowledge and adapt from experience. Context engineering offers a practical alternative to weight updates by improving model behavior through instructions, strategies, and evidence supplied at inference ... Read full story
18. 04:41 <Ada> arXiv.org | Sequential Capacity of Quantum Processes with Finite Memory Yibin Wang · 4h ago · 3 min read How complex can the responses of a quantum device become as it runs longer with a fixed internal memory? We quantify this complexity through sequential response capacity: how many adaptive testing stages, each using a fresh run, can continue to separate possible processes by a prescribed gap in response probabilities. For fixed system and memory sizes, we... Read full story
19. 04:42 <Ada> arXiv.org | Completion Aware Guidance for World Action Models 4h ago · 2 min read World Action Models (WAMs) predict visual futures and robot actions, yet they remain susceptible to task-incomplete imagination, where plausible, action-consistent predictions omit the transition needed for task completion. In this paper, we show that this failure is not inherent to the world model backbone, but emerges when adapted for short-chunk contro... Read full story
20. 04:42 <Ada> arXiv.org | FineART: Fine-Grained Annotated Robotic Trajectory Dataset and Vision-Language-Action Model for Bimanual Manipulation 4h ago · 3 min read Robots operating in real-world environments must often execute complex, multi-step bimanual tasks over long horizons rather than single, isolated actions. Current manipulation datasets provide limited support for this capability: although single-arm datasets reach hundreds of thousands of trajectories, they typically provide only one high-level instructio... Read full story
21. 13:34 <Ada> Hugging Face | One Model Family, Two Gold-Level Results: Fine-Tuning Nemotron for IOI and IMO Aleksander · 49m ago · 5 min read A Blog post by NVIDIA on Hugging Face Read full story
22. 13:34 <Ada> Hugging Face | Introducing Falcon ASR Abdul Muneer · 13m ago · 7 min read A Blog post by Technology Innovation Institute on Hugging Face Read full story
23. 17:03 <Ada> Hugging Face | Multimodal open d1 decision models for the edge Aurelien Lac · 8m ago · 5 min read A Blog post by Liquid AI on Hugging Face Read full story
24. 17:05 <Vega> OpenAI News | Helping teens learn, plan, and shape the future of AI 5h ago College Planner is coming to ChatGPT for Teens to help students manage college applications, alongside new flashcards, quizzes, and a teen AI council. Read full story
25. 18:06 <Vega> OpenAI News | Radisson Hotel Group brings hotel discovery into ChatGPT 11h ago Radisson partnered with Accenture to build a ChatGPT plugin using OpenAI technology, helping travelers find, compare, and book hotels while planning their trips. Read full story
26. 19:07 <Vega> OpenAI News | GPT-6 and Intelligent UI for everyone 19h ago GPT‑6 is rolling out globally in ChatGPT with Intelligent UI, delivering faster responses with visuals and interactive experiences you can explore and use directly. Read full story
27. 21:05 <Vega> The latest research from Google | Does better work always mean better workers? 45m ago · 8 min read Does better work always mean better workers? October 7, 2026 David Autor, Technology & Society Visiting Fellow, and Tanya Rodchenko, Principal, AI & Economy In a three-month randomized controlled trial with practicing patent attorneys, we found that, while routinely using AI lifts work quality across the board, its effects on on-the-job learning depend on... Read full story
28. 21:12 <Vega> MIT News | Massachusetts Institute of Technology | Discovering the value of humanistic inquiry 5h ago · 5 min read Students in in MIT’s Concourse first-year learning community study the humanities as well as taking their core math, science, and engineering courses together. In this tight-knit community, they learn to think deeply about the human condition, debate challenging topics, and develop judgment about non-quantitative issues. Read full story

Tous les jours · Rejoindre #ai