Knowledge catalogue
Today we release a study on decoupling the benefits of subword tokenization for language model training, by simulating each suspected benefit one at a time inside a 1.7B byte-level pretraining pipelin
Token Maxxing, April 2026-May 2026, RIP 🪦 🦔Microsoft canceled its internal Claude Code licenses this week after token-based billing made the cost untenable, even for a company with effectively infinit
arXiv:2605.20963v1 Announce Type: new Abstract: The proliferation of unmanned aerial vehicles (UAVs) has created urgent demand for precise UAV monitoring. Existing RGB-based systems rely on spatial cu
arXiv:2410.04155v2 Announce Type: replace Abstract: How to defend large language models (LLMs) from generating toxic content is an important research area. Yet, most research focused on various model
arXiv:2605.20885v1 Announce Type: new Abstract: Precision oncology requires predicting which drugs will suppress a specific tumor from its molecular profile, but drug-blind sensitivity prediction has
arXiv:2605.20477v1 Announce Type: cross Abstract: Language agents can adapt from experience in interactive environments, but current reflection-based methods can only self-correct within a single task
arXiv:2408.08812v2 Announce Type: replace Abstract: Deployed reinforcement learning agents often face safety requirements that are specified only after training, such as new hazard maps, revised risk
Try Composer 2.5 New CursorBench results just dropped. Two big takeaways. Composer 2.5 is way better than most people think. 63.2% score at $0.55 per task. Nearly matching Opus 4.7 Max and GPT 5.5 Ext
arXiv:2605.20803v1 Announce Type: new Abstract: Continual learning (CL) aims to train models sequentially on multiple tasks while mitigating catastrophic forgetting of previously learned knowledge. Re
arXiv:2605.20202v1 Announce Type: new Abstract: I study whether emotionally framed evaluation follow-ups change both the behavior and the calm-relative internal representations of small, locally deplo
arXiv:2511.09557v4 Announce Type: replace-cross Abstract: As large language models (LLMs) continue to grow in size, distributed inference has become increasingly important. Model-parallel strategies m
arXiv:2505.19075v3 Announce Type: replace-cross Abstract: Large Language Models (LLMs) have demonstrated remarkable general capabilities, but enhancing skills such as reasoning often demands substanti
arXiv:2605.21079v1 Announce Type: new Abstract: Capturing digital screens with smartphones frequently induces severe banding due to hardware synchronization mismatches. Existing video restoration meth
arXiv:2605.21002v1 Announce Type: cross Abstract: Generative artificial intelligence now synthesizes photorealistic imagery, audio, and video at a cost that defeats traditional forensic intuition. The
arXiv:2605.21130v1 Announce Type: new Abstract: Large Multimodal Models (LMMs) have shown promise for video quality assessment, but most methods still predict an absolute score for each video. Such po
arXiv:2605.21268v1 Announce Type: new Abstract: Land Use Scene Classification (LUSC) from remote sensing imagery plays a critical role in environmental monitoring, urban planning, and sustainable reso
arXiv:2605.20901v1 Announce Type: new Abstract: We propose VISTA, a V-JEPA Integrated StillFast Temporal Anticipator for the Ego4D Short-Term Object Interaction Anticipation (STA) Challenge at EgoVis
arXiv:2605.20676v1 Announce Type: new Abstract: Establishing a clear link between model predictions and the visual evidence that supports them is critical for transparency and reliability in multimoda
arXiv:2605.20774v1 Announce Type: new Abstract: Vision-Language-Action (VLA) models have shown strong promise for general-purpose robotic manipulation, but their real-world evaluation remains limited
arXiv:2605.20821v1 Announce Type: new Abstract: Detecting what has changed in an environment is essential for long-term autonomy, yet most change detection settings assume fixed viewpoints, mild misal
arXiv:2605.20311v1 Announce Type: new Abstract: Guided-wave structural health monitoring enables damage localization in composite plates using sparse networks of bonded piezoelectric transducers. Howe
arXiv:2605.21097v1 Announce Type: new Abstract: Web content extraction - isolating a page's main content from surrounding boilerplate - is a prerequisite for search indexing, retrieval-augmented gener
arXiv:2605.20441v1 Announce Type: new Abstract: Transformers trained on modular arithmetic exhibit sharp transitions between memorization, generalization, and collapse. We show that weight decay acts
Google DeepMind has launched 'AI for the Planet,' a three-month accelerator program in Asia Pacific focused on leveraging advanced AI to combat environmental challenges like climate change, biodiversi
arXiv:2605.20537v1 Announce Type: new Abstract: Biomedical named entity recognition (NER) and entity linking (EL) strongly depend on annotated corpora, but the utility of these resources for benchmark
arXiv:2605.21404v1 Announce Type: new Abstract: We read twelve well-known LLM agent benchmark papers and recorded, dimension by dimension, what each paper actually says about how its evaluation was ru
arXiv:2605.20558v1 Announce Type: new Abstract: Neural morphological generation systems often achieve high aggregate accuracy on benchmark datasets, yet such performance can conceal systematic errors
arXiv:2603.09024v2 Announce Type: replace Abstract: Sudden concept drift makes previously trained predictors unreliable, yet deciding when to retrain and what post-drift data size is sufficient is rar
arXiv:2605.21479v1 Announce Type: new Abstract: Visual Question Answering (VQA) benchmarks have largely emphasized perception-based tasks that can be solved from visual content alone. In contrast, man
arXiv:2605.20306v1 Announce Type: new Abstract: We introduce WildRoadBench, a wild aerial road-damage grounding benchmark that couples direct visual grounding by vision-language models with autonomous
arXiv:2605.20922v1 Announce Type: cross Abstract: Oscillations and synchronization are widely believed to play a fundamental role in representation and computation. However, existing machine learning
This post shows a Wordle game result where the player solved puzzle #1,796 in 4 attempts, with the final answer being a five-letter word with the pattern shown in green squares. The emoji grid display
arXiv:2605.21468v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) has become a dominant paradigm for improving reasoning in large language models (LLMs), yet the
arXiv:2605.20485v1 Announce Type: new Abstract: As autonomous agents increasingly execute end-to-end tasks under fixed monetary budgets, the pressing open question shifts from whether the budget is re
Google I/O 2026 unveiled new models, agents and tools to help users build, search, create, discover, shop and get more done. Key announcements included Gemini Omni, Google Antigravity, and Universal C
arXiv:2605.19407v1 Announce Type: cross Abstract: We investigate data filtering for large model pretraining via new scaling studies that target the high compute, data-scarce regime. In spite of an app
arXiv:2605.19988v1 Announce Type: cross Abstract: Documentation has long guided computer system tuning by distilling expert knowledge into per-parameter recommendations. Yet such guides capture only w
arXiv:2605.19618v1 Announce Type: new Abstract: Validating interpretable surrogate models for ensemble learners requires measuring agreement between the ensemble's internal representation and its surr
arXiv:2603.15411v2 Announce Type: replace Abstract: Accurate prediction of crop states (e.g., phenology stages and cold hardiness) is essential for timely farm management decisions such as irrigation,
arXiv:2605.18802v1 Announce Type: cross Abstract: Cardiovascular stability estimation from wearable photoplethysmography (PPG) requires a principled nonlinear framework, yet major gaps persist in heur
arXiv:2605.18780v1 Announce Type: cross Abstract: Reasoning-based Large Language Models (LLMs) like PO4ISR have set new benchmarks in session-based recommendation. However, the reproducibility of thei
arXiv:2605.19020v1 Announce Type: new Abstract: Vision foundation models have demonstrated strong transferability across diverse visual recognition tasks and are increasingly considered for biometric
arXiv:2605.18898v1 Announce Type: new Abstract: We apply the Weibull distribution -- a two-parameter family from extreme-value theory -- as a diagnostic framework for element-wise weight magnitude dis
arXiv:2506.14148v2 Announce Type: replace-cross Abstract: This paper presents a novel non-invasive object classification approach using acoustic scattering, demonstrated through a case study on hair a
arXiv:2512.00667v2 Announce Type: replace-cross Abstract: Effective medical simulators necessitate realistic haptic rendering of biological tissues that exhibit viscoelastic material properties, such
arXiv:2002.09053v3 Announce Type: replace Abstract: Pedestrian detection benefits from deep learning technology and gains rapid development in recent years. Most of detectors follow general object det
arXiv:2602.11454v3 Announce Type: replace-cross Abstract: We study left(epsilon,eltaright)-differentially private algorithms for the problem of approximately computing the top singular vector of a mat
This article covers the NVIDIA AI-Q Blueprint for building specialized deep research agents that empower AI systems to gather context, synthesize information, and support complex decision-making acros
arXiv:2511.01959v2 Announce Type: replace-cross Abstract: Pulsar Timing Arrays (PTA) provide a powerful framework to measure low-frequency gravitational waves, but accuracy and robustness of the resul
arXiv:2605.19009v1 Announce Type: new Abstract: Humanoid robots are difficult to deploy safely because they have high-dimensional bodies, many collision constraints, and must operate near people and o
arXiv:2605.19728v1 Announce Type: new Abstract: Foundation video models produce visually impressive results, but their use in embodied AI remains limited because they are primarily trained on natural
arXiv:2605.19149v1 Announce Type: new Abstract: Agents operating with computer and Web use inevitably encounter errors: inaccessible webpages, missing files, local and remote misconfigurations, etc. T
arXiv:2605.19010v1 Announce Type: new Abstract: Natural language to SQL (NL2SQL) conversion is an important problem for researchers and enterprises due to the ubiquitous importance of relational datab
Google I/O 2026 featured several new AI model releases including Gemini 3.5 Flash, an Omni model codenamed NanoBanana for video processing, Spark for background agent tasks, and Antigravity 2.0. These
arXiv:2605.19325v1 Announce Type: new Abstract: Nonnegative matrix factorization (NMF) seeks a low-rank approximation X approx UV^T with nonnegative factors and is commonly solved using interior metho
Samantha Kelly / Bloomberg: An interview with Match Group CEO Spencer Rascoff about plans for Tinder, including a redesign, AI features, live events, and group dating to win over Gen Z — The dating ap
arXiv:2605.13793v2 Announce Type: replace Abstract: Arguments are a fundamental aspect of human reasoning, in which claims are supported, challenged, and weighed against one another. We present an end
An OpenAI AI model successfully disproved a longstanding conjecture in discrete geometry, a mathematical field studying geometric properties of discrete objects. This achievement demonstrates the pote
Anyone understand what Google mean by 'Gemini Spark runs on Gemini 3.5 and uses the Antigravity harness' - is 'Antigravity' a generic term they're using for their agent harnesses now or is their Claw-
arXiv:2605.19852v1 Announce Type: new Abstract: Tool-augmented reasoning has emerged as a promising direction for enhancing the reasoning capabilities of multimodal large language models (MLLMs). Howe