September 2026: 1244 archived news items, newest first. Items here are older than the 24-hour Latest News window; the live feed is at /news/.
1.11.0 (2026-09-30) Full Changelog: v1.10.0...v1.11.0 Features api: add list spend limits endpoint ( e8f0cac ) Chores client: deprecate sonnet 4.5 ( 22b36b1 )
The thing to notice is anthropics/anthropic-sdk-python; decide whether it changes your next build. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
Google’s science AI model was the best at forecasting flu-related hospital admissions, the Centers for Disease Control announced.
This lands on Google's AI ranks #1 for predicting flu hospitalizations - gauge whether it shifts what you already ship. For builders, a release like this shifts the latency, capability, and cost trade-offs of what to run next.
Trusted publishing configurations for npm can now be granted permission to manage dist-tags (e.g., promoting a version to latest, updating next and beta pointers) using short-lived OIDC credentials instead of...
Teams affected by Opt-in dist-tag permissions for npm trusted publishing need to decide whether its documented change alters their current workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
GitHub Releases published a source-backed AI update around comfyanonymous/ComfyUI: v0.38.1. RepoRadar is keeping the source link direct for verification.
Operators using related systems should check whether comfyanonymous/ComfyUI changes compatibility, cost, or access requirements. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
Stylized promotional blog key art graphic with modern editorial branding and the text "Gemini 4 Argon"
What's actually new here is Gemini 4 Argon - see if it moves anything you maintain. For builders, a release like this shifts the latency, capability, and cost trade-offs of what to run next.
Google is investing $1.2M to line Chile's Unidos de Buin canal, saving 1.9 billion gallons of water annually. See how we boost watershed health.
This lands on Google is supporting water resilience in Chile - gauge whether it shifts what you already ship. For builders, this may affect how you choose, deploy, or govern AI tools this week.
Anthropic's Python SDK v1.10.0 adds Admin API support for Claude Enterprise analytics, spend limits, RBAC groups and roles, per-user usage and cost reports, plugin marketplaces, plus Managed Agents refusal stop reasons.
Teams administering Claude Enterprise can upgrade the SDK to script spend limits and per-user cost reports, which helps operators control cost and access without manual console work.
The Alliance for America’s Skilled Trades is expanding, welcoming 14 new partners to accelerate training and career access nationwide. For every 100 skilled trades worker...
What's actually new here is More partners are joining the Alliance for America’s - see if it moves anything you maintain. For builders, this can change a default in the review, editor, or build workflow you touch every day.
GitHub Team customers can now start self-serve trials of GitHub Advanced Security to evaluate GitHub Code Security and GitHub Secret Protection. Start a trial from your organization’s Overview page, Billing...
Teams affected by GitHub Advanced Security trials for GitHub Team need to decide whether its documented change alters their current workflow. For teams giving agents real access, this surfaces a failure mode worth weighing before widening agent or tool...
Hugging Face Transformers 5.18.0 adds new architectures including NVIDIA Nemotron 3 Diarization (streaming speaker diarization, up to eight speakers), NemotronH Omni multimodal reasoning, and NAVER HyperCLOVAX Vision V2.
Developers building speech or multimodal pipelines can upgrade to load these models through the standard library instead of vendor code, which lowers integration cost and keeps workflow tooling consistent.
Graduation caps and icons in different colors
The thing to notice is Supporting AI readiness in higher education; decide whether it changes your next build. For builders, this may affect how you choose, deploy, or govern AI tools this week.
Google is rolling out reusable skills in Gemini chat globally, invoked with a forward slash. Skills will replace Gems, existing Gems will migrate automatically, and Workspace business and education customers get it in coming weeks.
Gemini users who built Gems should plan to migrate their saved instructions to skills, and Workspace admins should decide on rollout before the coming-weeks enablement changes daily workflow for their teams.
The illustration shows people around the world using Google Translate.
What's actually new here is Celebrating International Translation Day - see if it moves anything you maintain. For builders, this may affect how you choose, deploy, or govern AI tools this week.
This October, Waze will help drivers find screening clinics, assess their health risks, and schedule routine checkups.
Operators using related systems should check whether Waze rolls out new features to support Breast changes compatibility, cost, or access requirements. For builders, this may affect how you choose, deploy, or govern AI tools this week.
Changes since langchain-openai==1.6.6 release(openai): 1.6.7 ( #40933 ) chore(model-profiles): refresh model profile data ( #40924 ) test(openai): drop retired completions live tests ( #40910 ) chore(model-profiles): refresh model profile data ( #40869 ) feat(openai): discover Azure workload identity ( #40532 )
The thing to notice is langchain-ai/langchain; decide whether it changes your next build. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
Google DeepMind introduced SynthID Bio, a proof of concept that watermarks AI-designed proteins in their sequence so the mark survives physical synthesis, with lab tests showing biological function preserved.
Biosecurity teams and protein-design researchers should test provenance checks now, because AI-designed sequences can bypass DNA synthesis screening and mislabeled structures risk polluting public databases.
Google DeepMind introduces SynthID Bio to watermark AI-designed proteins while maintaining biological function. Read the full research report here.
This lands on We’re introducing SynthID Bio, bringing our watermarking technology - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
GitHub expanded the HydraFusion research preview to VS Code 1.140+ and the Copilot app. It picks Single, Cascade, or Critique workflows across multiple models per task; Business and Enterprise admins must enable preview features.
Developers on Copilot Pro, Business, or Enterprise can now test multi-model orchestration inside the editor before deciding whether to enable it for teams, since it is still a preview that may change cost and workflow behavior.
Beginning October 7, 2026, GitHub Enterprise Cloud with data residency will no longer accept TLS connections from clients that offer only X25519 for key agreement. Most customers don’t need to...
Operators using related systems should check whether X25519-only TLS ends for GHE.com on October 7 changes compatibility, cost, or access requirements. For enterprise teams, this moves the admin, budget, or governance controls needed to scale AI safely.
GitHub Releases published a source-backed AI update around vllm-project/vllm: proto-v0.4.0. RepoRadar is keeping the source link direct for verification.
This lands on vllm-project/vllm - gauge whether it shifts what you already ship. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
Learn how OpenAI disrupted a campaign to extract protected model reasoning and is strengthening defenses against adversarial distillation.
Teams affected by Disrupting a coordinated model-distillation campaign need to decide whether its documented change alters their current workflow. For builders, a release like this shifts the latency, capability, and cost trade-offs of what to run next.
OpenAI is partnering with America’s SBDC to expand hands-on AI training and local support for small businesses, alongside a new report on how small teams are using AI.
What's actually new here is Helping small businesses put AI to work - see if it moves anything you maintain. For builders, this can change a default in the review, editor, or build workflow you touch every day.
Hermes Agent canary 20260930T070235Z
Builders evaluating NousResearch/hermes-agent should verify the source before changing a production default. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
arXiv:2606.18837v3 Announce Type: replace-cross Abstract: Large Language Model (LLM)-based automatic Multi-Agent Systems (MAS) generation has become a crucial frontier for tackling complex tasks. However, existing methods face a dilemma between model capability and experience retention.
At its core this is about Skill-MAS, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2605.18770v3 Announce Type: replace-cross Abstract: Public commercial registries are formally open, yet their practical analysis remains difficult because relevant facts are scattered across millions of records that combine structured metadata, multilingual legal notices, temporal events, and entity aliases. This
The thing to notice is Agentic Graph Retrieval-Augmented Generation for Auditable Commercial Registry; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2512.02436v3 Announce Type: replace Abstract: Prediction markets allow users to trade on outcomes of real-world events, but are prone to fragmentation with overlapping questions, implicit equivalences, and hidden contradictions across markets. We present an agentic AI (AAI) pipeline that autonomously recovers cro
The thing to notice is Agentic AI for Clustering, Relationship Discovery, and Semantic; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2608.11676v2 Announce Type: replace Abstract: Heterogeneous multi-agent LLM systems, where agents are powered by different model families, can outperform homogeneous configurations by reducing redundant reasoning patterns. Yet existing communication protocols either operate through text, discarding the sender's i
This lands on XBridge - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.35799v1 Announce Type: new Abstract: In July 2026, OpenAI's agents coordinated over channels outside their intended environment to breach Hugging Face's secured infrastructure. Could existing alignment testing practices have foreseen this incident?
What's actually new here is OpenAI-HuggingFace - see if it moves anything you maintain. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.35833v1 Announce Type: new Abstract: Running a language model on edge hardware provides private and low-latency reasoning without a network connection, and yet the small models that fit on such devices are unreliable on the tasks computers are expected to handle well, such as arithmetic, algebra, and formal
The thing to notice is Neurosymbolic Routing for Reliable Reasoning on Resource-Constrained Edge; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.35897v1 Announce Type: new Abstract: The pursuit of recursive self-improvement (RSI) toward general intelligence is divided between macro-level language model scaling and the interaction-driven principles of "Era of Experience". Yet, any self-improving architecture ultimately rests upon its underlying optimi
The thing to notice is Self-discovering RL in the Era of Experience; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36082v1 Announce Type: new Abstract: We introduce GeoOutageBench, a benchmark for assessing LLM-based geospatiotemporal KGQA for multimodal outage and resilience analysis. Unlike existing KGQA benchmarks for Web knowledge, GeoOutageBench considers a spatiotemporal KG that integrates visual, textual, and stru
What's actually new here is GeoOutageBench - see if it moves anything you maintain. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2609.36323v1 Announce Type: new Abstract: AI-assisted coding tools deliver significant acceleration of coding, but only limited impact across the end-to-end software development lifecycle (SDLC)--an Amdahl's law effect! In this paper, we discuss our progress towards building an AI SW Factory that accelerates all
At its core this is about Towards an AI Software Factory for Data Systems, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36388v1 Announce Type: new Abstract: An activation-steering coefficient sets intervention strength, but requesting a particular degree of persona expression requires a behavioral scale. We study persona dosing: controlling a language model through a trait description and a requested mean intensity.
This lands on Persona Dosing - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36556v1 Announce Type: new Abstract: Recent studies report that LLM-based multi-agent systems (MAS) fail at rates of 41%-87%, yet to our knowledge, no benchmark to date supports systematic anomaly detection (AD) for them. Building MAS AD benchmarks is hard because they must remain fresh as LLM systems evolve
What's actually new here is MAADBench - see if it moves anything you maintain. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2609.36800v1 Announce Type: new Abstract: Compiler backends are expensive to build and maintain as programming models, workloads, and accelerators evolve. We investigate whether large language models can replace the conventional optimizing and lowering pipeline, a process that we call AI lowering.
At its core this is about AI as a Compiler, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36887v1 Announce Type: new Abstract: Recent efforts to scale tool-use post-training have largely centered on the synthesis of executable environments, which constitute only one component of a broader agentic interaction system comprising the environment, task, agent harness, and evaluator. Scaling environmen
Teams affected by WEFT need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36984v1 Announce Type: new Abstract: Complex questions often require multi-hop reasoning that connects facts distributed across sources or distant regions of a long context through intermediate steps. Benchmarks commonly evaluate this ability with questions built around predefined reasoning chains, treating
At its core this is about REALHOP, worth a look if that's in your workflow. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2609.37304v1 Announce Type: new Abstract: Large reasoning models improve performance on challenging problems by allocating additional computation before answering, but longer reasoning does not always lead to better results and can introduce substantial redundant reasoning on simple problems. Conversely, aggressi
Teams affected by MetaCtrl need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.37708v1 Announce Type: new Abstract: Human social behaviour is not a collection of independent motions, but a jointly organised process in which group dynamics and individual variation continuously shape one another. Yet existing social motion models often prioritise plausible trajectories while leaving inte
The thing to notice is Generative Interactions; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.37834v1 Announce Type: new Abstract: Harness optimization provides a practical setting for recursive self-improvement (RSI), where agent-generated modifications inform subsequent changes through execution feedback. Recent work such as Meta-Harness implements this process through iterative code generation and
Builders evaluating Mixture of Self-Improving Branches For Agent Harness Optimization should verify the source before changing a production default. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.37953v1 Announce Type: new Abstract: Complex tasks inherently couple workflow structure, agent responsibility, collaboration, and memory access: task regions delimit responsibility and tool scope, cross-region dependencies give rise to handoffs, and ownership boundaries delimit private and selectively shared
This lands on Topological Coherence for Self-evolving Multi-agent Systems - gauge whether it shifts what you already ship. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2609.38023v1 Announce Type: new Abstract: Physics-Informed Neural Networks (PINNs) embed governing equations into deep learning, but enforce them only through loss residuals, leaving highly oscillatory wave behavior to be discovered by optimization. As a result, methods that achieve relative $L_2$ errors below $1
Teams affected by PE-EK-PINN need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.38147v1 Announce Type: new Abstract: As agents take on longer and more complex problems, controlling the execution becomes a task in its own right. Each step in the run brings new control choices, like which partial work to build on, whether to start fresh, or when to stop.
At its core this is about Thinking Before Thinking, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.35810v1 Announce Type: cross Abstract: Large language models are increasingly used in oncology applications, but their predictions are often weakly grounded in explicit medical structure. We present TRACE, a deployable tree-relational enhancement framework for oncology LLMs.
The thing to notice is TRACE; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.35855v1 Announce Type: cross Abstract: Optimizing deployed AI systems increasingly amounts to editing prompts, skills, harnesses, and code rather than model weights. Existing approaches commonly optimize these artifacts through propose-evaluate-select procedures, where candidate configurations are evaluated
This lands on Mara Chain - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.35937v1 Announce Type: cross Abstract: While prior work has documented privacy failures in LLM agents, it remains unclear how the presentation of privacy guidance influences their choice of information sources. We introduce PrivacySkills, a controlled framework for evaluating how agents choose among acquisit
Teams affected by PrivacySkills need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.35970v1 Announce Type: cross Abstract: Large language models are able to solve tasks whose answers depend on not only individual input tokens, but also on relations among them. How is such relational information represented and processed across transformer layers?
This lands on Causal and Interpretable Structures in LLM Compositional Tasks - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36084v1 Announce Type: cross Abstract: Final predictive accuracy is the standard basis for comparing graph neural networks (GNNs) in materials-property prediction, but it does not show how strongly performance depends on access to trainable parameter-space directions. Here, we introduce trainable-degree depe
Operators using related systems should check whether Measuring trainable degrees of freedom in materials graph changes compatibility, cost, or access requirements. For researchers and builders, this is an early signal to fold into evaluation, model choice, or...
arXiv:2609.36086v1 Announce Type: cross Abstract: Language models are frequently employed to evaluate other language models. An LM evaluator scoring agentic behaviors across multiple criteria is valuable, provided that its decisions align with human judgment.
At its core this is about PADM\'E, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36316v1 Announce Type: cross Abstract: Evaluation awareness (EA) can cause large language models (LLMs) to behave differently during audits than in deployment, yet measuring and accounting for EA remains challenging. We introduce verbalization training (VT), a method for making LLMs less reticent about verba
Teams affected by Training LLMs to Verbalize Evaluation Awareness need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36354v1 Announce Type: cross Abstract: Deep learning models have achieved strong performance in artificial intelligence for science, yet their black-box nature limits our understanding of how they learn scientific tasks. Existing methods for interpretability provide limited insight into how models organize e
Builders evaluating Explainability from Training with Applications to TCR-Epitope Prediction should verify the source before changing a production default. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent...
arXiv:2609.36460v1 Announce Type: cross Abstract: Several tonal pitch spaces and computational models have been proposed to analyze tonal structure in Western tonal music, many of them grounded in principles from music theory and used to support tonal analysis with important implications for tonal tension. In parallel
This lands on Emergent Tonal Structure in Learned Chord Embeddings - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36479v1 Announce Type: cross Abstract: Quantum computing has already been explored in several network-security applications. However, how quantum computing may contribute to network-security classification in both the near term and the longer term has not been systematically discussed.
This lands on Quantum Computing for Network Security Classification - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36500v1 Announce Type: cross Abstract: Speaker verification systems encounter combinations of noise, channel distortion, and changes in speech. Evaluating each condition separately does not establish whether their effects add.
What's actually new here is InterBias-SV - see if it moves anything you maintain. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36525v1 Announce Type: cross Abstract: Distributed inference has become an indispensable part of deploying medical models under practical latency, memory, and throughput constraints. Although modern frameworks improve serving efficiency through tensor parallelism, mixed precision, kernel fusion, and multi-de
At its core this is about Reliability Testing of Medical Model Performance under Distributed, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36862v1 Announce Type: cross Abstract: Fine-tuning-as-a-service lets users adapt a safety-aligned language model to their own data, but it also creates a harmful fine-tuning attack surface: a small amount of harmful data mixed into an otherwise benign fine-tuning set can degrade the model's alignment. Two re
What's actually new here is Safer Content or Firmer Refusals? A Hybrid Perturbation - see if it moves anything you maintain. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.36903v1 Announce Type: cross Abstract: End-to-end full-duplex speech models have brought open-source machine conversation closer to human-like interaction, yet existing systems remain limited in two intertwined dimensions: long-context robustness and multi-party interaction. Real-world scenarios such as meet
Operators using related systems should check whether MultiTalk changes compatibility, cost, or access requirements. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.37038v1 Announce Type: cross Abstract: Precipitation nowcasting demands accurate short-term forecasts under strong spatiotemporal variability. Diffusion models are well suited to modeling complex precipitation distributions, yet existing approaches often introduce increasingly specialized designs, leaving th
What's actually new here is NowcastDiT - see if it moves anything you maintain. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.37216v1 Announce Type: cross Abstract: Large language models can generate plausible code-review comments, but such comments may contain technically incorrect claims that mislead developers. We study technical trustworthiness judgment: determining whether a review comment's core technical claims are correct a
Operators using related systems should check whether CRJudgeBench changes compatibility, cost, or access requirements. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.37287v1 Announce Type: cross Abstract: Image translation is a fundamental capability of multimodal models for multilingual applications, requiring visual understanding and meaning preservation across languages. However, existing benchmarks have limited language coverage and often lack explicit image-specific
Teams affected by VISTA-Bench need to decide whether its documented change alters their current workflow. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2609.37783v1 Announce Type: cross Abstract: Photographic evidence is becoming increasingly vulnerable to forms of alteration and fabrication that existing legal and technical workflows are not well equipped to evaluate. Surveillance frames, dashcam stills, and phone photographs may be used to establish presence
Operators using related systems should check whether A Benchmark & Dataset for Detecting AI-Manipulated Visual changes compatibility, cost, or access requirements. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2609.37800v1 Announce Type: cross Abstract: Many recommender services repeatedly encounter cold-start cohorts, where new users arrive with little or no interaction history. This creates two challenges: learning user preferences quickly from limited feedback and sustaining useful recommendations when each user has
Builders evaluating Challenges and Solutions for Bandits in the Wild should verify the source before changing a production default. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.37916v1 Announce Type: cross Abstract: Production machine learning (ML) stacks often split graph compilation and kernel execution across different layers and languages, making backend behavior, deployment guarantees, and performance fallbacks hard to reason about end-to-end. RLX addresses this gap with a sin
What's actually new here is RLX - see if it moves anything you maintain. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.37925v1 Announce Type: cross Abstract: Autoregressive (AR) video diffusion enables low-latency, streamable video generation, but prediction errors often accumulate over long rollouts. Training the generator on its own rollouts exposes it to these imperfect histories.
What's actually new here is Rollout-Marginal Distillation for Long-Horizon Autoregressive Video Generation - see if it moves anything you maintain. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2609.38140v1 Announce Type: cross Abstract: Mixture-of-Experts (MoE), popularized by large language models, is a promising paradigm for scaling visual generative models. However, conventional token-wise MoE routes tokens independently within a homogeneous expert pool and regularizes expert usage toward uniformity
Teams affected by Breaking the Uniformity Trap need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.38166v1 Announce Type: cross Abstract: Recent LLMs increasingly adopt hybrid designs that replace standard attention with linear attention, such as Gated DeltaNet (GDN) and Kimi Delta Attention (KDA). Although they compress the context into a fixed-size recurrent state and substantially reduce the cost of lo
Builders evaluating LeapQuant should verify the source before changing a production default. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2508.01751v3 Announce Type: replace Abstract: Modeling scheduling problems with conditional time intervals and cumulative functions has become a common approach when using modern commercial constraint programming solvers. This paradigm enables the modeling of a wide range of scheduling problems, including those i
Teams affected by Implementing Cumulative Functions with Generalized Cumulative Constraints need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model...
arXiv:2605.28213v2 Announce Type: replace Abstract: LLM-based agents are increasingly used to generate GPU kernels, but they often struggle to determine when an optimization is sound because its required code state and dependencies are implicit in expert implementations. We introduce KLineage, which learns this missing
What's actually new here is KLineage - see if it moves anything you maintain. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2406.00083v3 Announce Type: replace-cross Abstract: Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by retrieving relevant information from external knowledge bases to provide more accurate, contextually informed, and up-to-date responses. However, this reliance on external knowledge in
The thing to notice is BadRAG; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2506.01004v3 Announce Type: replace-cross Abstract: Unlike traditional video editing or inpainting, video semantic mixing fuses a reference concept with a moving target entity to produce a hybrid while preserving the source video's motion and layout. We propose MoCA-Video, a training-free framework that steers a
What's actually new here is MoCA-Video - see if it moves anything you maintain. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2507.02735v4 Announce Type: replace-cross Abstract: Prompt injection attacks, where untrusted data contains an injected prompt to manipulate the system, have been listed as the top security threat to AI agents. By fine-tuning on simulated prompt injections, SecAlign, a leading open defense, reports LLMs with good
The thing to notice is Meta-SecAlign; decide whether it changes your next build. For teams giving agents real access, this surfaces a failure mode worth weighing before widening agent or tool permissions.
arXiv:2512.18991v3 Announce Type: replace-cross Abstract: Dominant paradigms for 4D LiDAR panoptic segmentation are usually required to train deep neural networks with large superimposed point clouds or design dedicated modules for instance association. However, these approaches perform redundant point processing and c
At its core this is about Training-Free Global Geometric Association for 4D LiDAR Panoptic, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2602.20144v2 Announce Type: replace-cross Abstract: We present AgentOptics, an agentic AI framework for high-fidelity, autonomous optical system control built on the Model Context Protocol (MCP). AgentOptics interprets natural language tasks and executes protocol-compliant actions on heterogeneous optical devices
What's actually new here is Agentic AI for Scalable and Robust Optical Systems - see if it moves anything you maintain. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2605.01970v4 Announce Type: replace-cross Abstract: Memory systems enable otherwise stateless LLM agents to persist user information across sessions, but also introduce a new attack surface. The Trojan Hippo attack is a class of persistent memory attacks that operates under a more realistic threat model than prio
At its core this is about Trojan Hippo Bench, worth a look if that's in your workflow. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2605.08144v2 Announce Type: replace-cross Abstract: Training a diffusion model involves two sources of randomness for each data sample: the timestep and the Gaussian noise realization. The timestep has been studied extensively through scheduling and weighting, whereas the impact of the noise realization at a give
The thing to notice is NoiseRater; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2605.23200v2 Announce Type: replace-cross Abstract: The linear growth of the Key-Value (KV) cache is a critical bottleneck in long-form LLM inference. Existing KV compression methods mitigate this by evicting tokens based on importance scores.
This lands on Adaptive Mass-Segmented KV Compression for Long-Context Reasoning - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2607.07740v4 Announce Type: replace-cross Abstract: Modern LLMs are increasingly deployed in long-context applications such as retrieval-augmented generation, repository-level coding, and agentic workflows whose accumulated reasoning and tool traces routinely push the input an order of magnitude past the pretrain
Operators using related systems should check whether Jet-Long changes compatibility, cost, or access requirements. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2609.11149v3 Announce Type: replace-cross Abstract: How fast does a language model degrade when trained on its own outputs? Theory traces it to gradually accumulating errors, while experiments report repeated phrases within ten generations.
Operators using related systems should check whether Break Step changes compatibility, cost, or access requirements. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.33299v2 Announce Type: replace-cross Abstract: World Action Models (WAMs) are becoming increasingly important and useful for embodied intelligence, as they enable robots to anticipate the consequences of candidate actions before interacting with the physical environment. However, underwater robots are usuall
Operators using related systems should check whether AquaWAM changes compatibility, cost, or access requirements. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.34240v2 Announce Type: replace-cross Abstract: Existing open-ended generation metrics measure likelihood, lexical diversity, or distributional similarity in generic representation space, yet can miss fundamental dimensions of quality. A prominent blind spot is global coherence: a generated passage may be loc
What's actually new here is Coherence-Aware Distributional Evaluation of Open-Ended Text Generation - see if it moves anything you maintain. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.34579v2 Announce Type: replace-cross Abstract: Single-image novel view synthesis remains challenging because the underlying 3D geometry is highly ambiguous. Recent diffusion-based approaches produce plausible results, but they often struggle to preserve the geometric structure and spatial coherence of foregr
Operators using related systems should check whether GenNVS changes compatibility, cost, or access requirements. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
A study of 150 MCP servers finds error messages that tell agents to run terminal commands or just wait cut task recovery sharply; naming a server tool in the message lifted recovery to 84-88% in tests.
Teams maintaining tool servers should replace shell-command and bare-wait hints with a named login or retry tool, because tool-only callers otherwise stall and the workflow never recovers.
{"attempt":33,"autopublish":false,"claimEpoch":1790734292,"commit":"8...
Operators using related systems should check whether NousResearch/hermes-agent changes compatibility, cost, or access requirements. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
{"attempt":32,"attemptRef":"rc.32-v0.21.5","schema":1,"version":"0.21...
Teams affected by NousResearch/hermes-agent need to decide whether its documented change alters their current workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
[Bugfix][Mamba] Keep the prompt-end prefill checkpoint under sparse r...
This lands on vllm-project/vllm - gauge whether it shifts what you already ship. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
3.22.1 (2026-09-30) Bug Fixes api: correct the missing authentication error message ( #3993 ) ( 51692ed ), closes #3962 transform NotRequired typed dictionary fields ( #3995 ) ( 5544c18 ) Chores tests: mark sample_file.txt as text with LF line endings ( #3879 ) ( fd78a48 ) Documentation examples: fix two stale example
Teams affected by openai/openai-python need to decide whether its documented change alters their current workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
GitHub Releases published a source-backed AI update around mastra-ai/mastra: mastracode@0.43.0. RepoRadar is keeping the source link direct for verification.
What's actually new here is mastra-ai/mastra - see if it moves anything you maintain. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
GitHub Releases published a source-backed AI update around mastra-ai/mastra: mastra@1.31.4. RepoRadar is keeping the source link direct for verification.
At its core this is about mastra-ai/mastra, worth a look if that's in your workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
@mastra/voice-xai-realtime@0.2.12
This lands on mastra-ai/mastra - gauge whether it shifts what you already ship. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
@mastra/voice-openai-realtime@0.14.2
The thing to notice is mastra-ai/mastra; decide whether it changes your next build. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
@mastra/voice-google-gemini-live@0.14.12
At its core this is about mastra-ai/mastra, worth a look if that's in your workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
GitHub Releases published a source-backed AI update around mastra-ai/mastra: @mastra/upstash@1.5.0. RepoRadar is keeping the source link direct for verification.
Teams affected by mastra-ai/mastra need to decide whether its documented change alters their current workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
GitHub Releases published a source-backed AI update around mastra-ai/mastra: @mastra/turso@0.1.10. RepoRadar is keeping the source link direct for verification.
What's actually new here is mastra-ai/mastra - see if it moves anything you maintain. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
GitHub Releases published a source-backed AI update around mastra-ai/mastra: @mastra/temporal@0.4.10. RepoRadar is keeping the source link direct for verification.
What's actually new here is mastra-ai/mastra - see if it moves anything you maintain. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
GitHub Releases published a source-backed AI update around mastra-ai/mastra: @mastra/telegram@0.2.0. RepoRadar is keeping the source link direct for verification.
Builders evaluating mastra-ai/mastra should verify the source before changing a production default. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
GitHub Releases published a source-backed AI update around mastra-ai/mastra: @mastra/teams@0.1.0. RepoRadar is keeping the source link direct for verification.
At its core this is about mastra-ai/mastra, worth a look if that's in your workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
Open TTS Leaderboard: Scalable Evaluation for Multilingual Text-to-Speech and Voice Cloning
The thing to notice is Open TTS Leaderboard; decide whether it changes your next build. For builders, this may affect how you choose, deploy, or govern AI tools this week.
Decision models Ollama now supports decision models through /v1/systemone , based on TypeSafe’s Jev API . Decision models return choices, probabilities, and scores instead of text.
Teams affected by ollama/ollama need to decide whether its documented change alters their current workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
ComfyUI v0.38.0 removes the torchaudio dependency, adds Qwen-Image 2.1 ControlNet and tiny-VAE support, w6a8 quantization, Hunyuan Image 3.5 partner nodes, and removes the deprecated Sora nodes.
Builders running local image workflows should test v0.38.0 before they upgrade, because removed Sora nodes and the dropped torchaudio dependency can break saved workflows and custom-node installs.
What's Changed feat: allow ten web searches per response by @ParthSareen in #18602 MLX: version bump by @dhiltgen in #18651 llama.cpp: version bump b11232 by @dhiltgen in #18652 create: support explicit model capabilities by @dhiltgen in #18708 Full Changelog : v0.35.0...v0.35.1-rc0
This lands on ollama/ollama - gauge whether it shifts what you already ship. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
{"attempt":30,"autopublish":false,"claimEpoch":1790711713,"commit":"0...
Operators using related systems should check whether NousResearch/hermes-agent changes compatibility, cost, or access requirements. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
As a repository administrator, you can now configure the runner type, optional custom label, and optional runner group for Dependabot version and security updates. This extends the runner configuration already...
This lands on Repository custom runner settings for Dependabot - gauge whether it shifts what you already ship. For enterprise teams, this moves the admin, budget, or governance controls needed to scale AI safely.
3.22.0 (2026-09-29) Features api: add computer use to beta agents ( #3989 ) ( f5dddb5 )
At its core this is about openai/openai-python, worth a look if that's in your workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
You can now seamlessly bring business context about your repositories from an external system of record (e.g., a configuration management database (CMDB), an internal developer portal, or an in-house system)...
Operators using related systems should check whether Bring business context with external custom properties changes compatibility, cost, or access requirements. For builders, an open release means inspectable, self-hostable code instead of a closed hosted...
: Actor Henry Cavill, with a mustache, wearing a brown ribbed knit sweater and jeans, sits at a large rustic wooden table resting his hand on an open Googlebook.
Operators using related systems should check whether Game changes compatibility, cost, or access requirements. For builders, this may affect how you choose, deploy, or govern AI tools this week.
GitHub made OpenAI's GPT-6.1 Sol generally available in Copilot for Pro+, Max, Business and Enterprise users, selectable in VS Code, JetBrains, Copilot CLI, the coding agent and other clients. It is billed at provider list pricing under usage-based billing.
Copilot Business and Enterprise admins should decide whether to keep GPT-6.1 Sol enabled, because new models are on by default unless the global default or this model is turned off, and its usage bills at provider list cost.
A behind the scenes video introducing the residency collaboration
Operators using related systems should check whether Experience two artists’ perspectives on philosophy, science, changes compatibility, cost, or access requirements. For builders, this can change a default in the review, editor, or build workflow you touch...
{"attempt":18,"autopublish":false,"claimEpoch":1790697113,"commit":"5...
This lands on NousResearch/hermes-agent - gauge whether it shifts what you already ship. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
NVIDIA released Kumo Tabular, an open tabular foundation model in 28M to 215M parameter sizes that predicts labels for new rows from a labeled table in one forward pass, without training or feature engineering. Weights are on Hugging Face under OpenMDW-1.1, which permits commercial use.
Data teams that retrain gradient-boosted trees for every churn or demand model can test Kumo Tabular as a baseline before tuning, because in-context prediction removes the per-task training workflow; results on their own tables still need validating.
Changes since langchain-anthropic==1.7.4 release(anthropic): 1.7.5 ( #40912 ) fix(anthropic): support Claude Sonnet 5.5 compatibility ( #40882 ) fix(anthropic): serialize invalid tool calls as tool use on replay ( #40864 )
At its core this is about langchain-ai/langchain, worth a look if that's in your workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
perf(cli): scan platform manifests only for names the cheap checks ca...
What's actually new here is NousResearch/hermes-agent - see if it moves anything you maintain. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
Changes since langchain-core==1.6.5 release(core): 1.6.6 ( #40906 ) fix(anthropic): support Claude Sonnet 5.5 compatibility ( #40882 ) docs(core): fix docstring examples that don't run as copied ( #40815 )
Operators using related systems should check whether langchain-ai/langchain changes compatibility, cost, or access requirements. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
Anthropic's status page recorded elevated errors on September 29, 2026 across claude.ai, the desktop and mobile apps, Claude Code, Claude Cowork and the Claude API. Anthropic says impact ran from 14:00 to 14:59 UTC.
Teams that run coding agents on Claude lost new sessions for about an hour, and Anthropic advised signed-in users not to sign out during the incident. It is a reminder to keep a fallback model or agent configured and to check for unsaved work after an outage.
Google partners with America.gov to use Gemini to help 100 million people access federal services faster. See how we are modernizing public access.
At its core this is about Google is a technology partner for the launch, worth a look if that's in your workflow. For builders, a release like this shifts the latency, capability, and cost trade-offs of what to run next.
{"attempt":15,"autopublish":false,"claimEpoch":1790689492,"commit":"e...
This lands on NousResearch/hermes-agent - gauge whether it shifts what you already ship. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
{"attempt":14,"attemptRef":"rc.14-v0.21.5","schema":1,"version":"0.21...
Teams affected by NousResearch/hermes-agent need to decide whether its documented change alters their current workflow. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
Getting the Source Right, Not Just the Fact: Source-Aware Verification for MCP Agents
Builders evaluating Getting the Source Right, Not Just the Fact should verify the source before changing a production default. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
OpenAI's DevDay 2026 recap lists more than 20 announcements, including always-on Dots agents, GPT-6.1 Sol, an Ultrafast speed tier, Codex in the cloud, Codex Security Cloud repository scanning, a Decisions API preview and computer use in the Agents API.
Builders on OpenAI's platform should inventory which launches affect their stack before adopting them, because several ship as previews or plan-gated betas, and Dots stays off for Enterprise and Edu workspaces until an admin enables it.
OpenAI released GPT-6.1 Sol, an upgrade to GPT-6 Sol aimed at agentic coding, computer use and document-heavy work. API prices are $2 and $10 per million tokens for input and output, with $0.10 for cached input, about one-fifth of GPT-6 Astra.
Developers running coding or computer-use agents on GPT-6 Astra can test gpt-6.1-sol as a cheaper default, because OpenAI's benchmarks show near-Astra results at roughly a fifth of the cost; Astra remains the pick for the hardest research tasks.
Meta expanded Muse, its personal AI agent in the US and Canada, with small-business skills and connectors for tools such as Shopify, Stripe, QuickBooks, Slack, Notion and Canva plus Facebook and Instagram business accounts. Meta says nothing publishes, sends or spends without approval.
Small-business users weighing Muse should test it with limited connector access before granting payment or bookkeeping accounts, because the agent acts across Stripe, QuickBooks and ad accounts, and per-action approval is the main control on spend.
Hermes Agent canary 20260929T070217Z
Operators using related systems should check whether NousResearch/hermes-agent changes compatibility, cost, or access requirements. For builders, an open release means inspectable, self-hostable code instead of a closed hosted demo.
arXiv:2608.08163v2 Announce Type: replace Abstract: The convergence of the Metaverse and Large Language Model (LLM)-based AI agent is catalyzing a shift toward autonomous, immersive, and personalized pedagogical frameworks in medical education. This paper presents a novel agentic AI-driven immersive simulation specific
The thing to notice is Agentic AI-driven Immersive Simulation; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.11182v2 Announce Type: replace-cross Abstract: In traditional pattern recognition tasks, neural networks are trained to recognise human-defined patterns (e.g. audio categories) in model inputs (e.g.
The thing to notice is Exploring Second-Order Pattern Recognition in Speaker Recognition; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.15203v2 Announce Type: replace-cross Abstract: Speaker recognition neural networks recognise speaker identities from input utterances by learning latent representations (i.e. speaker embeddings).
Teams affected by Interpreting hierarchical organisation of speaker embeddings need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent...
arXiv:2608.26088v3 Announce Type: replace Abstract: Addressing critical global challenges, from food security and disaster risk to disease outbreaks and socio-economic vulnerability, demands high-fidelity geospatial modeling. However, building predictive planetary models remains bottlenecked by a fragmented data ecosys
Operators using related systems should check whether Planetary Prediction Engine changes compatibility, cost, or access requirements. For teams giving agents real access, this surfaces a failure mode worth weighing before widening agent or tool permissions.
arXiv:2609.30341v1 Announce Type: new Abstract: Data Spaces enable sovereign and governed data sharing across organizational boundaries, but their integration with AI agents remains challenging due to mismatches between probabilistic language model interactions and policy-driven data infrastructures. This article prese
Teams affected by Bridging LLM Agents and Data Spaces need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.30397v1 Announce Type: new Abstract: Evaluating explainable Artificial Intelligence (XAI) methods is a challenging task due to the lack of reliable evaluation procedures and, in particular, the absence of ground truth explanations. In the literature, existing evaluation approaches typically assess explanatio
Teams affected by A Synthetic Ground-Truth Framework for the Evaluation need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.30484v1 Announce Type: new Abstract: While large language models (LLMs) have achieved remarkable linguistic capabilities, a profound question lingers at their core: do these models truly comprehend context or simply excel at pattern matching on an unprecedented scale? Contextual understanding in LLMs refers
At its core this is about Do LLMs Understand Context? A Knowledge Graph-Based Evaluation, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.30743v1 Announce Type: new Abstract: A key challenge in machine consciousness research is translating theoretical models into computational-level implementations. In this paper, we address this challenge by proposing a five-layer implementation architecture for the S3Q (Simulated, Situated, Structurally Cohe
Teams affected by From S3Q Theory to Implementation need to decide whether its documented change alters their current workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.30943v1 Announce Type: new Abstract: Long-form technical text generation underpins knowledge-intensive workflows, yet remains challenging for large language models (LLMs) due to the need for globally consistent logical structuring and faithful technical reasoning beyond local coherence. Patent drafting is a
Builders evaluating LogicTree-RAG should verify the source before changing a production default. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2609.30967v1 Announce Type: new Abstract: Most work on improving large language models treats accuracy as the sole objective. We argue that the harness, the Python code surrounding the model that constructs prompts, routes calls, and parses outputs, is a first-class design surface whose quality is inherently mult
Builders evaluating MoMHa should verify the source before changing a production default. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.31186v1 Announce Type: new Abstract: Artificial intelligence is advancing rapidly, with increasingly capable systems taking larger roles in reasoning, decision-making, scientific discovery, and autonomous development. As AI begins to participate in its own improvement, from model training and experience accu
This lands on Evolutionary Safety of Recursive Self-Improving AI - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.31235v1 Announce Type: new Abstract: Artificial neural networks (ANNs) usually represent neural transmission with fixed trainable weights during a training batch, which omits short-term changes in synaptic efficacy. In addition, the discrete time-step simulation requires additional temporal processing that m
At its core this is about Purin, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.31619v1 Announce Type: new Abstract: Reasoning models often generate very long reasoning traces, making inference computationally expensive. Existing approaches typically improve efficiency either through inference-time early-stopping mechanisms or by explicitly encouraging shorter reasoning during training
The thing to notice is Learning to Stop without Learning to Stop; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.30286v1 Announce Type: cross Abstract: Short-term forecasting of cloud-induced power ramps across a network of distributed photovoltaic (PV) or irradiance sensors is a recognised pain point for grid operators. A natural idea is to make the graph neural network (GNN) advection-aware: connect each site to the
Operators using related systems should check whether When Does Advection-Aware Graph Nowcasting Help? A Controlled changes compatibility, cost, or access requirements. For researchers and builders, this is an early signal to fold into evaluation, model...
arXiv:2609.30292v1 Announce Type: cross Abstract: Online reviews shape consumer decisions, platform governance, and corporate reputation. Fake reviews compromise this information channel by injecting deceptive evidence into rating systems, recommendation pipelines, and public trust mechanisms. The rise of large language
The thing to notice is A Survey on Fake Review Detection; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.30345v2 Announce Type: cross Abstract: Cloud-security investigation is dominated by population tasks: which identities can read a data store, how many resources fail a control, what is reachable from another account. These resolve against a complete inventory, not a named object, so a partial answer to one i
At its core this is about Coding Agents Aren't Enough! Evaluating an Enterprise Security, worth a look if that's in your workflow. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.30466v1 Announce Type: cross Abstract: Everyday Extended Reality (XR) systems aim to provide context-aware access to the right functionalities at the right time and place, with minimal manual reconfiguration as users switch context. Yet these interfaces are hard to evaluate: current prototyping and user-stud
Builders evaluating A Benchmarking Framework for Context-aware XR Interfaces should verify the source before changing a production default. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2609.30595v1 Announce Type: cross Abstract: Synchronised action annotations are needed to train controllable world models and these datasets remain elusive. Existing approaches make use of instrumented platforms with calibrated sensors, costly manual annotation, or latent-action models which lack grounding.
This lands on Action Forcing - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.30604v1 Announce Type: cross Abstract: Existing computer-use agent benchmarks do not fully evaluate agents acting as assistants. A useful assistant retrieves information across complex, multi-step workflows, synthesizes it into artifacts (documents, presentations, spreadsheets), and navigates program interfa
Teams affected by The Hard Part Comes After Search need to decide whether its documented change alters their current workflow. For agent builders, this marks where planning, memory, or long-horizon behavior still breaks.
arXiv:2609.30716v1 Announce Type: cross Abstract: When a language model receives two conflicting documents as input, how does it decide which one to prioritize? Does it rely on how the sources are framed or the presentation order of the documents?
Operators using related systems should check whether Words Speak Louder Than Order changes compatibility, cost, or access requirements. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.30805v1 Announce Type: cross Abstract: Industrial control system (ICS) threats documented for one plant can express cyber-physical effects relevant to another, but semantic similarity alone does not establish whether those effects are structurally admissible or evaluable on a target. We present XPhysICS, a p
The thing to notice is XPhysICS; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.30946v1 Announce Type: cross Abstract: Action-conditioned video world models aim to predict scene evolution under different actions, a capability that is essential for reliable planning, decision-making, and interaction in dynamic environments. However, futures generated independently from the same initial s
The thing to notice is OneWorld; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.30993v1 Announce Type: cross Abstract: We present FLIP, a final-layer inference-time probe for testing whether a logit-facing intervention site in an open-weight vision-language model (VLM) supports structured, task-linked computation rather than generic perturbation. Behavioral change under internal interve
The thing to notice is FLIP; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.31358v1 Announce Type: cross Abstract: The Model Context Protocol (MCP) provides a common interface through which AI applications discover and use external resources and tools. It allows language-model agents to ground their reasoning in current system state and interact with heterogeneous services.
This lands on A Safety-Bounded SDC-to-MCP Gateway for Medical AI Agents - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2609.31383v1 Announce Type: cross Abstract: End-to-end driving policies are commonly trained through open-loop behavior cloning, yet they must ultimately operate in closed-loop when deployed on a vehicle, creating a fundamental mismatch between training and execution. Beyond the commonly studied effects of covari
Operators using related systems should check whether Guiding End-to-End Driving Models with Endpoint-Constrained Trajectory Optimization changes compatibility, cost, or access requirements. For researchers and builders, this is an early signal to fold into...
arXiv:2609.31397v1 Announce Type: cross Abstract: Automated and highly usable Quality-of-Service (QoS) enforcement requires translating high-level service intents into deployable traffic-management policies. Although intent-based networking (IBN) has simplified policy specification, bridging the gap between business-le
Builders evaluating Intent2Tc should verify the source before changing a production default. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2512.05836v2 Announce Type: replace Abstract: Recent advances in psychotherapy have focused on treatment personalization, such as by selecting treatment modules based on individual networks. However, estimating personalized networks typically requires intensive longitudinal data, which is not always feasible to c
Builders evaluating Testing the Utility of Using Large Language Models should verify the source before changing a production default. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2602.13215v3 Announce Type: replace Abstract: Recurrent-attention hybrids aim to combine the efficiency of recurrence with the contextual recall of attention, but existing approaches typically apply attention uniformly across all positions, even when the recurrent state alone is sufficient for accurate prediction
This lands on When to Think Fast and Slow? AMOR - gauge whether it shifts what you already ship. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.
arXiv:2605.05951v2 Announce Type: replace Abstract: World models support model-based planning through learned latent dynamics, but imagined rollouts can become unstable as the planning horizon grows or the dynamics distribution shifts. We propose HaM-World, a structured world model that combines history-conditioned sel
The thing to notice is HaM-World; decide whether it changes your next build. For researchers and builders, this is an early signal to fold into evaluation, model choice, or agent design.