Search: “gui”

JarvisGUI: Towards Cross-Device GUI Agents with Dynamic Task Composition

JarvisGUI benchmark tests GUI agents on cross-device workflows across Android, Windows, and Ubuntu, revealing major gaps in state transfer and long-horizon reasoning.

JarvisGUI is a dynamic benchmark that formulates GUI tasks as input-output transformations under a lightweight type system, automatically composing multi-step cross-device workflows across Android, Windows, and Ubuntu virtual environments. Evaluation shows state-of-the-art open-source GUI agents struggle with state-transfer awareness, cross-platform contextual reasoning, and long-horizon dependency management, exposing a capability gap invisible to existing single-device benchmarks.

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 16:59 UTC · 6d agoAI research1

CUA-Universe: A Scalable and Dynamic Environment for Hybrid GUI+CLI Agents

Researchers introduce CUA-Universe, a pipeline turning real desktop software into hybrid GUI+CLI agent environments, lifting a 9B model's OSWorld success rate.

CUA-Universe is an environment-to-data pipeline that converts real desktop applications into hybrid GUI+CLI environments, scaling to 16 applications via App-Forge, Task-Weave, and Path-Steer. Training on its harvested trajectories shifted a 9B model toward effective GUI+CLI orchestration, yielding +39.3 points on CUA-Verse, +16.8 points success rate on OSWorld, and +7.84 points on OSWorld-MCP while cutting steps and tokens by up to 57% and 60%. The work addresses the scarcity of scalable hybrid environments for computer-use agents.

arXiv cs.AI / cs.LG / cs.CL·Sep 4, 17:26 UTC · 11d agoAI research1

Show-Harness: Just a VLM Agent Can Play Robots

Show-Harness lets VLM agents control robots via discrete semantic action units, outperforming VLA baselines zero-shot and after light fine-tuning.

Show-Harness is an embodied agent harness that exposes discrete semantic action units a VLM reasons over, with embodiment-specific interpreters grounding them into local robot actions. It enables zero-shot robot control with closed-source frontier VLMs and low-cost adaptation of small open-source VLMs using only a few GPU-hours of fine-tuning. The companion GUMI (GUI Manipulation Interface) extends the same semantic action space to GUI-based demonstration collection without specialized teleoperation hardware. Experiments show robust generalization across tasks, embodiments, and environments, beating representative agentic and VLA paradigms.

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 17:53 UTC · 6d agoAI research

DOUBLECUP Uses ClickFix and Cached PNGs to Deliver CountLoader and DeviceManager RAT

The Hacker News·Aug 4, 09:03 UTC · Aug 4, 2026Malware

NASA AIT-GUI Flaws Could Let Unauthenticated Attackers Issue Spacecraft Commands

Researchers disclosed 9.4-CVSS flaws in NASA/JPL's AIT-GUI allowing unauthenticated attackers to issue arbitrary spacecraft commands via exposed command endpoints.

Cycode disclosed a flaw chain in AIT-GUI, the browser-based operator console for NASA/JPL's open-source AMMOS Instrument Toolkit, tracked as GHSA-p9r8-2q67-fp86 with CVSS 9.4, affecting versions 2.5.1 and earlier. The server binds to 0.0.0.0:8080, issues sessions without credential checks, and lets unauthenticated users issue spacecraft commands via POST /cmd and reach files outside intended directories via /script/run and /seq (CWE-306, CWE-352, CWE-22). Version 2.5.2 adds host binding and same-origin checks but does not add authentication, and a related missing-authentication flaw was published separately as CVE-2026-60112 with CVSS 9.3.

The Hacker News·Aug 20, 17:20 UTC · 26d agoVulnerabilityCVE-2026-601121

LLaDA-UI: Bringing Block-wise Diffusion to Vision-Language GUI Agents

LLaDA-UI, a 16.7B block-wise diffusion vision-language GUI agent, outperforms Qwen2.5-VL-7B and beats Qwen3-VL-8B on four of six GUI benchmarks.

LLaDA-UI is a 16.7B-parameter MoE-based, block-wise diffusion vision-language GUI agent built on the LLaDA2.0-mini-base diffusion language backbone with a native-resolution vision encoder. It uses a two-stage pipeline: general multimodal pre-training followed by GUI-agent supervised fine-tuning on mobile, desktop, web, and grounding data. It substantially outperforms Qwen2.5-VL-7B and surpasses Qwen3-VL-8B on four of six reported GUI benchmarks, establishing block-wise diffusion as a practical paradigm for latency-sensitive multimodal agents.

Hugging Face daily papers·Sep 8, 20:00 UTC · 7d agoAI research

Testing race conditions with memory access tracing and stack-based delay injection

Google Project Zero released MAccConc, Linux kernel tooling that traces memory accesses to explore and test race condition interleavings.

A Google Project Zero researcher published MAccConc (Memory Access Concurrency), tooling for exploring possible interleavings of multithreaded test cases in the Linux kernel, available on GitHub. The tools use KCOV with ASAN outline-mode instrumentation to record per-access memory traces, enabling automatic testing of all A-B-A interleavings plus terminal and GUI explorers for manual analysis. The work targets confirming race condition candidates, building reliable regression tests, and enabling concurrency fuzzing, drawing on ideas from SKI and Ned Williamson's sockfuzzer.

Google Project Zero·Sep 8, 07:00 UTC · 8d agoResearch1

TRACE: Trajectory-robust Admission with Evidence Ordering for Efficient GUI Agents

TRACE, a training-free visual token pruning framework, cuts GUI agent inference latency and memory while keeping trajectory-wide visual evidence reusable.

TRACE is a training-free framework for trajectory-robust admission and coverage-aware evidence ordering that prunes high-resolution screenshot tokens accumulated in GUI agent trajectories. It ranks visual evidence using a query-independent layout-derived interaction prior combined with instruction relevance and feature novelty, and reserves part of the budget for native tokens distributed across the screen to repair spatial coverage. A monotone KV contraction incrementally compresses retired frames into compact session state, avoiding repeated visual encoding or pruning. Experiments across six GUI benchmarks and diverse models verify effectiveness under tight budgets, with source code to be released.

Hugging Face daily papers·Sep 8, 20:00 UTC · 7d agoAI research

Critical Flaw in NASA/JPL Open-Source Spacecraft Command Software Allowed Unauthenticated Command Execution

NASA/JPL's AIT-GUI web console (CVSS 9.4) lacked authentication and CSRF protection, letting anyone send unauthenticated commands to spacecraft instruments; fixed in 2.5.2.

Cycode researchers found that AIT-GUI, the browser console of NASA/JPL's open-source AMMOS Instrument Toolkit, ran an HTTP server with no authentication, no session checks, and no CSRF protection on state-changing endpoints, allowing arbitrary commands to spacecraft instruments. The issue combines four weaknesses: binding to 0.0.0.0 regardless of configured host, missing authentication, missing CSRF defense, and path traversal in the /seq and /script/run endpoints. Tracked as GHSA-p9r8-2q67-fp86 with CVSS 9.4, it is fixed in AIT-GUI 2.5.2. The findings came from AI-assisted code analysis confirmed with working proofs-of-concept, including a headless-browser CSRF demonstration.

Security Affairs·Aug 22, 08:04 UTC · 25d agoVulnerability1

Mistral X Mozilla: Private, Multilingual AI Browsingnew

Mistral and Mozilla partnered to power Firefox's Smart Window AI browsing assistant in France and North America, with zero data retention.

Mozilla's Firefox Smart Window (beta) AI browsing assistant is now powered by Mistral models for users in France and North America, with the UK and Germany expected later this year. Conversations are not saved on Mozilla's servers by default, and Mistral agreed to zero data retention. Both companies frame the partnership as advancing open-source, privacy-first, and regionally fine-tuned AI, with models trained on regional languages, dialects, and cultural context.

Hacker News · security·08:08 UTC · 2h agoAI industryHN 35↑ · 8 comments

Homebrew 7.0.0 gets built-in GUI, better security controls

Homebrew 7.0.0 adds a built-in vulnerability scanner (brew vulns), a Homebrew advisory database, stronger sandboxing, and the BrewUI GUI.

Homebrew 7.0.0 ships a native BrewUI graphical interface for macOS 26 Tahoe and later, plus a new 'brew vulns' command that queries OSV.dev in batch to flag vulnerabilities in installed formulae, specific packages, or Brewfile dependencies. The release also introduces a Homebrew-specific OSV-format advisory database distinguishing outstanding vulnerabilities from backported fixes. Sandbox improvements block home directory access by default and separate network-enabled dependency downloads from offline installs.

BleepingComputer·Sep 14, 19:51 UTC · 1d agoTools1

Context Engineering Inside the Harness: 4 Mechanisms That Beat Context Overflow and Goal Loss on Long-Horizon Tasks

Survey of four harness mechanisms—context budgeting, compaction, todo-state, and memory—that keep long-horizon LLM agents on task across 200+ tool calls.

The article details how agent harnesses, not larger context windows, solve context overflow and goal loss on long-horizon tasks, citing Chroma's Context Rot report showing 18 LLMs (GPT-4.1, Claude 4, Gemini 2.5, Qwen3) degrade on long inputs. Concrete implementations include LangChain Deep Agents offloading tool responses over 20,000 tokens to the filesystem and truncating old tool calls at 85% window usage, and Claude Code capping auto memory at 25KB while re-reading the 5 most recently modified files after compaction. OpenAI's Responses API now offers server-side compaction via context_management with a standalone /responses/compact endpoint, which Codex uses for long-running coding tasks. Manus reports a roughly 100:1 input-to-output token ratio per ~50-tool-call task, motivating todo.md state recitation to prevent goal drift.

MarkTechPost·Sep 13, 05:56 UTC · 3d agoAI research1

JWT used for authentication in web GUI signed with static key

Fortinet warns that FortiMonitorOnSight's web GUI uses JWTs signed with a static key, enabling remote unauthenticated authentication bypass.

Fortinet advisory FG-IR-26-170 discloses a CVSS 9.6 vulnerability (CWE-540, sensitive information in source code) in the FortiMonitorOnSight web portal. The web GUI's JWT authentication tokens are signed with a static key, allowing a remote unauthenticated attacker to bypass authentication via forged or reused JWTs. The advisory was revised on September 8, 2026.

Fortinet PSIRT·Sep 8, 07:00 UTC · 8d agoAdvisory

Bring your spreadsheet data to life with Sheets canvas

Google added Sheets canvas to Workspace, letting users generate interactive dashboards, trackers, and charts from spreadsheet data via prompts.

Google introduced Sheets canvas, a Workspace feature that turns spreadsheet data into interactive dashboards, custom study trackers, and seating charts from a simple prompt. The announcement is a productivity feature launch with no described security impact or risk. No metrics, model details, or pricing are provided in the notice.

Google · AI·Aug 13, 16:45 UTC · Aug 13, 2026AI industry

Fortinet FortiSandbox Vulnerability Allows Unauthenticated Attackers to Access Sensitive Information

Fortinet fixed CVE-2026-26084, an unauthenticated access-control flaw in FortiSandbox GUI rated 8.9 CVSS, with no known exploitation yet.

Fortinet disclosed CVE-2026-26084 (advisory FG-IR-26-166), a CWE-284 improper access control flaw in the GUI of FortiSandbox, FortiSandbox Cloud and FortiSandbox PaaS, rated 8.9 CVSS v3.1. An unauthenticated remote attacker can send specially crafted HTTP requests to control NAT rules and expose sensitive information. Affected versions include FortiSandbox 4.4.0-4.4.8 and 5.0.0-5.0.5 (plus Cloud/PaaS 5.0.4-5.0.5), fixed in 4.4.9 and 5.0.6. Fortinet researcher Adham El Karn found the flaw internally and the September 8 advisory reports no known exploitation.

GBHackers·Sep 9, 07:15 UTC · 7d agoVulnerabilityCVE-2026-26084

Show HN: LLM Attention Visualization

A developer released a browser-based tool that visualizes which past tokens influence each LLM output token using aggregated, value-weighted attention scores.

A Show HN project presents a React application built on Transformers.js that renders per-token attention influence by aggregating attention weights scaled by value-vector magnitudes across all attention heads and layers. To expose internal tensors, the author instrumented the ONNX computation graph, hosted a modified model on Hugging Face, and pre-generated prompts to avoid long model downloads in the browser. Demos with a 600-million-parameter model show how verbatim copying draws heavily on source tokens and how single outputs blend information from multiple phrases.

Hacker News · AI·Sep 8, 16:59 UTC · 7d agoAI tools & infraHN 41↑ · 9 comments

RANCOR: Targeted Attacks in South East Asia Using PLAINTEE and DDKONG Malware Families

Unit 42 names RANCOR, a previously unidentified espionage group targeting Southeast Asian political entities with newly discovered PLAINTEE and DDKONG malware families.

Unit 42 tracks a previously unidentified group it dubs RANCOR conducting highly targeted espionage attacks in Singapore and Cambodia using two custom malware families, DDKONG and PLAINTEE. Attacks appear to have begun with spear phishing, with decoy political news documents hosted on legitimate sites including a Cambodian government website and Facebook. Infrastructure links to IP 89.46.222[.]97 and KHRAT-associated domains such as facebook-apps[.]com, and the activity is grouped into two clusters tied together by PLAINTEE usage and similar targeting.

Palo Alto Unit 42·Aug 17, 13:24 UTC · 29d agoThreat actor1

27.5KB language-agnostic WebGPU syntax highlighter

A developer released gpu-lexer, a 27.5KB language-agnostic syntax highlighter that uses a tiny WebGPU model to label code tokens in the browser.

gpu-lexer splits source into words, whitespace, and symbols, then a small WebGPU model uses local and whole-file context to assign nine token classes, working on languages never seen in training. On held-out files, 12.57% of token labels differ from Shiki, though this measures agreement with Shiki rather than objective correctness. In benchmarks against Shiki 4.4.3, Prism.js, Highlight.js, Sugar High, and Starry Night, it highlighted 10 concatenated copies of three.min.js (5.56M characters) about 10x faster on an Apple M4 Pro in Chrome 152. The author frames it as an experiment, not a grammar-equivalent highlighter.

Hacker News · security·Sep 9, 01:14 UTC · 7d agoAI tools & infraHN 21↑ · 5 comments1

Server-Side Request Forgery (SSRF)

Fortinet discloses a low-severity SSRF in the FortiSIEM GUI allowing authenticated attackers to send requests from targeted devices.

Fortinet PSIRT advisory FG-IR-26-159, revised 2026-08-12, describes a server-side request forgery (CWE-918) in the FortiSIEM GUI, scored CVSSv3 3.4. An authenticated attacker can send HTTP requests originating from the targeted device via specially crafted requests, potentially enabling internal network probing. No CVE identifier or exploitation status is included in the advisory text.

Fortinet PSIRT·Aug 12, 07:00 UTC · Aug 12, 2026Advisory

Show HN: Panel – A research workspace where the agent can build its own panes

Panel is an open-source research workspace where AI agents dynamically build their own panes, posted on Hacker News.

Panel, shared as a Show HN project on GitHub, is a research workspace in which the agent can construct its own panes rather than using a fixed interface. The post received 41 points and 8 comments on Hacker News. It targets agentic research workflows with a dynamically generated UI.

Hacker News · AI·Sep 15, 13:58 UTC · 20h agoAI tools & infraHN 41↑ · 8 comments

NASA Ground Control Software Flaw Enables Unauthenticated Commands

Critical flaws in NASA's AIT-GUI ground control software let unauthenticated attackers send spacecraft commands and execute scripts.

NASA's AIT-GUI ground control software contains critical flaws that expose spacecraft command and script execution to unauthenticated attackers. Anyone able to reach the ground control interface could send unauthorized commands without valid credentials. The report does not indicate that the flaws have been exploited in the wild.

Infosecurity Magazine·Aug 18, 14:30 UTC · 28d agoVulnerability

SAP Patches CVSS 10.0 Kernel Flaw Enabling Unauthenticated Remote Code Execution

SAP patched CVE-2026-44756 (CVSS 10.0), an unauthenticated kernel memory corruption allowing OS command execution, plus three other critical flaws.

SAP's September security updates include CVE-2026-44756 (OVERPASS, CVSS 10.0), a missing boundary validation during deserialization of Extended Passport (EPP) data in the SAP kernel, enabling unauthenticated attackers to run arbitrary OS commands with SAP administrative privileges. Also patched: CVE-2026-58240 (S4GET, CVSS 9.8), a missing authentication check in SAP NetWeaver Message Server yielding RCE as <sid>adm; CVE-2026-76969 (9.4), credential disclosure in SAP Cloud Application Programming Model multi-tenant apps; and CVE-2026-66768 (9.0), improper access control in SAP NetWeaver SAP GUI for Java. Onapsis, which discovered the flaws, says none have been exploited to date and recommends prioritizing internet-facing SAP systems.

The Hacker News·Sep 9, 06:25 UTC · 7d agoVulnerabilityCVE-2026-44756CVE-2026-58240CVE-2026-76969+1 CVEs

Studying Image Tokenizers as Visual Languages in Unified Multimodal Models

A controlled autoregressive testbed shows validation losses must be analyzed per task, and image tokenizer choice affects joint multimodal text modeling.

Researchers built a pure-autoregressive testbed to study image tokenizers as the 'visual language' of unified multimodal models, tracking task-specific validation losses during multimodal continual pretraining across text, image, text-to-image (T2I), and image-to-text (I2T) prediction. They found that losses exhibit distinct scaling behavior per task and rank tokenizers differently, and that I2T loss over a shared text vocabulary gives a more consistent loss–performance signal than T2I loss. Better reconstruction does not necessarily yield lower task-specific losses or stronger downstream performance, and tokenizer choice can affect text modeling under joint optimization. Case studies examine the discriminator, semantic supervision, and vocabulary size design axes.

Hugging Face daily papers·Sep 7, 20:00 UTC · 8d agoAI research1

xHunt Campaign: Attacks on Kuwait Shipping and Transportation Organizations

xHunt deployed custom backdoors Hisoka, Killua, Gon, and EYE against Kuwait shipping and transportation organizations, using DNS tunneling and email-draft C2 channels.

Unit 42 observed the xHunt campaign targeting Kuwait transportation and shipping organizations between May and June 2019, installing backdoors named after characters from the anime Hunter x Hunter, including Hisoka, Killua, Sakabota, and Netero. The tools use HTTP and DNS tunneling for C2, and Hisoka v0.9 added an unusual email-based C2 channel that abuses Exchange Web Services with stolen credentials to exchange email drafts. The Gon tool provides port scanning, file transfer, screenshots, remote command execution, and RDP session creation, while EYE acts as a cleanup failsafe if a legitimate user logs in. Related activity targeting Kuwait between July and December 2018 was reported by IBM X-Force IRIS, suggesting the campaigns are likely related.

Palo Alto Unit 42·Aug 17, 13:21 UTC · 29d agoThreat actor1

CanvasAnneal: Curriculum Reinforcement Learning for Diffusion Language Models

CanvasAnneal injects teacher reasoning traces into diffusion canvases during curriculum RL, improving diffusion LLMs on MATH500, Countdown, and Tau2.

CanvasAnneal is a curriculum-guided reinforcement learning framework for diffusion language models that addresses exploration bottlenecks in standard RL. It warm-starts exploration by injecting teacher-generated reasoning traces into the initial diffusion canvas, then gradually removes this guidance so the model generates reasoning trajectories independently. Across mathematical reasoning and tool-use benchmarks, it improves over standard diffu-GRPO on MATH500, Countdown, and Tau2 and accelerates reward improvement, though gains are task-dependent.

arXiv cs.AI / cs.LG / cs.CL·Sep 11, 16:59 UTC · 4d agoAI research1

SAP September 2026 Security Update Fixes 4 Critical Vulnerabilities and 15 Other Flaws

SAP's September 2026 Patch Tuesday ships 19 security notes, including four criticals led by a CVSS 10.0 memory-corruption flaw in Extended Passport Processing.

SAP released 19 new Security Notes plus one update, fixing four critical vulnerabilities and 15 additional flaws. The most urgent is CVE-2026-44756 (CVSS 10.0), memory corruption in Extended Passport (EPP) Processing affecting many SAP Kernel and Web Dispatcher releases. Other criticals include CVE-2026-58240 (NetWeaver Message Server missing authentication, 9.8), CVE-2026-76969 (credential disclosure in CAP sap/cds-mtxs, 9.4), and CVE-2026-66768 (improper access control in SAP GUI for Java, 9.0). High-severity fixes cover XXE in Integration Suite Trading Partner Management, deserialization in NetWeaver Business Client, and memory corruption in NetWeaver AS for ABAP.

GBHackers·Sep 8, 10:24 UTC · 8d agoVulnerabilityCVE-2026-44756CVE-2026-58240CVE-2026-76969+8 CVEs1

TANGO: Humanoid Navigation in Cluttered Environments with a Whole-Body Vision-Language-Action Model

Researchers introduce TANGO, a whole-body vision-language-action model enabling zero-shot language-guided humanoid navigation on the Unitree G1 robot.

TANGO addresses humanoid navigation in cluttered indoor environments by predicting 29-DoF joint-space actions directly from natural-language instructions and egocentric RGB, rather than 2D path planning. It is trained entirely in simulation via a pipeline combining global path planning, kinematic whole-body motion generation, obstacle-aware motion editing, and RL-based tracking. In simulation it achieves state-of-the-art vision-language navigation performance and transfers zero-shot to a Unitree G1 humanoid without any real-world navigation data.

Hugging Face daily papers·Sep 7, 20:00 UTC · 8d agoAI research1

Show-Harness: Just a VLM Agent Can Play Robots

Show-Harness enables VLM agents to control robots via a semantic action interface, achieving zero-shot frontier control and few-GPU-hour adaptation of small VLMs.

Show-Harness exposes discrete semantic action units that VLMs reason over, with embodiment-specific interpreters deterministically grounding them into local robot actions. It enables zero-shot closed-source frontier VLM control and adapts small open-source VLMs for low-cost deployment with a few GPU-hours of fine-tuning. The companion GUMI interface extends the same semantic action space to GUI-based demonstration collection without teleoperation hardware, and Show-Harness-equipped agents outperform representative agentic and VLA paradigms.

Hugging Face daily papers·Sep 8, 20:00 UTC · 7d agoAI research

Axis Robotics Releases AXIS: A Browser-Based Data Engine With 207 Robot Manipulation Tasks and 50,129 Trajectories

Axis Robotics and academic partners released AXIS, a browser-based teleoperation system yielding 207 manipulation tasks and 50,129 trajectories that lifts pi0.5 to 88.8 on LIBERO-Plus.

A team from Axis Robotics, UC Berkeley, Georgia Tech, and NTU introduced AXIS, a browser-based data engine where contributors teleoperate a simulated Franka Research 3 in a MuJoCo WebAssembly frontend while GPU backends handle task generation, training, and evaluation. The released snapshot holds 207 tasks, 50,129 episodes, and 60K+ task or scene variants from more than 70,000 community contributors. Continual pretraining of pi0.5 on AXIS data raises LIBERO-Plus performance from 83.9 to 88.8, versus 57.5 for a volume-matched RoboCasa365 control; the 2.36 TB dataset is gated for non-commercial academic use.

MarkTechPost·Sep 7, 18:38 UTC · 8d agoAI research

nex-agi/Nex-N2.5-mini — new model trending #30 on Hugging Face

Nex-AGI releases Nex-N2.5 agentic model family (mini, Pro, Max) with a 1.6-trillion-parameter MoE Max, open weights, and hosted access via OpenRouter.

Nex-AGI launched Nex-N2.5, a family of agentic models in mini, Pro, and Max sizes, with the Max version built on a 1.6-trillion-parameter text-only Mixture-of-Experts foundation and the company's first complete post-training effort at trillion-parameter scale. The models target long-horizon computer use, web browsing, and visually grounded agentic tasks, with expanded agent training environments. Reported benchmarks include Max scoring 86.1 on Terminal-Bench 2.1 and 65.7 on SWE-Bench Pro, trailing Claude Opus 5. Weights are being released openly on Hugging Face and ModelScope, with hosted access through OpenRouter.

Hugging Face trending models·Sep 8, 05:08 UTC · 8d agoModel release1

HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harnessnew

Researchers introduce HarnessVLN, a zero-shot training-free agent harness that sets new training-free SOTA on vision-language navigation benchmarks including R2R and HM3D.

HarnessVLN is a zero-shot, training-free framework for embodied vision-language navigation that coordinates perception, retrieval, grounding, navigation, recovery, and termination through a unified tool interface. It validates planner proposals against spatial evidence, geometric feasibility, and subgoal consistency, using hierarchical event memory and a persistent Spatiotemporal Graph that stores reusable spatial evidence and failure annotations. It reports success rates of 60.8% on R2R, 53.9% on RxR, 76.0% on HM3D-v2, and 59.3% on HM3D-OVON, surpassing prior training-free state of the art, with real-world humanoid deployment demonstrated.

Hugging Face daily papers·Sep 13, 20:00 UTC · 2d agoAI research

2026-011: Critical Vulnerabilities in SAP Kernel and NetWeaver Message Server

SAP patched two critical flaws, OVERPASS (CVE-2026-44756, CVSS 10.0) and S4GET (CVE-2026-58240), allowing unauthenticated attackers to execute OS commands on SAP hosts.

On SAP's September 2026 Security Patch Day, SAP released Security Notes 3747649 and 3759472 fixing two critical unauthenticated remote vulnerabilities found by Onapsis. CVE-2026-44756 ('OVERPASS', CVSS 10.0) is a memory corruption flaw in Extended Passport (EPP) deserialisation in the SAP Kernel; CVE-2026-58240 ('S4GET', CVSS 9.8) is a missing authentication check in the NetWeaver Message Server that lets attackers register as trusted cluster nodes. Successful exploitation of either can yield OS command execution as the SAP service account, leading to full system and business data compromise, and CERT-EU urges immediate patching. No in-the-wild exploitation is reported.

CERT-EU Advisories·Sep 9, 00:00 UTC · 7d agoVulnerabilityCVE-2026-44756CVE-2026-58240

MetroLLM-Bench: Evaluating Language Models as Transit Kiosk Runtimes

MetroLLM-Bench is a 955-case benchmark testing language models as transit kiosk tool-calling runtimes across six real metro systems.

The benchmark covers 37-414-station metro systems and eleven task categories including routing, fare calculation, disruptions, accessibility, and adversarial input, with 14 deterministic and 8 semantic scoring components. Of 26 models from six vendors, a PEFT-tuned 4B Qwen 3.5 student scored 91.3 on Tier 1, exceeding GPT-5.6 (90.6/90.0), while Muse Glimmer 30B led the composite ranking. A deterministic rule-based baseline reached 84.6, and PEFT gains over base models shrank from +7.03 points at 2B to -0.91 at 27B.

Hugging Face daily papers·Sep 8, 20:00 UTC · 7d agoAI research

Securing Claude Code: The New Compliance API, Local Visibility, and Identity Governance

Anthropic's new Compliance API endpoints expose Claude Code local session transcripts, highlighting governance gaps for endpoint AI agents.

Anthropic added local session transcript endpoints to its Compliance API on August 11, 2026, giving security teams visibility into prompts, bash commands, file operations, and MCP commands run by Claude Code harnesses on endpoints. The article argues local harnesses break the classic shared-responsibility model, citing Token Security data that 68.6% of discovered AI agents run on endpoints, and a Cloud Security Alliance survey of 418 IT and security professionals in which 82% found an unknown agent within the past year. It outlines three governance layers: Anthropic managed settings as a policy baseline, the Compliance API for cloud-visible transcripts, and endpoint telemetry to connect agent activity to identity, credentials, and permissions.

The Hacker News·Aug 31, 11:31 UTC · 15d agoAI safety & security1

How Hugging Face Inference Endpoints, Jobs, and Buckets Power Search on Papers with Code

Hugging Face explains how Inference Endpoints, Jobs, and Buckets power semantic search on Papers with Code.

Hugging Face describes the infrastructure behind search on Papers with Code, built on its Inference Endpoints, Jobs, and Buckets services. The post is a product-focused engineering walkthrough with no security impact.

Hugging Face Blog·Aug 21, 00:00 UTC · 26d agoAI tools & infra

Tracking OceanLotus’ new Downloader, KerrDown

Unit 42 identifies KerrDown, a new OceanLotus (APT32) downloader active since 2018 targeting Vietnamese speakers via malicious macros and DLL side-loading.

Unit 42 tracks KerrDown, a previously undocumented downloader family used by OceanLotus (APT32) since at least early 2018, primarily targeting Vietnam or Vietnamese-speaking individuals. Delivery uses macro-laced Microsoft Office documents embedding base64-encoded 32-bit and 64-bit DLLs, and RAR archives containing a legitimate program abused for DLL side-loading. KerrDown is dropped as main_background.png, downloads a DES-encrypted payload from a URL, and executes it directly in memory. Researchers used Jaccard-index similarity analysis to identify the new family, connect campaign samples, and infer patterns in the group's working hours and days.

Palo Alto Unit 42·Aug 17, 12:53 UTC · 29d agoMalware in the wild1

CodeTD: Topology of Attention Detects Hallucinations in Code LLMs

CodeTD detects hallucinations in code LLMs before execution by analyzing topological patterns of attention maps, outperforming recent baselines.

CodeTD applies topological data analysis (TDA) to code LLM attention maps to quantify prompt-generation mismatch as a pre-execution correctness signal. Experiments cover HumanEval, MBPP, BigCodeBench, and MultiPL-E across 5 programming languages and 10 code LLMs up to 34B parameters. The method outperforms recent baselines and transfers between coding benchmarks, helping catch code that fails the task or embeds security vulnerabilities.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:26 UTC · 8d agoAI research1

SAP security advisory – September 2026 monthly rollup (AV26-894)

Canada's Cyber Centre relayed SAP's September 2026 Patch Day rollup covering vulnerabilities across NetWeaver, kernel components, CAP, and Integration Suite.

The Canadian Centre for Cyber Security published advisory AV26-894 noting that as of September 8, 2026, SAP is affected by vulnerabilities in multiple products, including SAP Extended Passport, NetWeaver Message Server, SAP Cloud Application Programming Model, SAP GUI for Java, Integration Suite Cloud Integration, NetWeaver Business Client, and NetWeaver AS for ABAP. The advisory corresponds to SAP Security Patch Day September 2026 and lists many affected kernel and web dispatcher versions. Administrators are urged to review the SAP advisory and apply updates as available.

Canadian Centre for Cyber Security·Sep 8, 18:03 UTC · 7d agoAdvisory

ZGCM-1: A Fully Open and Extremely Efficient Foundation Model for Math and Agentic Search

ZGCM-1 is a fully open 7B foundation model with 256K context that stays competitive with frontier models on math reasoning and agentic search.

ZGCM-1 is a fully open 7B dense foundation model trained from scratch using an efficiency-focused recipe: interleaved gated sliding-window and full attention, a stable FP8 Muon optimizer, and MDP-based mid-training with context scaling across 16K, 64K, and 256K. On mathematical reasoning and agentic search suites it remains competitive with much larger frontier models such as Qwen3-235B-A22B and GLM-5.1. The recipe yields a ~4.2x improvement in 16K pre-training time-to-loss, and all weights, checkpoints, training code, data recipes, and W&B logs are open-sourced.

Hugging Face daily papers·Sep 10, 20:00 UTC · 5d agoModel release

SAP Patches Critical Extended Passport Processing Vulnerability

SAP patches CVSS 10.0 flaw CVE-2026-44756 (OVERPASS) enabling unauthenticated RCE in S/4HANA and NetWeaver, plus three other criticals; no in-the-wild exploitation reported.

CVE-2026-44756 is a CVSS 10.0 memory corruption flaw in Extended Passport (EPP) deserialization, triggered when a user session opens across web, SAP GUI, and RFC vectors. Exploitation yields command execution as the SAP OS account, database credential and password hash recovery, session reading, and data or binary modification; affected products include S/4HANA, ERP/ECC, NetWeaver, Web Dispatcher, and BW/4HANA. Three other criticals were patched: CVE-2026-58240 (S4GET, missing authentication affecting all S/4HANA 2025 and earlier releases), CVE-2026-76969 (CAP credential disclosure), and CVE-2026-66768 (NetWeaver access control). Onapsis and SAP report no indicators of in-the-wild exploitation.

SecurityWeek·Sep 8, 14:55 UTC · 7d agoVulnerability 2 sourcesCVE-2026-44756CVE-2026-58240CVE-2026-76969+1 CVEs