Search: “itdr”

40 stories

DomainAll Cyber AIWindow24h 3d 7d 30d AllIn the wild Important only Show duplicates

CategoryAll Vulnerability 19 Industry 18 Policy & legal 16 Ransomware 16 Data breach 14 Threat actor 13 Malware 8 Phishing & fraud 4 AI safety & security 3 AI tools & infra 2 Exploit / PoC 2 Tools 2 AI policy 1 Research 1

SeverityAll critical 57 high 17 medium 15 low 13 info 18

Smart search ranks by meaning as well as keywords (one row per story, last 45 days).

The 12 Best Extended Detection & Response (XDR) Platforms, Compared and Priced

Buyer's guide compares 12 XDR platforms, favoring Microsoft Defender XDR, Stellar Cyber and CrowdStrike, and warns ingestion pricing inflates costs.

The article compares 12 extended detection and response platforms, distinguishing native XDR (CrowdStrike, Palo Alto Cortex XDR, Microsoft Defender XDR, SentinelOne) from open XDR (Stellar Cyber, Arctic Wolf, Rapid7). It argues data ingestion pricing, not per-endpoint fees, is the main budget risk and should be modeled before signing. It repeats the consolidation note that Sophos acquired Secureworks for approximately $859 million in February 2025, and flags that ExtraHop is NDR rather than full XDR.

GBHackersupdated Sep 9, 08:07 UTC · 7d agofirst Sep 9, 07:51 UTC · 7d agoIndustry 3 sources1

New Huntress Managed ITDR Dashboard: Faster Identity Investigations

Huntress redesigned its Managed ITDR dashboard, adding Rapid Identity Triage, Failed Login Characterization, and Quick SIEM search to speed identity investigations.

Huntress announced a redesigned Managed ITDR dashboard aimed at accelerating identity threat investigations. New capabilities include Rapid Identity Triage, Failed Login Characterization, and Quick SIEM search. The update is a vendor product change with no incident or vulnerability details attached.

HuntressAug 27, 14:00 UTC · 20d agoTools

Credential Theft: How Attackers Steal & Use Stolen Credentials

Huntress explains how attackers steal credentials through phishing, AitM, infostealers, and dumping, then use them for lateral movement, BEC, and ransomware.

Huntress published an educational overview of credential theft, citing that roughly 70% of confirmed data breaches begin with stolen credentials. It details acquisition methods including phishing, adversary-in-the-middle attacks that capture MFA session tokens, infostealers (nearly a quarter of threats Huntress observed in 2025), Mimikatz-based credential dumping, credential stuffing, and password spraying. The piece then covers post-theft actions such as lateral movement, privilege escalation, account takeover, business email compromise, and ransomware, and closes with behavioral detection guidance and layered prevention strategies.

HuntressSep 10, 16:00 UTC · 6d agoResearch

Notes on gotchas while migrating 35kb preprompts from Opus to self-hosted Ollama

Opinion piece urges migrating 35KB preprompts from Anthropic/OpenAI to self-hosted Ollama, citing session privacy risks and safety filters blocking security research.

The author documents gotchas migrating 35KB preprompts from Claude Opus to self-hosted Ollama, motivated by fears that frontier providers train on user sessions, citing the OpenAI Navier-Stokes controversy. The piece argues inference providers cannot audit their own retention or training pipelines and that only self-hosted hardware offers verifiable privacy. It also criticizes frontier safety filters for refusing vulnerability research tasks and calls for models that support exploitability testing in CI/CD pipelines.

Hacker News · AISep 14, 13:59 UTC · 2d agoAI industry HN 41↑ · 13 comments1

2026 Cyber Insurance Trends Report: What's Changed and What You Need to Know

Huntress survey: CIRCIA reporting mandates now live, BEC claims exceed ransomware, exfiltration-heavy attacks cost twice as much, premiums rising.

Huntress's 2026 cyber insurance trends report, based on its own survey, finds 79% of respondents carry cyber insurance while 58% report shrinking coverage over five years. New CIRCIA federal reporting mandates and EU NIS2 requirements are reshaping policies, business email compromise now drives more claims than ransomware, and data exfiltration has replaced encryption as the dominant ransomware tactic at roughly twice the cost. After three years of declining premiums, rates are climbing again, and most businesses now refuse to pay ransoms.

HuntressSep 1, 17:00 UTC · 15d agoIndustry1

Anatomical Grounding and Leakage-Aware Multimodal Contrastive Learning for Alzheimer's Disease Classification from Structural MRI

Study of label leakage and anatomical grounding in multimodal MRI models for Alzheimer's staging shows cognitive-score fusion accuracy of 87.3% is leakage-driven.

The authors train a ResNet18 slice-based encoder with a one-layer Transformer on 1,075 ADNI-1 T1 MRI scans, using FastSurfer segmentations and YOLOv8 localization (mAP_50 above 0.96) as anatomical reference. Grad-CAM shows the image-only classifier often attends to skull and background rather than disease-relevant structures. A CLIP-style image-tabular contrastive framework organized along a label-leakage spectrum yields 87.3% three-way accuracy with cognitive scores versus 73.0% with regional volumes, and cropping to the medial temporal lobe raises image-only accuracy from 58.7% to 65.1%. Results come from single runs on a small balanced test set with reported confidence intervals.

arXiv cs.AI / cs.LG / cs.CLSep 14, 17:13 UTC · 2d agoAI research1

35 Actionable Password Statistics for Businesses in 2026 | Huntress

Huntress compiles 2026 password statistics showing 94% of 19 billion leaked passwords were reused and 37% of identity threats used stolen credentials.

Huntress published a compilation of password security statistics drawing on sources including Cybernews, Verizon's 2026 DBIR, IBM, and Bitwarden. Cybernews found 19 billion exposed passwords from roughly 200 incidents between April 2024 and April 2025, with only 6% unique and 94% reused across accounts. Huntress telemetry reports 37% of identity-based threats in 2026 involved stolen or suspicious credentials, while Verizon cites credential abuse in 39% of breaches. The piece argues weak and reused passwords remain a top entry point and recommends improved password hygiene.

HuntressSep 10, 13:00 UTC · 6d agoIndustry

Anchoring Clinical Events in Time: UID-Preserving Multimodal Reconstruction and Source-Grounded Adjudication

UID-preserving multimodal framework plus GAVEL LLM judge improves clinical timeline reconstruction, boosting event recovery 43% over prior matching.

The paper introduces a UID-preserving framework linking each narrative clinical event to its source span through text-only estimation, structured-evidence retrieval, timestamped source-row grounding, and joint revision. GAVEL, an LLM judge, compares UID-aligned timelines against narrative and structured records. Across six open-weight models and 40 mixed-critical-care summaries, GLM 5.2 multimodal revision improved temporal agreement without reducing event recovery and performed competitively with clinician annotations, while DeepSeek V3.2 did not benefit from multimodality. The pipeline achieves 43% increased event recovery with occurrence-level provenance.

arXiv cs.AI / cs.LG / cs.CLSep 11, 17:01 UTC · 5d agoAI research1

Top 10 Best Cloud Detection & Response (CDR) Solutions in 2026

Editorial scorecard ranks ten 2026 cloud detection and response platforms; Sysdig, Wiz, and CrowdStrike lead, with Wiz's Gem Security acquisition highlighted.

The editorial scorecard rates ten CDR platforms on real-time detection (30%), cloud telemetry depth, response automation, correlation, and value. Sysdig earns the best real-time detection score for its Falco- and eBPF-powered runtime telemetry, Wiz (8.7) folds acquired Gem Security's real-time CDR into its security graph, and CrowdStrike (8.7) leads response automation. Specialists Stream.Security, Skyhawk Security, Sweet Security, and the open-source Falco project are also assessed.

Cyber Security News07:41 UTC · 12h agoIndustry

Cortex Xdr

Stub page for Palo Alto Cortex XDR with no article text, offering no news content for classification.

The item contains only the title 'cortex xdr' and no article body. It references Palo Alto Networks' Cortex XDR product but provides no event, release, or threat information. Classified as general vendor/industry content with minimal relevance.

Palo Alto Unit 42Sep 8, 14:27 UTC · 8d agoIndustry

Cybersecurity jobs available right now: August 18, 2026

Help Net Security lists current cybersecurity job openings at ADI Global Distribution, Schneider Electric, L'Oreal, GovCIO, Elastic, and others across multiple countries.

A recurring job roundup featuring roles such as CISO at ADI Global Distribution, Cybersecurity Analyst at Schneider Electric in India, Cybersecurity Architect at L'Oreal in France, and security engineering positions at GovCIO, Neros Technologies, Elastic, and Chamelio. Most listed positions are no longer accepting applications; several emphasize AI-enabled security environments and identity threat detection and response. No vulnerability, incident, or research content is included.

Help Net SecuritySep 9, 12:47 UTC · 7d agoIndustry 2 sources

Three smart ways SMBs can improve cybersecurity

Opinion piece urges small and midsize businesses to adopt proactive prevention, threat detection, and 24/7 MDR or XDR services.

This opinion article argues SMBs face outsized cyber risk due to limited budgets, small IT teams, and reactive security postures. It recommends proactive prevention, a defined threat detection and response strategy, and managed detection and response (MDR) or extended detection and response (XDR) services. It cites ransomware costs up to $10,000 per device and an attack every 39 seconds as motivation for adopting vendor-operated 24/7 monitoring.

Help Net SecurityAug 27, 06:50 UTC · 20d agoIndustry

Cybersecurity jobs available right now: June 30, 2026

Help Net Security lists cybersecurity job openings across the US, UAE, France, Israel, Japan, Ireland, India, and Canada at major employers.

The roundup includes roles such as AI Offensive Security Engineer at AGAPI, Cloud Security Engineer at Spotify, IAM Engineer at Proton, DFIR in Israel, and a Senior Inspector overseeing NIS2 and AI regulation compliance in Ireland. Most listings were marked as no longer accepting applications, with a few still open including DFIR at Cye, IAM Engineer at Proton, and Detection Engineering at Saronic. The list covers employers across finance, infrastructure, and technology sectors.

Help Net SecurityAug 20, 10:38 UTC · 27d agoIndustry1

Technical Manual for a Toolkit for Measuring Contextual Individuation in Transformer Language Models

An open methodology toolkit measures whether transformer language models contextualize fixed word forms across domains using bridge forms and layer-wise silhouette analysis.

The manual documents an open toolkit built around 'bridge forms' - identical written words recurring across two or more subject domains with a different sense in each - to test whether transformer language models individuate word occurrences by context beyond the embedding layer. It covers declarative specification of bridge forms, Wikipedia corpus acquisition, occurrence localization, layer-wise representation extraction, domain-pairwise silhouette measurement, and visualization, justifying each choice against failure modes such as sense contamination and subword-tokenization misalignment. It is a methodological and implementation reference and reports no empirical results.

arXiv cs.AI / cs.LG / cs.CLSep 4, 16:38 UTC · 12d agoAI research

The 20 Most Common Passwords Hackers Target in 2026

Huntress details the 20 most common passwords of 2026 and how attackers use brute force, spraying, and credential stuffing against weak credentials.

Huntress published an awareness piece based on NordPass's seventh annual list of the 200 most commonly used passwords, compiled from exposed data in cyberattacks across 44 countries. The top passwords remain simple sequences and variants such as "123456", "admin", "password", and "P@ssw0rd", all crackable in under a second. The article explains four password attack types: brute force, password spraying, credential stuffing, and dictionary attacks. Huntress cites its own data showing more than 1 in 4 IT professionals consider employees' password habits their biggest weakness, and recommends avoiding common passwords, not reusing credentials, and combining letters, numbers, and symbols.

HuntressSep 10, 13:00 UTC · 6d agoPhishing & fraud

OmniMed-FL: A Robust Multimodal Federated Learning Framework for Clinical Diagnosis

OmniMed-FL benchmarks multimodal federated learning for chest radiograph diagnosis across 3-20 clients, with FedProx leading under severe non-IID skew.

OmniMed-FL studies multimodal federated learning combining chest radiographs and clinical notes for five-class condition classification under HIPAA/GDDR-compliant decentralized training. It benchmarks eight fusion strategies, imputation rules, and federated baselines under Dirichlet non-IID partitioning across 3-20 hospital clients. With 5 clients and severe skew (alpha=0.1), FedProx scored 0.737 macro-F1 versus 0.662 for FedAvg and 0.297 for local-only training. Multimodal fusion beat unimodal inputs (0.956 vs 0.934 text, 0.664 images) on the synthetic corpus.

arXiv cs.AI / cs.LG / cs.CLSep 9, 15:56 UTC · 7d agoAI research1

Zero trust AI agents demand a different kind of security

Teleport's Chris Webber argues zero trust must extend to AI agents through trusted runtimes with zero initial privileges and continuous per-action enforcement.

In an interview, Teleport VP of Product Marketing Chris Webber says point-in-time authentication and static least privilege fail for agents that act fast, unpredictably, and continuously, sometimes spawning dozens of clones with the credentials of the human who invoked them. Teleport Trusted Runtimes give each agent a unique attestable identity, zero starting privileges, and expiration after task completion to eliminate standing privilege and stored data. Teleport Identity Security monitors agent actions against declared objectives in real time, intervening up to termination and runtime destruction, replacing anomaly-based ITDR detection with continuous enforcement.

Help Net SecuritySep 7, 00:00 UTC · 9d agoAI safety & security

H Company Releases NeoMME: A Family of 260M and 800M Single-Tower Multimodal Encoders That Drop the Vision Tower and Causal Decoder

H Company released NeoMME, 260M/800M single-tower multimodal encoders matching 3.75B ColQwen2.5 on ViDoRe v3 while being 14.4x smaller, under Apache 2.0.

H Company released NeoMME, a family of 262,937,906- and 793,715,032-parameter bidirectional encoders that process text and raw 32x32 image patches in a single tower, pretrained via masked diffusion and released under Apache 2.0 with day-zero Hugging Face Transformers support. NeoMME-Retriever-260M reaches 0.523 nDCG@10 on ViDoRe v3, matching 3.75B-parameter ColQwen2.5 while being 14.4x smaller; the 800M model scores 0.556. Hierarchical token pooling with int8 and binary quantization shrinks late-interaction indexes from roughly 1.5 MB to 6 kB per page while retaining 95.19% of nDCG@10; text-only BEIR retrieval remains a weak spot.

MarkTechPostSep 6, 21:06 UTC · 9d agoAI research

Top 10 Best Cloud Infrastructure Entitlement Management (CIEM) Tools in 2026

2026 CIEM guide ranks Wiz, Prisma Cloud, Okta, Entra Permissions Management and specialists Sonrai, Britive, Tenable/Ermetic for cloud entitlement right-sizing.

Buyer's guide covers ten CIEM products across three market routes: CNAPP-bundled (Wiz, Prisma Cloud), identity-suite (Okta, CyberArk, SailPoint, Saviynt) and specialists (Sonrai, Britive, Tenable/Ermetic). It cites machine identities outnumbering humans 10:1 plus effective-permissions sprawl as core drivers, with JIT elevation as the fix. Notable consolidation includes Tenable acquiring Ermetic and Zscaler acquiring Canonic.

Cyber Security NewsSep 15, 06:30 UTC · 1d agoTools

FLAT: Resampling Image and Text into 1D Flexible-Length Aligned Transmodal Tokens for Retrieval and Generation

FLAT jointly trains a multimodal encoder with text-to-image and image-to-text decoders, producing flexible-length tokens that hit 83.1 GenEval on T2I after fine-tuning.

FLAT (Flexible-Length Aligned Transmodal representations) is a pre-training framework that jointly optimizes a shared multimodal encoder with T2I and I2T decoders, combining contrastive alignment with bidirectional cross-modal generative objectives. It maps visual and textual inputs into a unified continuous 1D sequence space and uses nested dropout over prefix-K tokens for dynamic output lengths. A single pre-training stage supports cross-modal retrieval and generation (71.1 GenEval), with task-specific fine-tuning reaching 83.1 GenEval on T2I, 40.5 BLEU-4 and 138.6 CIDEr on MS-COCO captioning, and strong Recall@5 on MS-COCO and Flickr30K.

Hugging Face daily papersSep 14, 20:00 UTC · 2d agoAI research

Introducing context-aware vulnerability discovery and remediation with Cloudflare Managed Defense and OpenAI Daybreak models

Cloudflare launches invitation-only Vulnerability Discovery and Remediation within Managed Defense, using OpenAI Daybreak models and WAF context to prioritize and patch vulnerabilities.

Cloudflare announced early access to Vulnerability Discovery and Remediation, an invitation-only service within Cloudflare Managed Defense. The service uses OpenAI Daybreak models, including GPT-5.6 Cyber, via the Daybreak Defense Network to hunt and validate vulnerabilities in customer-authorized codebases across Workers and proxied applications. Findings are prioritized using production traffic, WAF rule, and security event context, and proposed patches and WAF mitigations are automatically checked before customer review.

Cloudflare BlogSep 3, 21:03 UTC · 12d agoTools

Top 10 Best Endpoint Privilege Management (EPM) Tools in 2026

A 2026 scorecard ranks ten endpoint privilege management tools, led by BeyondTrust, ThreatLocker and Delinea for elevation, coverage and policy depth.

The article ranks ten endpoint privilege management (EPM) tools using weighted criteria covering elevation workflow, platform coverage, policy depth, time-to-value and value. BeyondTrust scored highest overall (8.4) for cross-platform breadth, with ThreatLocker (8.2), Delinea (8.1) and Admin By Request (8.0) highlighted for allowlisting integration, cloud administration and deployment speed respectively. It also notes that Netwrix acquired CoSoSys in 2024, which affects bundling when shortlisting both EPM and device control.

Cyber Security NewsSep 10, 07:58 UTC · 6d agoIndustry

Decomposition-Guided Diffusion Language Models for Inertial Confinement Fusion Prediction

ICF-DLM, the first language-model-based inertial confinement fusion predictor, cuts peak-timing error from 11.6 to 9.2 steps versus LLaMA-3-8B.

Each National Ignition Facility shot costs roughly one million dollars, motivating accurate AI surrogates for predicting 512-step neutron-rate waveforms from laser pulses and target parameters. ICF-DLM combines physics-typed decomposition into yield, peak timing, and local waveform; bidirectional denoising that defers commitment to peak location; and a physics-driven PPO reward. On ICFBench (50,000 simulations plus 232 experimental shots) it outperforms a matched autoregressive LLaMA-3-8B, classical sequence models, and LLM-based time-series predictors.

arXiv cs.AI / cs.LG / cs.CLSep 7, 17:00 UTC · 9d agoAI research

12 Best CIEM Tools Compared (2026): Features & Pricing

Buyer's guide compares twelve CIEM tools; Microsoft discontinued Entra Permissions Management, while Tenable (Ermetic), CyberArk, and Wiz lead the 2026 scorecard.

The scorecard evaluates twelve cloud infrastructure entitlement management vendors on permission analytics depth, JIT enforcement, non-human identity coverage, pricing predictability, and bundle leverage. Tenable (Ermetic) leads at 4.70, followed by CyberArk and Wiz, while Microsoft's retirement of Entra Permissions Management (CloudKnox) forces existing customers into migration cycles. Pricing structures span per-identity, per-resource, per-workload, credit-based, and quote-based models.

GBHackersupdated 09:52 UTC · 10h agofirst 07:37 UTC · 12h agoIndustry 14 sources

LandingAI Releases Agentic Document Extraction Gen2 with DPT-3 Pro and DPT-3 Verity

LandingAI shipped Agentic Document Extraction Gen2 with DPT-3 Pro and DPT-3 Verity parsing models, adding usage-based billing, block-tree outputs, and word-level grounding.

LandingAI has generally released Agentic Document Extraction Gen2, rebuilt around two parsing models: DPT-3 Verity for deterministic transcription of digital documents with per-word bounding boxes and confidence scores, and DPT-3 Pro for layout-aware parsing of scans, handwriting, non-Latin scripts, and LaTeX math. Billing changes from a flat 3 credits per page to a page-plus-output-character model (Pro: 1 credit/page plus 0.5 credits per 1,000 output characters on priority; Verity: 0.3 plus 0.2), with an asynchronous standard tier at 0.5x price and vendor-claimed 25-80% cost reductions. Parse v2 returns a document-page-block tree with semantic IDs, normalized bounding boxes, and line- or word-level atomic grounding, replacing flat chunks; Gen1 client code will not run against Gen2 endpoints. Deployment options include US/EU cloud, VPCs on AWS, Azure, and Google Cloud, Snowflake, and air-gapped on-premises environments, with automated model routing planned for fall 2026.

MarkTechPostSep 10, 02:13 UTC · 6d agoAI tools & infra

Studying Image Tokenizers as Visual Languages in Unified Multimodal Models

A controlled pure-autoregressive testbed shows task-specific validation losses rank image tokenizers differently, with I2T loss the most consistent signal.

Researchers built a controlled pure-autoregressive testbed and tracked task-specific validation losses during multimodal continual pretraining across text, image, text-to-image (T2I), and image-to-text (I2T) prediction. They find losses should be analyzed per task because they exhibit distinct scaling behavior and rank tokenizers differently, and that the loss-performance relationship depends on the predicted token space. I2T loss, computed over a shared text vocabulary, correlates consistently with both generation and visual understanding performance after supervised finetuning. Case studies revisit the discriminator, semantic supervision, and vocabulary size as tokenizer design axes.

arXiv cs.AI / cs.LG / cs.CLSep 8, 17:57 UTC · 8d agoAI research1

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.CLSep 7, 17:26 UTC · 9d agoAI research1

Risky Bulletin: Dutch intel services to get extensive new powers

Netherlands proposed a bill granting AIVD and MIVD expanded warrantless tapping, faster hacking powers, and forced data disclosure, citing Russia, China, and Iran threats.

The Dutch government introduced a bill greatly expanding surveillance powers of intelligence agencies AIVD and MIVD, allowing up to one year of tapping without pre-approval and simplified hacking operations against 'foreign adversaries'. Agencies could compel Dutch companies or citizens to provide data under threat of charges, share data with the private sector, and oversight bodies would merge into a new CTT board. The bill follows similar overhauls in Ireland, Germany, and France after Russia's invasion of Ukraine. The newsletter also reports Moonwell hacked for $8.7M, a Cosmos EVM bug exploited for ~$3M, ShinyHunters listing McKesson with claimed hundreds of millions of records, and a pro-Kremlin DDoS claim against Norway's government network.

Risky Business NewsAug 31, 03:25 UTC · 16d agoPolicy & legal

NIST Seeks Public Input on AI-Ready NVD Modernization

NIST is seeking public comment on modernizing the National Vulnerability Database to support AI-powered vulnerability research.

The US National Institute of Standards and Technology announced it is soliciting public input on modernizing the National Vulnerability Database. The initiative aims to make the NVD AI-ready to support AI-powered vulnerability research and analysis.

Infosecurity MagazineAug 12, 12:30 UTC · Aug 12, 2026Policy & legal

A Princeton Researcher Proposes Recurrent Looped Transformer (RLT) that Carries Decoder State across Every Token, Fixing 96 Blocks per Token with Unbounded Temporal Depth

Princeton researcher Yifan Zhang proposes Recurrent Looped Transformer, carrying full decoder state across every token for unbounded temporal depth.

Yifan Zhang's technical report defines the Recurrent Looped Transformer (RLT), pairing a causal encoder with a recurrent decoder whose final output and layerwise sliding-window attention cache carry into every subsequent token with no prompt-response boundary reset. The reference configuration ties 48 encoder and 48 decoder layers, executing 96 logical blocks per token while the state path grows to 48t blocks after t tokens at fixed per-token compute. The report details RL replay contracts that rebuild all states under current parameters and exact prefix snapshots for multi-turn serving, but explicitly reports no measured efficiency, reasoning quality, or scaling results.

MarkTechPostSep 13, 17:02 UTC · 3d agoAI research1

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 papersSep 8, 20:00 UTC · 8d agoAI research

PIDS-Bench: Evaluating Prompt-Injection Detectors Under Over-Defense, Obfuscation, and Distribution Shift

PIDS-Bench shows prompt-injection detectors scoring F1 above 0.98 still misclassify about one-third of external benign security-adjacent prompts, revealing provenance-sensitive over-defense.

PIDS-Bench is a frozen multi-axis benchmark that jointly evaluates prompt-injection detectors on attack detection and benign false-positive behavior at fixed thresholds, spanning in-distribution inputs, hard-benign prompts, obfuscated attacks, and domain/structural distribution shifts. It evaluates seven detectors plus a rule-based lower-bound reference. A detector exceeding F1 = 0.98 on held-out data still misclassifies roughly one-third of an externally-sourced benign security-adjacent subset, and no internal detector reaches F1 >= 0.95 with hard-benign FPR <= 0.10 on the stress distribution. Hard-negative augmentation nearly eliminates over-defense on curated stress inputs but leaves it intact on externally-sourced prompts, a pattern termed provenance-sensitive over-defense.

arXiv cs.CRSep 14, 04:28 UTC · 2d agoAI safety & security

Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal

A self-distillation safety framework tunes narrow-boundary refusals in Qwen3-8B, raising target-domain refusal to 84.75% while cutting over-refusal from 15.20% to 5.20%.

The paper formulates narrow-boundary safety, where deployments need refusals within specific topics rather than whole subjects, and proposes an offline self-generated framework with controlled topic generation, escalating retries, and harmful-benign boundary pairs. On political persuasion with Qwen3-8B, the method raised target-domain refusal from 9.47% to 84.75% and cut the mean unsafe-response rate across three broader benchmarks from 26.26% to 0.14%. Verified target-model responses reduced over-refusal from 15.20% to 5.20%, and boundary-pair data cut comply-side over-refusal on held-out pairs from 32.94% to 4.16%. Results show data composition controls the safety-usability trade-off and alignment should be evaluated on both sides of the refusal boundary.

Hugging Face daily papersSep 2, 20:00 UTC · 14d agoAI safety & security1

Rethinking Heterogeneous System Disaggregation for Subquadratic Attention

SQD disaggregates LLM inference by quadratic versus subquadratic attention layers, improving energy efficiency up to 56% on heterogeneous systems versus GPU-only baselines.

SQD (SubQuadratic Disaggregation) splits decode not by operator type but by quadratic versus subquadratic attention, matching their distinct arithmetic intensity and memory footprints. For sparse attention LLMs it separates top-k selection (requiring full KV indexing) from top-k attention plus FFN; for linear and sliding-window models it separates dense attention layers from subquadratic layers plus FFN. On an adjusted 8xB200 heterogeneous proxy, tokens-per-joule improves 53% on GLM 5.2, 31% on Nemotron 3 Ultra, and 56% on Gemma 4 31B. A Rubin plus LPX analytical model shows 1.2x-1.5x tighter achievable latencies and up to 3.6x higher throughput versus attention-FFN disaggregation.

arXiv cs.AI / cs.LG / cs.CLSep 11, 17:55 UTC · 5d agoAI research

Perturbation Probing: A New Diagnostic for the Fragility of LLM Safety

Unit 42 research shows LLM safety refusals concentrate in a thin neural layer, motivating external, multi-layered AI security controls.

Palo Alto Networks Unit 42 introduces Perturbation Probing, a diagnostic technique for measuring the fragility of LLM safety mechanisms. The research finds that safety refusal behavior is localized within a thin neural layer, implying small perturbations can undermine built-in refusals. The authors argue this motivates external, multi-layered security defenses on top of model-internal safety training.

Palo Alto Unit 42Aug 28, 22:00 UTC · 18d agoAI safety & security

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 papersSep 7, 20:00 UTC · 9d agoAI research1

TokenRhythm/NeoHorse-1-4B — new model trending #30 on Hugging Face

TokenRhythm releases NeoHorse-1-4B, an Apache-2.0 agentic fine-tune of Qwen3.5-4B claiming +5.93 benchmark macro-average gain.

NeoHorse-1-4B is a roughly 4B-parameter text-only causal language model post-trained by TokenRhythm from Qwen/Qwen3.5-4B for agent harnesses, tool use, coding, and instruction following. It applies routing-guided curriculum SFT and routing-guided on-policy distillation over execution trajectories as an early prototype toward recursive self-improvement (RSI). The release reports a 64.87 macro average across ten benchmarks versus 58.94 for Qwen3.5-4B (+5.93) and is distributed under Apache-2.0, trending #30 on Hugging Face.

Hugging Face trending modelsSep 5, 08:21 UTC · 11d agoModel release1

How Fragile Is Safety Alignment at Frontier Scale? A Single-Direction Attack on a 320B MoE

Researchers show directional ablation breaks refusal in GLM-5.3-Flash, a 320B-parameter MoE, cutting refusal by 41–89 points across seven benchmarks.

The study extends directional ablation, a white-box attack that removes an aligned LLM's refusal behavior, from dense models up to ~70B parameters to GLM-5.3-Flash, a 320B-parameter mixture-of-experts model with 288 routed experts, four-wide hyper-connection residual, and block-FP8 quantization. Editing attention, dense, and routed-expert writers jointly removes 0.776 of refusal, with 74% of the effect existing only under the joint intervention; the conventional module-name-based recipe reaches only 0.066 and fails silently on MoE architectures. The attack yields 41–89 percentage-point reductions in refusal across seven harmful benchmarks with no detected capability change, and a category-concentrated refusal residue survives all edits at ranks 1 to 12.

arXiv cs.CRSep 9, 06:48 UTC · 7d agoAI safety & security