ZeroHour
DashboardNewsCVEsExploitsIncidentsAIBriefAnalyticsEntitiesWatchlistIOCsAsk the feedSourcesFeeds & APIKiosk mode

Search: “Kevin Beaumont”

40 stories

Searchkeywordsmart
DomainAllCyberAIWindow24h3d7d30dAllIn the wildImportant onlyShow duplicates
CategoryAllRansomware 91Vulnerability 87Exploit / PoC 76Data breach 27Malware 26Threat actor 14Industry 5Advisory 2Policy & legal 2Research 1Phishing & fraud 1
SeverityAllcritical 193high 95medium 37low 7info 0

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

ClickFix attacks are tricking Mac and Windows users into hacking themselves

ClickFix campaigns trick users into pasting terminal commands via fake CAPTCHAs; latest wave abused HBO Max's hacked Reddit account to push infostealer ads.

ClickFix attacks use fake CAPTCHA or anti-bot prompts on hacked or counterfeit websites to convince victims to copy a command into Windows Command Prompt or macOS Terminal, instantly installing infostealers that steal passwords, logged-in sessions, and crypto wallets. The latest campaign compromised HBO Max's official Reddit account to post hundreds of fake ads linking to a spoofed HBO Max page carrying the lure, per researchers at Hudson Rock. It is unclear how many users were compromised. Because execution happens in the terminal, many attacks evade antivirus tools; enterprises can block terminals domain-wide and Mac users can deploy BlockBlock.

TechCrunch · Security · 2d agoPhishing & fraud in the wild1

Necessary or Sufficient? Evaluating LLM Explanations With Behavioural Evidence

New framework tests whether LLM-cited explanation factors are necessary or sufficient, finding weak correlation across Claude, GPT, and Gemini models.

An arXiv paper introduces black-box intervention tests measuring whether factors LLMs cite in their explanations are necessary or sufficient for their outputs in agent oversight workflows. Across eight models from the Claude, GPT, and Gemini families, Spearman correlations between cited rankings and measured influence ranged from 0.349-0.354 (advisor recommendation) to 0.431-0.580 (prompt monitoring). Uncited factors scored above the lowest cited factor in up to 57.6% of advisor responses, showing cited top-three factors do not reliably identify the most influential inputs.

arXiv cs.AI / cs.LG / cs.CL · 12d agoAI safety & security1

Researcher Publishes CrowdStrike Privilege Escalation Zero Day

Researcher NightmareEclipse published FalconFlank, a zero-day local privilege escalation in CrowdStrike Falcon Sensor abusing its Office malicious macro remediation feature.

A researcher known as Nightmare Eclipse published details of FalconFlank, a zero-day privilege escalation that abuses the Microsoft Office malicious macro removal feature in CrowdStrike Falcon Sensor, verified working on fully updated Windows 11 25H2 and Windows Server 2025 with Falcon Phase 3 Optimal Protection. CrowdStrike urged customers to disable the Office File Suspicious Macro Removal policy while it investigates, noting customers remain protected via Cloud Anti-malware for Microsoft Office Files; no CVE has been assigned. The same researcher previously released the Exploitarium dump of over 30 PoC exploits, and researcher Kevin Beaumont confirmed FalconFlank works.

Infosecurity Magazine · 9d agoExploit / PoC

Turn it off and on again, but for critical infrastructure

KTH researchers trained a reinforcement-learning intrusion response agent on an emulated segmented OT network that autonomously resets hosts and processes to disrupt intruders.

Researchers at KTH Royal Institute of Technology built a containerized replica of a segmented industrial network, attacked it across 14 days, and captured 40,000 30-second traffic intervals to train a defense agent under partial observability. The agent observes six packet-count numbers per interval, maintains 500 running state hypotheses, and can reset supervisory hosts, water tank processes, or entire subnets, with resets rebooting the target, renewing credentials, and changing its IP. The best agent approached a full-visibility baseline but depends on an assumed attacker behavior model; the testbed comprised three supervisory hosts, two PLCs, two tanks, weak credentials, and CVE-2017-7494 exposure. The team released its implementation and plans validation on a real industrial testbed with a partner.

Help Net Security · 2d agoResearchCVE-2017-74942· 1 read

Serial Microsoft 0-day hunter drops yet another Defender exploit

Researcher Nightmare Eclipse released ShieldCrash, a Microsoft Defender zero-day PoC that bypasses September patches and reads files as SYSTEM.

Zero-day researcher Nightmare Eclipse (MSNightmare) published ShieldCrash, their 11th Microsoft zero-day, a proof-of-concept bypass of the ShieldBreak patch (CVE-2026-69414), which itself had bypassed the fix for RoguePlanet (CVE-2026-50656). ShieldCrash works on Windows 10, Windows 11, and Windows Server systems that have applied the September 2026 patches and allows arbitrary file reads as SYSTEM, though not arbitrary writes or a full SYSTEM shell. Microsoft has not said when it plans to patch the issue. The researcher recently also released zero-days affecting CrowdStrike Falcon (FalconFlank), Kaspersky endpoint antivirus (HardBreacher, patched), and Gen Digital's Avast (PrettyPrague).

The Register · Security · 7d agoExploit / PoCCVE-2026-69414CVE-2026-50656

The Model Proposes, the Code Disposes: A Pre-Registered Ablation of a Verifier-and-Acceptance Stage in an LLM-Orchestrated Offensive-Security Agent

Pre-registered ablation finds a model verifier stage in an LLM offensive-security agent suppresses findings; removing it eliminated suppression with precision tradeoff.

The paper evaluates a verifier-and-acceptance stage in an LLM-orchestrated offensive-security agent via a pre-registered 20-run confirmatory ablation and a 2x2 factorial study with 40 runs on vulnerable lab targets. Removing the stage eliminated pre-report suppression (median 2 vs 0 findings, p = 0.00003) but reduced model-blinded shipped precision (0.471 vs 0.353, p = 0.0087). Suppression was attributed to the model verifier rather than deterministic acceptance rules, and an instrumented canary recorded zero external contacts in all 60 runs. The full design retained 93.8% of model-adjudicated true candidates but failed its pre-registered non-inferiority floor of 0.90.

arXiv cs.CR · 2d agoResearch

Pro-Palestinian operation claims dozens of data breaches against Israeli firms

Hacktivist group Cyber Toufan claims 60 data breaches of Israeli and allied firms, wiping systems and erasing backups; Check Point links it to Iran.

The pro-Palestinian group Cyber Toufan said it released stolen data from 60 Israeli and foreign firms, including SpaceX, Toyota and IKEA, as part of a month-long leak operation launched in late November. Researchers at Check Point and SOC Radar assess the leaks are genuine, likely stemming partly from a major attack on Israeli hosting company Signature-IT, and attribute the group to Iran; researcher Kevin Beaumont said roughly a third of victims remain offline weeks later with backups erased. Google blocked the group's Telegram leak channel, while about 10 Iranian-backed hacking groups are assessed to be attacking Israel in the ongoing cyberwar.

The Record · 9d agoThreat actor in the wild

Biology-in-the-loop: Amortized Adaptive Hit Discovery in CRISPR Screens

Researchers release AssayBench-Loop, a 1,389-screen CRISPR benchmark, and AssayLoop, a framework that learns adaptive hit discovery policies.

The paper introduces AssayBench-Loop, a large-scale benchmark of 1,389 CRISPR screens across five phenotype categories for adaptive hit discovery under budget constraints. It also introduces AssayLoop, which combines AssayFormer, a transformer-based amortized acquisition policy trained across historical screens, with LLM-derived biological priors via an adaptive handoff. On temporally held-out screens, AssayLoop achieves 5.67-fold enrichment over random selection and recovers 27.7% of hits after assaying roughly 5% of the candidate library, outperforming existing adaptive-design methods and standalone LLMs.

arXiv cs.AI / cs.LG / cs.CL · 6d agoAI research1

Prolific Microsoft 0-day hunter drops CrowdStrike Falcon exploit PoC

Researcher Nightmare Eclipse released a PoC for FalconFlank, a privilege escalation zero-day in CrowdStrike Falcon's Office macro remediation feature.

Researcher Nightmare Eclipse published a zero-day PoC dubbed FalconFlank targeting CrowdStrike Falcon's Microsoft Office malicious macro removal feature, achieving privilege escalation on fully updated Windows 11 25H2 and Windows Server 2025 with Optimal Protection and macro removal enabled. Kevin Beaumont independently confirmed the exploit works. The same researcher also released HardBreacher, an elevation-of-privilege bug in Kaspersky Endpoint (since fixed via update), PrettyPrague, an Avast Sandbox flaw that dumps the SAM database for a SYSTEM shell (Gen Digital is developing a patch), and GreenSection, an Nvidia memory corruption bug that crashes systems. CrowdStrike advises customers to disable the Suspicious Macro Removal Windows policy while it investigates.

The Register · Security · 13d agoExploit / PoC

When LLM judges agree, should we believe them?

Amazon ICML paper uses Ising models to correct correlated LLM-judge votes, beating accuracy-weighted panels by 9-14%.

Amazon Science describes an ICML paper, "Dependence-aware label aggregation for LLM-as-a-judge via Ising models," addressing how correlated judge outputs inflate majority-vote confidence. The unsupervised method models pairwise dependence between judges, learning both reliability and similarity without human reference labels. Tested on relevance, toxicity, and summarization tasks with 10 judge models at temperature zero, it outperformed accuracy-weighted voting by 9% to 14%.

Hacker News · AI · 2d agoAI researchHN 41↑ · 23 comments

ShieldBreak Zero-Day PoC Claims Microsoft Defender Patch Bypass With SYSTEM Access

Researcher Chaotic Eclipse released ShieldBreak PoC, a claimed patch bypass for Microsoft Defender flaw CVE-2026-50656, now tracked as CVE-2026-69414.

Researcher Chaotic Eclipse released a PoC for ShieldBreak, a claimed full patch bypass of Microsoft Defender flaw CVE-2026-50656 (RoguePlanet), which can yield SYSTEM privileges. Independent researchers Kevin Beaumont and Will Dormann validated the exploit on Windows 11 25H2 and Windows Server 2025, noting it requires Defender to be enabled. Microsoft assigned CVE-2026-69414 (CVSS 7.8) with an 'Exploitation More Likely' assessment and is investigating, but has not yet released a patch. The article also notes CISA's KEV addition of the actively exploited Windows AFD.sys flaw CVE-2026-68820.

The Hacker News · Aug 17, 2026Exploit / PoCCVE-2026-50656CVE-2026-62832CVE-2026-68820+2 CVEs

K-Bench: a clinically calibrated benchmark for evaluating large language models in high-risk mental health conversations

Clinician-calibrated K-Bench evaluates 125 LLM configurations on 200 high-risk mental health vignettes, exposing wide variation in suicide and violence risk handling.

K-Bench is a clinician-calibrated, protected benchmark evaluating 125 model configurations from 33 base models across 14 providers on 200 multi-turn vignettes covering suicide, self-harm, domestic violence, substance misuse and no-risk presentations. A frozen GPT-4o judge achieved 94.2% exact agreement with clinician consensus across 6,751 eligible comparisons from 151 clinician-rated transcripts. Leading models combined supportive conversation with combined-risk scores above 95, while risk exploration varied substantially among weaker configurations; therapeutic prompting helped weaker models and elevated reasoning produced no average improvement. A continuously updated public leaderboard is hosted at k-bench.ai with protected test materials.

arXiv cs.AI / cs.LG / cs.CL · 2d agoAI safety & security

153GB of stolen credentials surface after LiteLLM supply chain attack

153GB archive from the LiteLLM supply chain attack exposes secrets from about 2,500 companies, including AWS, Samsung, Cisco and Salesforce.

Hudson Rock obtained and analyzed a 153GB archive stolen in the LiteLLM supply chain attack, containing 433,909 files with secrets attributed to 2,488 corporate domains, including AWS, Samsung, Cisco, Salesforce, NVIDIA, Microsoft and Siemens. TeamPCP compromised Trivy on March 19, 2026, stole LiteLLM's PyPI publishing tokens through the build pipeline, and published malicious LiteLLM versions 1.82.7 and 1.82.8 on March 24. CloudSEK separately estimated close to 2,500 exposed organizations, and Kevin Beaumont confirmed the data is legitimate. Hudson Rock is running a global ethical disclosure effort and urges organizations to rotate secrets before the trove leaks publicly.

Help Net Security · Aug 13, 2026Data breach in the wild

One Symptom, Three Levers: A Critical Review of On-Policy Self-Distillation

A review paper frames on-policy self-distillation collapse as governed by three levers: token weighting, privileged information, and guidance decay.

The paper critically reviews On-Policy Self-Distillation (OPSD), where a language model trains on its own generations scored token-by-token by a teacher conditioned on privileged information such as reference solutions or environment feedback. It identifies collapse, the progressive narrowing of producible reasoning paths, as the dominant failure mode and analyzes it through three levers: signal weighting, the nature of privileged information, and teacher dynamics. The review is restricted to mathematical reasoning, reports no new experiments, and offers a shared vocabulary separating settled findings from disputed ones.

Hugging Face daily papers · 22d agoAI research

Underwriting Superintelligence: Backing Agents you can Sue — Rune Kvist, AIUC

AIUC raised a $40 million Series A to build AIUC-1, an agent security standard backed by insurance, serving Cursor, Harvey, Lovable, and ElevenLabs.

AIUC, cofounded by former Anthropic product hire Rune Kvist, announced a $40 million Series A led by Ribbit Capital and First Harmonic. The startup builds AIUC-1, an emerging standard for agent security, safety, and reliability, stress-testing agents for jailbreaks, hallucinations, and data leaks. It pairs standards with insurance underwriting through Lloyd's of London and counts Cursor, Harvey, Lovable, and ElevenLabs among its customers. Kvist argues trust and liability, not capability, are becoming the binding constraint on AI adoption.

Latent Space · 8h agoAI industry 2 sources

Microsoft patches 400+ vulnerabilities, one zero-day under attack (CVE-2026-68820)

Microsoft's August 2026 Patch Tuesday fixes 400+ vulnerabilities, including an actively exploited Windows zero-day (CVE-2026-68820) used by North Korean attackers.

Microsoft's August 2026 Patch Tuesday fixes over 400 vulnerabilities, including CVE-2026-68820, a use-after-free in the Windows Ancillary Function Driver exploited in the wild by North Korean actors deploying a kernel-mode rootkit in Operation Dream Job. Critical unauthenticated remote code execution flaws in Microsoft QUIC (CVE-2026-62815) and Windows DNS (CVE-2026-62878) were also patched, alongside a SharePoint RCE chain combining CVE-2026-63520 with CVE-2026-55040. Researcher Nightmare-Eclipse released ShieldBreak, a PoC bypassing the July RoguePlanet Microsoft Defender patch (CVE-2026-50656), confirmed working by Will Dormann on Windows 11.

Help Net Security · Aug 12, 2026Exploit / PoC in the wildCVE-2026-68820CVE-2026-62832CVE-2026-72971+6 CVEs1

Ask Before You Optimize: Dynamic Pre-Formulation Clarification for Interactive Optimization

Researchers release OR-Clarify, a benchmark testing whether LLM agents ask clarifying questions before formulating optimization models from incomplete requests.

OR-Clarify evaluates pre-formulation clarification in operations research: each task gives a partial problem description, withholds structured hidden slots, and scores agents via bounded interaction with a simulated user, measuring slot recovery, stopping behavior, silent assumptions, and interaction cost. The authors also propose InterOPT, a two-stage framework that identifies formulation-critical gaps to decide when to ask or stop. In choice-based experiments InterOPT substantially outperforms all baselines in exact slot recovery and remains competitive in the open-ended setting.

Hugging Face daily papers · 13d agoAI research1

Nightmare Eclipse Drops CrowdStrike, Nvidia, Avast Zero-Day Exploits

Researcher Nightmare Eclipse released working zero-day PoCs for Avast, CrowdStrike Falcon, and Nvidia, prompting a Gen patch, CrowdStrike mitigation, and Nvidia investigation.

Security researcher Nightmare Eclipse released three zero-day exploits within a short window: PrettyPrague targeting the Avast sandbox for full system privileges (possibly affecting other GenDigital products including AVG and Norton), FalconFlank exploiting CrowdStrike Falcon Sensor's Office malicious macros remediation feature for privilege escalation, and GreenSection targeting an out-of-bounds memory write in a shared global memory section used by Nvidia user-mode components. GenDigital said it has fixed the Avast issue; CrowdStrike advised disabling the Microsoft Office File Suspicious Macro Removal policy setting; Nvidia said it is actively investigating the PoC. Kevin Beaumont reported that the Avast, CrowdStrike, and Kaspersky exploits work. The researcher previously released the HardBreacher privilege escalation zero-day in Kaspersky endpoint security, patched August 31.

SecurityWeek · 9d agoExploit / PoC

PARSER: Read in Parallel, Reason in Depth for Long-Context LLM Agents

PARSER uses parallel reader subagents and an RL-trained lead agent for long-context QA, beating baselines and cutting latency up to 11x.

The PARSER paper decouples reading from reasoning: frozen subagents each read one document chunk in parallel while an RL-optimized lead agent iteratively broadcasts queries and aggregates evidence in scatter-gather rounds. On multi-hop QA with 7K to 896K token contexts, a 4B-backbone PARSER beats the strongest sequential memory baseline by 5.7 points on average and 12.0 points at 896K tokens, and a 9B version surpasses DeepSeek-V4-Pro by 6.3 points. Controlled experiments show robustness to evidence position, order, and distance perturbations, with inference latency reduced by up to 11x.

Hugging Face daily papers · 11d agoAI research1

One resignation turned the embers of AI fear into a wildfire

Interconnects argues a frontier-lab researcher's safety resignation went viral via media coordination, reigniting AI existential-risk discourse.

The essay analyzes why researcher Jacob Coxon's resignation over AI safety risks went viral, aided by a Wall Street Journal exclusive, advocacy-group amplification, and Daniel Kokotajlo's same-day Joe Rogan appearance. It notes Evan Hubinger's >10% extinction-risk figure and criticizes existential-risk discourse for conflating very different meanings of the term. The author assigns near-zero probability to complete extinction but argues concrete risks such as cyberattacks on critical infrastructure and bio-risks merit debate. It also rejects recursive self-improvement forecasts, proposing 'lossy self-improvement' where models excel at math and code but remain limited elsewhere.

Interconnects · 6d agoAI safety & security

ukisai/Swift-Qwen3.8-27b — new model trending #30 on Hugging Face

UkisAI releases Swift-Qwen3.8-27B, a Qwen3.8-27B derivative using 58.3% fewer thinking tokens with <1% performance loss and ~1.95x speed-up.

UkisAI released Swift-Qwen3.8-27B, a reasoning-efficient derivative of Qwen3.8-27B that cuts thinking-token usage by 58.3% while staying within 1% of base performance, yielding a 1.95x speed-up on several tasks. The model was fine-tuned by penalizing reasoning-marker tokens that trigger overthinking, plus a transfer component from BottleCap AI's ThinkingCap-Qwen3.6-27B. Benchmarks include GPQA-Diamond 88.28% (base 88.38%), MMLU-Pro 84.95% (base 85.47%), and AIME 2026 94.00% (base 98.67%), with mean-token reductions of roughly 27-46% across tests. GGUF weights are available on Hugging Face alongside enterprise licensing options.

Hugging Face trending models · 8d agoModel release

From Reweighting to Rewriting: Unlocking the Intervention Effects of Influential Samples in Training Data Attribution

Study shows rewriting responses of influence-selected training examples shifts LLM behavior more strongly than reweighting the same samples.

The paper examines training data attribution, arguing that influence functions identify high-leverage examples whose value goes unrealized under conventional weight-based reweighting interventions. It introduces influence-guided response rewriting, which replaces the responses of influence-selected examples with behavior-aligned or behavior-opposed supervision while keeping instructions fixed, tested across four open-weight LLMs using epistemic abstention as the primary testbed. Rewriting produces stronger, more persistent, and bidirectional behavioral shifts, including on safety refusal, while reweighting the same examples yields weak, inconsistent effects. The results motivate intervention-aware evaluation of TDA methods.

Hugging Face daily papers · 15d agoAI research

Lightning Weave: Improving the Accuracy-Efficiency Frontier of Reasoning Models through Capability Composition

Lightning Weave composes capabilities from independently post-trained models via on-policy distillation, improving Qwen3.5-4B reasoning accuracy while cutting tokens.

Lightning Weave is a post-training framework that merges accuracy and efficiency capabilities from independently post-trained specialist models into a single student via on-policy distillation. Each capability is represented as a policy shift, combined via aligned log-ratio shifts and Tilted-Target DOPD, enabling training without serving multiple live anchor models concurrently. On Qwen3.5-4B, it raises HMMT 2025 accuracy from 59.2% to 64.0% with 10.7% fewer response tokens, and LiveCodeBench v5 accuracy from 41.7% to 54.2% with 9.6% fewer tokens. The authors report a state-of-the-art accuracy-efficiency Pareto frontier across diverse students and math/code benchmarks, with code planned for release.

Hugging Face daily papers · 4d agoAI research

Domain-Incremental Learning for Multi-Channel Replay Speech Detection

First continual learning benchmark for multi-channel replay speech detection shows task-specific beamforming cuts catastrophic forgetting across 24 acoustic environments.

Researchers frame replay-attack detection for voice-controlled systems as domain-incremental learning over acoustic environments, evaluating a beamformer-based detector across all 24 environment orderings of the ReMASC corpus with five seeds. Naive sequential fine-tuning raises error rates on previously learned environments by 18.8 points, while elastic weight consolidation halves forgetting but loses plasticity and gradient projection memory is statistically indistinguishable from naive fine-tuning. A task-specific beamformer keeping one spatial front-end per environment significantly improves final and incremental accuracy, and the last environment in a sequence dominates final performance.

arXiv cs.CR · 6d agoResearch1

Forgetting Without Restarting: Execution-State Unlearning for Stateful LLM Agents

Researchers propose provenance-guided selective replay letting LLM agents forget revoked information without restarts, matching full reset behavior.

The paper formalizes execution-state unlearning for stateful LLM agents, requiring that agents behave as if a revoked memory record was never observed across transcripts, compressed memory, tool plans, and KV caches. It proves exact unlearning requires at least T-τ+1 recomputed transitions and that Provenance-Guided Selective Replay attains this bound via a provenance graph, KV cache cropping, and sanitized replay. In audits across three agent suites, nine baselines, and three model families, memory deletion left leakage unchanged, instruction-based forgetting collapsed under elicitation (Leak@probes = 1.00), and selective replay matched full resets at up to 9x fewer recomputed tokens.

arXiv cs.CR · 12d agoAI safety & security1

Jackrong/Qwopus3.8-27B-Flash-GGUF — new model trending #26 on Hugging Face

Community fine-tune Qwopus3.8-27B-Flash, built on Qwen3.8-27B, cuts agent reasoning latency with 12.8% faster decoding and 80.7% MTP acceptance.

Jackrong released Qwopus3.8-27B-Flash, a fine-tune of Qwen3.8-27B optimized for long-running agent workloads, reporting 12.8% faster decoding and 80.7% multi-token-prediction acceptance. Training used roughly 1.5 million teacher-scored SFT examples filtered to the top 10%, followed by reinforcement training with NVIDIA NeMo-RL and GSPO. The author notes an explicit trade-off: MMLU-Pro mixed-set scores are lower than the base model, and a known bug can produce incorrect Python indentation. Author-provided benchmarks have not been independently verified.

Hugging Face trending models · 12d agoModel release1

ActReview: Rebuttal-Guided Training Data and Rubric Rewards for Actionable Peer Review Generation

ActReview post-trains Qwen3-8B-Base on 40K rebuttal-derived instances with rubric rewards to generate actionable, grounded peer-review feedback, plus a 1,000-instance benchmark.

The framework builds ActReview-40K from real OpenReview review-rebuttal threads, aligning reviewer weaknesses with author responses and grounding feedback in localized paper evidence. Qwen3-8B-Base is post-trained with multi-task supervised fine-tuning followed by GRPO using candidate-aware, weakness-specific rubric rewards. Experiments show improved actionability and grounding over prior specialized review-generation models, supported by ActReview-Bench, a human-curated 1,000-instance evaluation set. Human evaluation confirms better revision usefulness while noting a remaining gap in technical accuracy.

arXiv cs.AI / cs.LG / cs.CL · 8d agoAI research

Competence-Gated Pooling of Language Models and Priors for Event Forecasting

Paper proposes a competence gate pooling language model forecasts with external priors, improving Brier score from 0.0771 to 0.0732 across 2,357 binary questions.

The paper defines a language model's relative competence as its marginal value beyond an available external forecast, and derives conditions under Brier loss where model disagreement improves that forecast. A competence gate estimates domain-level source weights from resolved outcomes, shrinks uncertain estimates toward a global weight, and recalibrates the pooled forecast. Across 2,357 resolved binary questions and five language models, it improves the external baseline from 0.0771 to 0.0732 Brier and beats global forecast combinations, though it defers to the market on ForecastBench. Across four Qwen models, verbal confidence failed to identify when the model outperformed the external forecast, while outcome-estimated competence supported better abstention.

Hugging Face daily papers · 7d agoAI research

Building a risk-based vulnerability management program that scales

Asimily CEO Shankar Somasundaram outlines a risk-based vulnerability management approach using inventory, attack paths, KEV and EPSS data.

In a Help Net Security video, Asimily CEO Shankar Somasundaram argues patching everything is infeasible as AI-driven attacks inflate vulnerability counts, with one customer finding a thousand unknowns for each known one. He recommends building a full inventory of devices, applications, and data flows, mapping attack paths for reachability, and prioritizing with KEV, EPSS, and business impact. Mitigations include patching, virtual patching via NACs and firewalls, segmentation, and configuration snapshots to detect drift.

Help Net Security · 23d agoIndustry

Former OpenAI researcher builds an AI model that judges options instead of writing text

TypeSafe AI launches Jev, a judgment-only model built by ex-OpenAI staff that classifies inputs with 70-500 ms latency instead of generating text.

Startup TypeSafe AI, co-founded by former OpenAI researcher and InstructGPT co-author Diogo Almeida, introduced Jev, a model that scores developer-defined answer options with probabilities rather than generating free-form text. The company claims 70-500 ms responses, parallel multi-question evaluation, and $0.042 per million input tokens with free outputs, targeting request routing, sales intent scoring, and assistant guardrail checks. Benchmarks are self-built and not independently verified, the 'no hallucination' guarantee only covers output structure, and access is currently via waitlist.

The Decoder · 11h agoAI industry

A*-Thought-V2: Efficient Latent Reasoning via Geometric Dynamics of LLM

A*-Thought-V2 compresses chain-of-thought into latent tokens using geometric hidden-state dynamics, cutting computation while improving accuracy on Qwen models.

A*-Thought-V2 models chain-of-thought as a hidden-state trajectory and interleaves explicit text with continuous latent tokens, compressing steps whose transitions deviate from the question-to-solution direction. Trained via stepwise embedding forcing and label forcing with soft multi-modal supervision, it was evaluated on Qwen3.5-9B and Qwen3.6-27B across six benchmarks. Reported results include up to 2.6% average accuracy gain, up to 50% shorter responses, 2.29x higher Accuracy per Computation Unit, 94.6% faster preprocessing, and up to 80.3% faster training.

Hugging Face daily papers · 9d agoAI research

Distill Globally, Adapt Locally: Reasoning Distillation and Product-Type Test-Time Training for Scalable Trade-Up Recommendation

A distillation framework compresses LLM reasoning into a 15.5M-parameter trade-up recommendation model reaching AUC 0.941 with product-type test-time training.

The paper targets trade-up recommendation, which identifies higher-quality alternatives that preserve customer purchase intent. A retrieval-augmented few-shot LLM teacher generates labels and rationales that supervise a compact embedding-pair classifier; at inference the 15.5M-parameter student uses only two precomputed 768-dimensional embeddings with no LLM calls. On 8,352 annotated pairs, label-only training scored AUC 0.912, reasoning distillation reached 0.924, and product-type test-time training lifted it to 0.941 with average precision 0.940. The distilled student is roughly 5,000x faster and 10,000x cheaper than direct LLM inference on a 100K-pair proxy catalog.

arXiv cs.AI / cs.LG / cs.CL · 12d agoAI research

The Pelican comparison grid for Astra is pretty interesting

Simon Willison's pelican SVG comparison shows GPT-6 Astra producing markedly better images than GPT-5.6 Sol, Terra, and Luna across reasoning levels.

Willison generated pelicans-riding-bicycles SVGs with newly accessed GPT-6 Astra at low through max reasoning levels and rendered them in a comparison grid against GPT-5.6 Sol, Terra, and Luna. Astra's outputs were markedly more coherent, while even the best GPT-5.6-Sol images remained largely abstract shapes. Astra does not support a reasoning=none setting, so all comparisons involved reasoning-enabled runs.

Simon Willison · 12d agoAI research

ToolLoop: Closed-Loop Tool-Use Data Synthesis via Decomposed Generation and Dynamic Self-Feedback

ToolLoop introduces a closed-loop synthetic data framework whose 11K examples lift a 4B model to 86.40% on BFCL tool-use evaluation.

ToolLoop decomposes tool-use data synthesis into function-name sampling, backward derivation of user queries, and forward derivation of tool calls, with dynamic self-feedback at each stage. This shifts the paradigm from generate-then-filter to generate-verify-refine, reducing inefficient and imbalanced synthetic data. A 4B model trained on 11K synthetic examples reaches 86.40% accuracy on BFCL non-reasoning mode (86.07% in an Isolate variant excluding BFCL-overlapping functions) and 72.1% on ACEBench using only 18.3% of baseline training data.

arXiv cs.AI / cs.LG / cs.CL · 8d agoAI research1

AI safety panic goes mainstream after Anthropic researcher's warnings land on CNN and Fox News

Departing Anthropic researcher Jacob Coxon's AI extinction warnings reached CNN, Fox News, and US politicians, pushing AI safety debate mainstream.

Jacob Coxon, a departing Anthropic researcher, warned that self-improving AI poses an existential threat, earning coverage on CNN and Fox News plus attention from US politicians and a Joe Rogan episode. Safety researchers at Anthropic and OpenAI supported his views, and Paul Christiano warned humanity could permanently lose control of a superintelligence without better oversight. Christiano joined the OpenAI Foundation board and its Safety and Security Committee, though without voting rights. The article cautions that the extinction scenario remains extreme and contested, while citing rogue hacking agents from OpenAI and Anthropic as concrete demonstrations of AI risk.

The Decoder · 6d agoAI industry

Coding Agents Have Converged: Why the SWE-bench Leaderboard Can No Longer Order Its Top Entries, and What to Measure Instead

Audit of 254 SWE-bench submissions finds top coding-agent entries statistically inseparable, so small leaderboard gaps no longer establish rank.

The paper audits 254 SWE-bench submissions across four splits without running models. On Verified, the top two entries each resolve 396 of 500 instances, and exact paired McNemar tests separate none of the 29 adjacent top-thirty pairs at alpha=0.05. Within-model scaffold score ranges reach 29.8 percentage points, versus an 8.8-point spread among the top thirty. The authors release a five-step audit protocol and recommend reporting comparison-set-specific resolution and model-scaffold provenance.

arXiv cs.AI / cs.LG / cs.CL · 1d agoAI research

Everything in Moderation: Per-Domain Coverage Optima and Alignment-Resistant Domain Gaps in Multi-Domain Mid-Training

Controlled mid-training experiments on Qwen3-8B-Base find each domain has a 10-40% coverage optimum and domain gaps survive alignment SFT.

Using Qwen3-8B-Base (with a 4B replication) across five semantically rule-disjoint KOR-Bench domains, the authors train 30 data allocations spanning the five-domain simplex at five seeds each. All five domains show interior optima in the moderate 10-40% coverage band, and domain gaps persist after a fixed-budget compensatory SFT pass, which raises 116/120 cells yet bridges 0/240 pairs at a 5% threshold. Zero coverage collapses mid-training-only accuracy, though a FineWeb-Edu-only control shows the collapse is partly generic drift. The results argue mid-training data composition requires principled design rather than reliance on later alignment.

arXiv cs.AI / cs.LG / cs.CL · 8d agoAI research1