Search: “regularization”

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

Attention Quantization for Tabular Foundation Models

FP8 quantization of attention queries, keys, and values speeds tabular foundation model inference up to 1.7x with no accuracy loss.

The paper develops an FP8 quantization strategy targeting attention calculations (queries, keys, values) in tabular foundation models, arguing attention matters more than weight or KV cache quantization given their differing size and serving patterns versus LLMs. Aligning quantization error between test rows and training rows proves crucial, since misalignment causes drastic accuracy drops. A Triton kernel using explicit FP8 matrix multiplication achieves up to 1.7x speedup over regular 16-bit kernels, with no relevant accuracy loss on TabPFN-v3 and TabICLv2 across TabArena and BeyondArena benchmarks.

arXiv cs.AI / cs.LG / cs.CL·Sep 11, 16:31 UTC · 5d agoAI research1

28

Beyond the Perimeter: Building Resilience Against Cloud and SaaS Supply-Chain Attacks

ShinyHunters exploited an Oracle PeopleSoft zero-day to steal data and extort roughly 100 organizations, including the Council of Europe, for up to $2.3M.

Between May and early June 2026, the ShinyHunters group exploited a critical zero-day in Oracle PeopleSoft across about 100 organizations and 300 instances worldwide, per reports cited by The Register. Stolen records included employee and student personal data, payroll, tax, financial and health information, plus immigration and passport documents. AgentCypher.ai estimates extortion demands of $400,000 to $2.3 million per victim, typically in Bitcoin; the Council of Europe refused to pay. The article uses the incident to argue for Zero Trust, supply-chain risk management, rapid patching, encrypted distributed backups and defined recovery-time objectives.

Cyber Security News·Sep 12, 10:46 UTC · 4d agoData breach in the wild1

78

On the Regularization Landscape for the Linear Recommendation Models

Study shows leading linear recommendation models reduce to nuclear-norm or Frobenius-norm regularization, with two new closed-form low-rank solutions proposed.

The paper unifies top-performing linear recommendation algorithms under a single regularization framework, showing they effectively apply either nuclear-norm or Frobenius-norm regularizers. Nuclear-norm solutions have a rigid structure, are low-rank, and have closed form, while Frobenius-norm solutions are more expressive but full-rank or require hard-to-tune procedures such as ADMM. The authors derive two new low-rank, closed-form solutions that combine the advantages of both regularization families.

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:45 UTC · 6d agoAI research

22

Cybersecurity jobs available right now: March 10, 2026

Help Net Security's roundup lists open cybersecurity roles at BioNTech, AIG, ServiceNow and others across Europe, the Middle East and Canada.

A job-board roundup of cybersecurity openings including Associate Director Application Security at BioNTech (Germany), CISO at AIG (Israel), Cloud Security Professional at ServiceNow (Italy), and SOC/GRC, analyst, engineer and data governance roles in the UK, UAE, India, Canada and France. Roles span application security, cloud security, SOC operations, compliance and OT environments. Most listings are marked no longer accepting applications.

Help Net Security·Aug 28, 09:06 UTC · 19d agoIndustry

10

A Unified and Constrained View of Regularization-Based Robust Reinforcement Learning

Paper unifies regularization-based robust RL methods via new performance-gap upper bounds and jointly learned Lagrange multipliers.

The authors derive new upper bounds on the gap between nominal and worst-case deep RL policies, each expressible as an existing regularization objective plus a KL-divergence penalty. Robust training is reformulated as constrained optimization, where prior methods correspond to a fixed Lagrange multiplier. The multiplier is instead updated jointly with the policy, auto-tuning the regularization weight. Adversarial evaluations across several continuous control tasks validate the theory.

arXiv cs.AI / cs.LG / cs.CL·Sep 11, 16:45 UTC · 5d agoAI research1

20

Quantile-based Loss Filtering for Outlier-Robust Stochastic Gradient Descent

Quantile-k-Loss SGD filters corrupted component losses by quantile sampling, proving linear convergence while outperforming standard and min-k-loss SGD.

The paper proposes Quantile-k-Loss SGD (Q(k)L-SGD), a loss-filtering framework for finite-sum optimization with corrupted components that samples k losses per iteration and updates using an index from the lower empirical q-quantile. The authors prove linear convergence under standard convexity, requiring sample size to scale with the number of corruptions, plus a complementary small-sample probabilistic analysis. Experiments on polynomial regression, regularized logistic regression, and hinge loss show intermediate quantiles often outperform both standard SGD and min-k-loss SGD.

arXiv cs.AI / cs.LG / cs.CL·Sep 11, 16:38 UTC · 5d agoAI research

18

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

40

Why The Vulnerability Backlog Is About To Get Worse

Recorded Future analysis says AI-driven vulnerability discovery and faster weaponization will grow the triage backlog while shrinking defenders' response windows.

Disclosed vulnerabilities rose from roughly 21,000 in 2021 to nearly 50,000 in 2025, while Recorded Future assessed only 446 as actively exploited in 2025. VulnCheck found nearly 29% of 2025 KEV entries were exploited on or before CVE publication. The authors argue AI-assisted discovery and automated exploit development will multiply credible reports, cut disclosure-to-exploit time toward minutes, and force re-evaluation of medium-severity flaws as exploit-chain components.

Recorded Future·Aug 26, 00:00 UTC · 21d agoResearch

45

A Theoretical Analysis of Generalization Dynamics in Neural Networks under Gradient Descent with Weight Decay

Theoretical framework bounds generalization for gradient descent with weight decay, deriving conditions that explain delayed generalization and grokking.

The paper proves convergence of gradient descent with weight decay to a neighborhood of global minimizers of the empirical l2 loss for a broad class of neural networks. It decomposes population error into data, optimization, and prediction variation errors, deriving cellwise and layerwise approximate-homogeneity bounds on prediction variation along the training trajectory. The resulting necessary and sufficient conditions explain layerwise generalization differences and provide a theoretical characterization of grokking.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 16:59 UTC · 9d agoAI research

12

UBoatRAT Navigates East Asia

Unit 42 discovers UBoatRAT, a new custom RAT targeting South Korean and video-game industry personnel, delivered via Google Drive with GitHub-based C2 and BITS persistence.

Unit 42 identified UBoatRAT, a new custom remote access trojan first found in May 2017, whose initial version used a public Hong Kong blog service and a compromised Japanese web server for command and control. The latest variants target personnel or organizations related to South Korea or the video games industry, are delivered through Google Drive, and masquerade as Microsoft Word, Excel, or folder icons. The RAT checks for virtualization software and domain join, retrieves its C2 address from a Base64-encoded string in a GitHub-hosted file, uses a custom XOR-encrypted C2 protocol, and maintains persistence via Windows Background Intelligent Transfer Service (BITS) jobs that survive reboots.

Palo Alto Unit 42·Aug 17, 12:55 UTC · Aug 17, 2026Malware in the wild1

44

Understanding the Impact of Model Pruning on Long-Tail Forgetting and Explanation Reliability in Medical Imaging

Systematic study finds model pruning causes frequency-dependent long-tail forgetting in medical imaging and that gradient-informed methods best preserve explanations.

Across two long-tailed medical imaging datasets, two CNN architectures, four pruning methods, and sparsity up to 95%, the study measures predictive performance, explanation stability, and faithfulness. Rare classes degrade earlier and more severely than frequent ones, while explanation reliability depends mainly on the pruning strategy, with gradient-informed methods degrading least. Mechanistic analysis ties explanation collapse to loss of class-discriminative gradients rather than vanishing feature activations, recommending class- and explanation-aware evaluation of compression.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:44 UTC · 9d agoAI research

27

The Future of Deepfakes and the Decline of Reality (With Hany Farid)

GetReal cofounder and PhotoDNA developer Hany Farid discusses deepfake detection and how AI-generated media is blurring perceptions of reality.

404 Media's podcast episode features Hany Farid, cofounder of deepfake detection company GetReal and developer of the PhotoDNA perceptual hashing algorithm, discussing deepfakes and AI-generated images. The conversation reflects on how the technology has evolved since 2017 and speculates on where detection and perceptions of reality are headed. The episode is part of 404 Media's interview series for subscribers.

404 Media·Aug 17, 14:19 UTC · Aug 17, 2026AI safety & security

22

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.CL·Sep 8, 17:57 UTC · 7d agoAI research1

30

When will average people feel AI’s impact?

Interconnects essay argues AI's impact is still a rounding error for average people, comparing looming wage stagnation to Engels' pause.

An Interconnects essay argues that AI currently touches daily life far less than previous industrial revolutions, since its benefits are concentrated in knowledge work and lack tangible consumer goods. The author invokes Engels' pause (1790-1840), when British wages stagnated amid rapid GDP growth, as a warning that popular backlash could kneecap AI's development. He contends the current phase is about building compounding infrastructure, and predicts daily life may look similar even 50 years from now.

Interconnects·Sep 9, 11:01 UTC · 7d agoAI industry1

22

AdamX: Cosine similarity meets gradient descent

Researchers propose AdamX, a cosine-similarity-based first-order optimizer with variance rectification that matches Adam-class convergence across benchmark datasets and architectures.

The paper introduces AdamX, a first-order optimizer that uses cosine similarity as an adaptive mechanism for controlling update magnitudes, plus a variance rectification scheme for smoother optimization early in training. The method is described as scalable, model-agnostic, and straightforward to integrate into existing pipelines. Empirically, AdamX shows competitive convergence rates measured by epochs to reach performance thresholds under a fixed hyperparameter budget, with code and experiments released on GitHub.

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:42 UTC · 6d agoAI research

22

Cyble Introduces Major Upgrade to its Executive Monitoring Module

Cyble upgraded Executive Monitoring in Cyble Vision, unifying impersonation, exposure, and mention findings with AI scoring and multi-channel alerting.

Cyble released a major upgrade to Executive Monitoring within its Cyble Vision platform, consolidating Mentions, Impersonations, Exposures, and a new Surface Mentions source (blog and news coverage) into a single findings stream. The release adds AI-generated verdicts and recommendations, per-executive Risk and Reputation scoring for board reporting, and richer onboarding with aliases, reference images, and addresses to cut same-name false positives. Alerting now supports data residency, bulk import, delivery via email, WhatsApp, and SMS, an Access API, and branded executive PDF reports; the upgrade is live with no customer action required.

Cyble·Sep 10, 12:26 UTC · 6d agoTools

15

New Phishing Campaign Abuses Windows Mshta.exe to Steal Credentials and Secrets

Fortra researchers track an active phishing campaign targeting Spanish speakers that abuses mshta.exe to run HTA payloads for reconnaissance and credential theft.

Fortra's FIRE team says a phishing campaign active since June 2026 targets Spanish-speaking users with invoice ('Facturación') and judicial ('Aviso Judicial') lures, many sent via libero.it/italiaonline.it infrastructure with SCL:-1 markings to bypass anti-spam. Clicking embedded links delivers an HTA launcher executed via mshta.exe, which hides its window off-screen, builds dynamic C2 URLs, and performs reconnaissance using WMI, PowerShell, and environment-variable inspection. A second-stage JavaScript dropper uses HTML smuggling to reconstruct a Base64-encoded ZIP in the browser, delivering a 7-Zip self-extracting executable disguised as a Firefox installer; staged design allows later delivery of credential stealers or ransomware. Defenders are urged to block archivogratuito[.]online and shortener domains goo[.]su, abrir[.]link, and abre[.]ai, and to restrict mshta.exe execution via AppLocker or ASR rules.

GBHackers·Sep 12, 06:16 UTC · 4d agoPhishing & fraud in the wild1

50

Training and Finetuning Multi-Vector Embedding Models with Sentence Transformers

Hugging Face published a tutorial on training and finetuning multi-vector embedding models using the Sentence Transformers library.

Hugging Face's blog walks through training and finetuning multi-vector embedding models with Sentence Transformers. Multi-vector approaches store multiple vectors per document to support late-interaction retrieval. The post is a practical guide for developers building retrieval pipelines with the library.

Hugging Face Blog·Aug 26, 00:00 UTC · 21d agoAI tools & infra1

20

Muse can shop, write emails, and negotiate prices for users, all through WhatsApp

Meta launched Muse, a WhatsApp-controlled agent running on an isolated VM with a Sentinel gatekeeper, able to shop, email, book travel, and negotiate.

Meta introduced Muse, an autonomous agent controlled through WhatsApp that runs on its own cloud virtual machine, plans multi-step tasks, browses, fills forms, and negotiates on users' behalf. Payments run through Stripe's Link using one-time cards, which Meta calls the first AI agent covered by Link's purchase protection, with Shop Pay and 1Password integration planned. A second agent, Sentinel, gates all Muse network access and holds credentials, and a Muse Confidential VM with user-held encryption keys is planned later this year. Muse's model reportedly scored 44-48 on Artificial Analysis Intelligence Index v4.3, up from 31 for Muse Spark in April, near GPT-5.6 Sol's 47; it launches first in the US on iOS and Android.

The Decoder·updated Sep 12, 20:00 UTC · 3d ago·first Sep 10, 12:27 UTC · 6d agoAI industry 10 sources1

72

Data Scarcity and Model Sparsity: Mixtures-of-Experts Overfit More to Repeated Data

Study finds Mixture-of-Experts models overfit faster than dense Transformers under repeated training data, with degradation tied to total parameter sparsity.

Across models from 80M to 1B active parameters (8.5B total), MoE architectures degrade more rapidly than dense models when training data is repeated, with the effect increasing with sparsity as dictated by total parameters. Dense 80M models tolerate 8x repetition with minimal loss while MoEs suffer at 4x and underperform dense models beyond 32x. Masking-based regularization such as dropout mitigates overfitting, letting MoEs beat dense models even at over 64x repetition, though no method matches all-unique training data. Routing stabilizes early and expert specialization correlates with overfitting to repeated data.

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:57 UTC · 5d agoAI research

36

Microsoft’s September 2026 Patch Tuesday addresses 964 CVEs (CVE-2026-81963, CVE-2026-85880)

Microsoft's September 2026 Patch Tuesday fixes a record 964 CVEs, including two zero-days actively exploited in the wild.

Microsoft patched a record 964 CVEs in its September 2026 Patch Tuesday release, with 101 rated critical and 824 rated important, making it the largest Patch Tuesday to date. The release includes fixes for two zero-days that were exploited in the wild, with CVE-2026-81963 and CVE-2026-85880 highlighted in the release. Affected products span Windows core components, Office, Exchange, .NET, Visual Studio, Active Directory services (including AD CS and AD FS) and numerous Azure services. Tenable notes the release surpasses July's as the biggest ever.

Tenable Blog·Sep 8, 18:07 UTC · 7d agoVulnerability in the wildCVE-2026-81963CVE-2026-858801

85

Likelihood-free inference with nuisance parameters through normalizing flows

Researchers decompose normalizing flows to derive near-pivotal statistics for likelihood-free inference with nuisance parameters, recovering the t-test and beating Welch limits.

A new paper decomposes neural-network normalizing flows to uncover pivotal statistics in the presence of nuisance parameters using only a sample generator from the distribution of interest. The statistic is near-pivotal in the sense of minimum average KL-divergence of its p-values and can incorporate prior knowledge of group invariances such as translation and scale. Experiments show it recovers the one-sample t-test almost exactly, outperforms the Welch test on worst-case size over a constrained variance-ratio range, and delivers higher power and much faster runtime than profile likelihood-ratio techniques on small-to-moderate samples.

arXiv cs.AI / cs.LG / cs.CL·Sep 9, 17:58 UTC · 6d agoAI research

16

25 Years of Mass Surveillance Is Enough

Bruce Schneier and Cindy Cohn argue post-9/11 mass surveillance expanded far beyond its counterterrorism justification and should be reevaluated for costs to rights.

An essay by Bruce Schneier and Cindy Cohn (originally in Lawfare) traces the post-9/11 shift from targeted surveillance to mass collection of telephone and internet metadata. It cites the Section 215 bulk phone records program, struck down in interpretation by the Second Circuit in 2015 and curtailed by the USA Freedom Act, and the NSA's Upstream program under Section 702 of the 2008 FISA Amendments Act, which ended content searches in 2017. The authors note mass surveillance now serves routine law enforcement and immigration actions, with FBI Director Kash Patel confirming purchases of Americans' data from brokers, and private systems like Flock license plate readers and venue facial recognition feeding government access.

Schneier on Security·Sep 15, 11:01 UTC · 1d agoPolicy & legal

25

General Quantification of Covariate and Concept Shifts

Paper proposes γ*-concept shifts via entropic optimal transport, deriving estimable generalization bounds unifying covariate and concept shift under distribution shift.

The authors show existing definitions of concept shift break when source and target supports mismatch and propose γ*-concept shifts grounded in entropic optimal transport. They derive a general error bound covering broad loss functions, label spaces and stochastic labeling, plus estimators with concentration guarantees. The resulting DataShifts algorithm quantifies distribution shifts and estimates the error bound in most applications, addressing learning bounds that were previously non-estimable from samples.

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:57 UTC · 5d agoAI research1

28

Trends in Web Threats in CY Q2 2022: Malicious JavaScript Downloaders Are Evolving

Unit 42 detected 751,000 landing URL incidents in Q2 2022 and documented malicious JavaScript downloaders evolving to evade detection.

Unit 42 detected 751,331 landing URL incidents (253,644 unique) and 1,744,629 malicious host URL incidents (256,844 unique) from April through June 2022. Total landing URL incidents rose compared with Q1 2022, and unique host URL incidents grew 42%, indicating attackers deploying more variants. The report includes a case study of a JavaScript downloader campaign demonstrating new evasion techniques. Personal sites, blogs, and business sites were the top apparently benign entry points.

Palo Alto Unit 42·Aug 17, 13:05 UTC · Aug 17, 2026Research

15

Quenched Ensemble Sampling

Quenched Ensemble Sampling generalizes nested sampling's hard energy constraint to repulsive potentials, traversing first-order phase transitions where tempering fails.

Quenched Ensemble Sampling generalizes nested sampling's hard energy constraint into a family of repulsive potentials at the energy boundary, preserving monotone energy descent while making the constrained target amenable to scalable gradient-based kernels. On synthetic phase-transition models it estimates marginal likelihood and draws posterior samples across first-order transitions where popular alternatives such as tempering fail. Applications include marginal likelihood estimation for Bayesian neural network architecture comparison and partition function estimation in a high-dimensional continuous lattice field theory.

arXiv cs.AI / cs.LG / cs.CL·Sep 14, 17:16 UTC · 2d agoAI research1

15

Meta AI builds detailed profiles of children from years of family posts

Meta AI suggested questions about a child and assembled detailed family profiles from years of Facebook posts, including a photo deleted years ago.

A mother reported that Meta AI on Facebook suggested the question 'Who is the child passenger?' and then aggregated her children's names, birth dates, videos, and a photo she had deleted years ago. The assistant also pieced together old posts to pinpoint her home location when prompted with 'Where does Kalie Robins live?'. Meta admitted the prompt 'never should have' appeared and said it fixed the suggestion issue, while noting the data came from posts the asker could already access. The article situates this among prior Meta AI privacy failures, including publicly shared chats and a bug exposing private conversations via guessable IDs.

Malwarebytes Labs·Sep 15, 09:44 UTC · 1d agoAI safety & security

28

Mask Forcing: Improving Autoregressive Video Diffusion Distillation via Dual-Noise Masking Rollout

Researchers propose Mask Forcing, a dual-noise masking rollout that mitigates mode collapse in autoregressive video diffusion distillation.

The paper targets over-saturation and over-smoothing in autoregressive video diffusion models distilled via Distribution Matching Distillation, attributed to reverse-KL mode-seeking behavior. Mask Forcing perturbs the student self-rollout with random masks along spatial and temporal axes, injecting cleaner tokens that act as denoising guidance for noisier tokens. Experiments show improved visual quality across multiple distillation methods without using real video data or extra post-training stages.

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

22

Risky Bulletin: Academics find source code overlaps between Geedge and China's Great Firewall

Academics linked Chinese vendor Geedge Networks' Tiangou Secure Gateway source code to one of the Great Firewall's three traffic filtering capabilities.

US researchers presenting at USENIX Security reconstructed Geedge Networks' Tiangou Secure Gateway firmware from over 100,000 leaked files, including Git repositories with commit history, and matched its filtering behavior to sections of China's Great Firewall. They found only 1 of 3 characterized DNS injectors matched Geedge code, noted the system relies on memory-unsafe C components and copied third-party code, and said its bugs could aid future circumvention tools. Geedge also exports censorship tools to Kazakhstan, Ethiopia, Pakistan, and Myanmar. The newsletter additionally rounds up multiple breaches.

Risky Business News·Aug 21, 02:55 UTC · 26d agoResearch2

48

MAxBench: A Multinomial Concept Recovery Benchmark

MAxBench evaluates multinomial concept recovery methods, finding affine subspaces steer most reliably but none consistently beats prompting.

MAxBench is a geometry-agnostic evaluation framework for multinomial concept representations in language models, based on sampling from recovered concept representations. It compares 10 localization methods covering 5 geometry types across 6 concepts and 4 models. Findings show affine subspaces steer more reliably than rank-one or linear subspaces due to better non-zero offsets, manifold steering is competitive where applicable, and no method consistently outperforms prompting.

arXiv cs.AI / cs.LG / cs.CL·Sep 11, 17:08 UTC · 5d agoAI research

18

Infostealers Target Claude, Cursor, Codex and Other AI Agents to Steal Credentials and Sensitive Data

Gen Digital researchers report infostealer families Amatera, Remus and CallbackBeaver now harvest Claude, Cursor and Codex agent data, including tokens and MCP configs.

Gen Digital researchers found commodity infostealers extending their collection rules to local AI coding agent data from Claude, Cursor, Codex, Cline, Continue and OpenCode on Windows and macOS. Amatera targets Cline and Continue, Remus targets Claude, Cursor and OpenCode, and CallbackBeaver added Claude and Cursor with more than 5,000 samples observed in 30 days; Djinn Stealer hits Claude, Codex, Gemini, Cline, OpenCode and Kilo on macOS. Stolen data includes access and refresh tokens, prompt histories, conversation databases and MCP configurations holding API keys, potentially exposing connected source-control, cloud and ticketing systems. Remus is assessed as a Lumma Stealer variant using EtherHiding C2 resolution via Ethereum smart contracts.

GBHackers·Sep 9, 07:05 UTC · 7d agoMalware in the wild1

62

Eliciting Weak-to-Strong Generalization with On-Policy Reverse Distillation

OPRD distillation enables weak-to-strong generalization by amplifying verifier-supported policy updates, outperforming existing RL and distillation methods with fewer student updates.

On-Policy Reverse Distillation (OPRD) evaluates a weak teacher's policy shift relative to its reference policy on student rollouts and amplifies the verifier-supported component of the student's policy gradient. This rescaling preserves the stationary points of policy optimization while letting the student learn beyond the teacher's capacity ceiling. In successive model transfer and multi-teacher distillation, OPRD achieves higher performance with fewer student updates than existing RL and distillation approaches, and response-style analysis shows students remain closer to verifier-RL-trained models than to their weak teachers.

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

33

Why the current tech backlash feels different

The Verge's Decoder mailbag discusses the current tech backlash, arguing AI hype overstates verifiability outside software engineering.

Nilay Patel's Decoder mailbag episode addresses listener feedback on the widely discussed 'software brain' essay. He argues AI hype is concentrated on software because code is verifiable through compilation, while domains like drug discovery, math and science lack equivalent verifiability. The episode also touches on AI backlash, surveillance, data centers and upcoming midterm coverage.

The Verge · AI·Sep 10, 14:00 UTC · 6d agoAI industry1

18

When Does Scale-Invariant Optimization Become Unstable? An Exact Schedule Law with Weight Decay

Researchers derive an exact law linking learning-rate schedules and weight decay in normalized networks, pinpointing when scale-invariant optimization destabilizes.

The paper shows that normalization makes large parts of neural networks scale-invariant, creating a hidden feedback loop where learning-rate schedules and weight decay interact through the parameter norm to control the effective optimizer step. An exact discrete-time law with a single scalar quantity separates contraction- and expansion-dominated effective learning-rate regimes, and the balance point is intrinsically unstable, so constant learning rate with weight decay produces recurrent behavior instead of a stable equilibrium. A unified homogeneous-optimizer framework explains why adaptive methods stabilize more weakly under normalization. The law is validated with high precision on MLPs, CNNs, and GPT-2 across MNIST, CIFAR, WikiText, and OpenWebText, with code released on GitHub.

arXiv cs.AI / cs.LG / cs.CL·Sep 8, 17:46 UTC · 8d agoAI research

30

Climate-ModernBERT: Revisiting Corpus Composition for Domain-Adaptive Continued Pretraining

Climate-ModernBERT domain-adapted encoders reach 76.3 average F1 across nine climate benchmarks, 2.8 points above vanilla ModernBERT-Base.

The authors continue pretraining ModernBERT-Base on three climate corpora - academic text, climate-filtered web data, and synthetic documents - and compare joint mixtures against parameter-space merging of specialized checkpoints. The best model achieves 76.3 average F1 across nine climate NLP benchmarks, a 2.8-point improvement over the vanilla baseline. Academic climate corpora provide the strongest adaptation signal, and parameter-space merging outperforms joint multi-source training while preserving complementary corpus information; all variants are released.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:41 UTC · 9d agoAI research

22

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·Sep 14, 18:08 UTC · 1d agoPhishing & fraud in the wild1

68

Learning 3D Editing without Paired Supervision via Generative Prior Distillation

New framework distills 2D editing and VLM priors into a feed-forward 3D editing model without paired 3D training data.

The method, PriorEdit3D, learns feed-forward instruction-guided 3D editing by distilling knowledge from foundation models instead of using ground-truth 3D pairs. Through a differentiable rendering pipeline it supervises a 2D visual prior from an image editing model at the main view and a Vision-Language Model semantic prior at novel views for instruction fidelity and identity preservation. A 3D-aware Distribution Matching regularization constrains outputs to the manifold of realistic 3D assets defined by a pretrained image-to-3D teacher. Experiments report superior instruction fidelity and cross-view consistency over state-of-the-art baselines, with code released on GitHub.

Hugging Face daily papers·Sep 3, 20:00 UTC · 12d agoAI research

30

Meta debuts its Muse AI agent. Will consumers trust it?

Meta launched Muse, a consumer AI agent powered by Muse Spark that connects to users' apps to execute tasks like emailing, booking travel, and payments.

Meta introduced Muse, a personal AI agent for US users that connects to email, calendars, payments, shopping, and other services to execute tasks such as booking travel, lowering bills, and completing purchases via Link by Stripe. The agent runs in a dedicated Muse Secure VM with a separate Sentinel agent kept apart at the system level, and Meta claims it cannot see passwords or payment data and does not share conversations with ad systems. Muse is free to start, with Power ($20/month) and Maximum ($100/month) subscription tiers, and is available on the web, iOS, Android, and WhatsApp, with Meta AI glasses support planned. The launch follows Meta's $18 billion multistate consumer-harms settlement and comes as rivals like Gemini Spark and Claude Cowork push agentic AI.

TechCrunch · AI·Sep 8, 19:00 UTC · 7d agoAI industry 3 sources1

74

Distance generalization in transformers: why bother with positional encoding?

arXiv study uses synthetic delay-copy tasks to show how RoPE, ALiBi, NoPE and training data diversity affect transformers' distance generalization.

The paper studies distance generalization in transformers: extrapolating when inter-token distances change between training and inference while context length stays fixed. Using two synthetic delay-copy tasks with finite source-recall distances, the authors test models on unseen delays. They investigate whether positional encodings such as RoPE and ALiBi outperform no positional encoding (NoPE), how the diversity of training distances affects performance, and when distance transfer learning is positive or negative.

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:57 UTC · 5d agoAI research2

30

The August 2026 Security Update Review

ZDI's August 2026 review covers a smaller-but-still-huge Microsoft and Adobe patch batch, with Adobe shipping 51 CVEs across five bulletins.

The Zero Day Initiative's August 2026 Security Update Review covers the latest Adobe and Microsoft patches. Adobe released five bulletins addressing 51 unique CVEs. The release is smaller than July's but still large by historical standards, which ZDI notes reflects a 'new normal' in patch density.

Zero Day Initiative Blog·Aug 11, 17:56 UTC · Aug 11, 2026Vulnerability1

50