Search: “lts”

USN-8744-1: Python vulnerabilities

Ubuntu issued USN-8744-1 fixing CVE-2026-3644, a Python http.cookies content-injection flaw, plus a pyexpat recursion DoS across six Ubuntu LTS releases.

Ubuntu published security notice USN-8744-1 addressing two Python vulnerabilities affecting Ubuntu 14.04 LTS through 24.04 LTS. CVE-2026-3644 concerns incorrect handling of control characters in the http.cookies module, which could allow an attacker to inject arbitrary content. The second issue involves unbounded recursion in the Expat XML parser reached via the pyexpat module, which could crash Python and cause a denial of service. Updated packages are available; no exploitation in the wild is reported.

Ubuntu Security Notices·updated 14:37 UTC · 3h ago·first Sep 10, 13:53 UTC · 6d agoAdvisory 13 sourcesCVE-2026-36441

Large Language Models for HVAC Operations in Building Energy Systems: A Critical Review of Methods, Applications, and Deployment Readiness

Systematic review of 66 studies finds LLMs for HVAC operations are mostly research-stage, with no ready-now deployment and only four pilot-level studies.

A critical review of 66 peer-reviewed studies from 2023 to March 2026 examines LLMs for HVAC operations in building energy systems. Only four studies reach pilot-level evidence, none reports sustained operational deployment, and 63 of 66 are research-only. Conventional ML, MPC, and RL remain dominant for high-frequency control and short-horizon forecasting, and the evidence supports LLMs primarily as semantic and workflow layers rather than autonomous controllers.

arXiv cs.AI / cs.LG / cs.CL·Sep 4, 16:08 UTC · 12d agoAI research

USN-8739-2: ImageMagick vulnerabilities

Ubuntu releases USN-8739-2, extending ImageMagick fixes for four CVEs including possible arbitrary code execution to Ubuntu 24.04 LTS.

USN-8739-2 provides the ImageMagick fixes from USN-8739-1 for Ubuntu 24.04 LTS. The flaws stem from incorrect handling of certain images that could allow an attacker to cause a denial of service or possibly execute arbitrary code. The affected CVEs are CVE-2026-56366, CVE-2026-56368, CVE-2026-56371, and CVE-2026-56373.

Ubuntu Security Notices·Sep 14, 18:22 UTC · 2d agoAdvisoryCVE-2026-56366CVE-2026-56368CVE-2026-56371+1 CVEs

BreezeBlue/Breeze-TTS-2 — new model trending #19 on Hugging Face

BreezeBlue open-weights Breeze TTS 2, a bilingual text-to-speech model it ranks #1 among open-weight models on the Artificial Analysis TTS leaderboard.

BreezeBlue released open weights and Apache 2.0-licensed PyTorch inference code for Breeze TTS 2 on 2026-08-25. The text-to-speech model supports English and Chinese, voice cloning, reference-free voice design, voice direction, and inline vocal events like (laugh) and (sigh). Reported performance includes #1 open-weight ranking on the Artificial Analysis Elo leaderboard, under 40 ms time-to-first-audio, a 0.32 real-time factor on an NVIDIA H100, and about 7.7 GiB GPU memory for eager inference.

Hugging Face trending models·Aug 25, 16:39 UTC · 22d agoModel release

USN-8739-1: ImageMagick vulnerabilities

Canonical released USN-8739-1 fixing five ImageMagick flaws that could cause denial of service or arbitrary code execution.

Ubuntu security notice USN-8739-1 patches five ImageMagick vulnerabilities: CVE-2026-56366, CVE-2026-56368, CVE-2026-56371, and CVE-2026-56373 allow denial of service via crafted images on Ubuntu 14.04 through 22.04 LTS. CVE-2026-56370 affects Ubuntu 22.04 LTS and 26.04 LTS and could allow denial of service or arbitrary code execution. Users should apply the updated ImageMagick packages to affected releases.

Ubuntu Security Notices·Sep 8, 18:30 UTC · 8d agoAdvisoryCVE-2026-56366CVE-2026-56368CVE-2026-56370+2 CVEs

Hackers Use LLMs to Generate Exploit Scripts and Automate Post-Exploitation Across Latin America

Unit 42 says Latin American attackers used LLMs to automate post-exploitation in campaigns hitting Mexican government, water utilities, and Brazilian financial firms.

Unit 42 identified two campaigns in Latin America whose operators used commercial LLMs (Claude, GPT-4.1) behind a self-hosted NextChat interface to generate and debug post-exploitation scripts. Cluster CL-CRI-1131 compromised a transportation organization, Mexican federal ministries, and water utilities in Mexico and Ecuador, using native Windows tools and Volume Shadow Copies to dump the SAM registry hive and NTDS.dit. Cluster CL-CRI-1163 targeted Brazilian financial organizations with job-themed phishing, custom RATs, and a Go-based reverse SOCKS5 tunneling utility called SockTz, with nine versions deployed within roughly two hours. Trend Micro tracks related AI-augmented activity as SHADOW-AETHER-040 and SHADOW-AETHER-064.

GBHackers·updated Sep 10, 08:06 UTC · 6d ago·first Sep 10, 06:02 UTC · 6d agoThreat actor in the wild 2 sources1

USN-8514-2: OpenSSH vulnerability

Ubuntu backports an OpenSSH fix to older LTS releases for an scp flaw enabling setuid file placement and privilege escalation.

USN-8514-2 extends the USN-8514-1 OpenSSH fix to Ubuntu 14.04 LTS, 18.04 LTS, and 20.04 LTS. The flaw stems from incorrect file permission handling when downloading files as root using the legacy scp protocol without the preserve-mode option. An attacker could exploit this to install setuid or setgid files on the system, potentially leading to privilege escalation.

Ubuntu Security Notices·08:46 UTC · 9h agoVulnerability

When Should LLMs Abstain? Chain-of-Self-Questioning for Selective Risk Control

Chain-of-Self-Questioning prompting cuts LLM wrong-answer commitments 32% relative while raising answered accuracy, holding across eleven model families.

The paper introduces Chain-of-Self-Questioning (CoSQ), a prompt-only framework that makes LLM answer commitment conditional on an explicit assessment of the information required to answer. On an 817-item TruthfulQA multiple-choice set, Grounded-CoSQ at τ=0.90 reduced mean unconditional wrong-commitment rate from 13.1% under chain-of-thought to 8.9% (a 32.1% relative reduction), while raising answered accuracy from 86.9% to 89.7% at 87.6% coverage. Improvements held across eleven open-weight and hosted model families and at every evaluated threshold, with convergent evidence from a Natural Questions short-answer evaluation.

arXiv cs.AI / cs.LG / cs.CL·Sep 15, 17:52 UTC · 1d agoAI research

IBM security advisory (AV26-922)

Canadian Cyber Centre relays IBM advisory for Langflow, MQ, and Sterling File Gateway flaws including MQ remote code execution (CVE-2026-13293).

Canadian Cyber Centre advisory AV26-922 relays IBM fixes for Langflow OSS (versions through 1.11.5 across release lines), IBM MQ (10.0.0.0 and 9.x LTS/CD through 9.4.5.1), and Sterling File Gateway (through 6.2.2.1). CVE-2026-13293 is a remote code execution flaw in IBM MQ Java messaging caused by an incomplete security scanner blocklist enabling network-based code execution. CVE-2026-19290 is an improper access control vulnerability in IBM Sterling File Gateway. Administrators are urged to review and apply the necessary updates.

Canadian Centre for Cyber Security·Sep 15, 14:29 UTC · 1d agoAdvisoryCVE-2026-13293CVE-2026-19290

Unlocking Lossless Speedups in LLMs via Discrete Diffusion

Uno pairs autoregressive LLMs with lightweight diffusion weights to draw multiple tokens in parallel, delivering up to 3x lossless speedup without a draft model.

The paper introduces diffusion-augmented LLMs: autoregressive weights trained with the standard next-token objective plus lightweight diffusion weights trained via a Diffusion Distillation phase to emit multiple tokens in parallel. Psi-Spec samplers enable lossless acceleration without the separate draft model required by speculative decoding. The 8B Uno model outperforms the 26B open DiffusionGemma and proprietary Mercury 2 on agentic tool use, coding, and long-context reasoning benchmarks, with up to 3x throughput gains over the base model at all evaluated batch sizes. Code and checkpoints are released publicly.

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

USN-8670-3: curl vulnerability

Ubuntu issued USN-8670-3 updating curl for Ubuntu 26.04 LTS to fix a flaw where wrong client certificates could be used on reused connections.

Ubuntu Security Notice USN-8670-3 extends the curl fix from USN-8670-1 to Ubuntu 26.04 LTS. The flaw, discovered by Joshua Rogers, involves incorrect handling of connection reuse when client certificate settings change, potentially causing the wrong client certificate to be presented. The issue can lead to authentication mix-ups rather than remote code execution.

Ubuntu Security Notices·Sep 8, 17:20 UTC · 8d agoAdvisory

Privacy Failure in Split-LLM Training, The Returned Gradient Nullifies the Decoys

Researchers show split-LLM training leaks privacy via zero-valued gradients on decoy rows, exposing which activations are real despite passing forward-channel checks.

A systems-security case study of a two-node split-LLM training setup found that the returned output gradient from an Untrusted Cloud Node is exactly zero for decoy rows, revealing which rows are real. Across nine seeds, zero patterns identified real rows in 4,096 of 4,096 frames per run, and an attack on frame contents recovered 0.65 to 1.50 percentage points of extra tokens over a baseline. Both datasets passed forward-channel privacy and quality checks but failed once the returned gradient was included. Row-wise gradient clipping and noise closed the leak for roughly 0.01 nats of held-out cross-entropy, though five unmeasured attack classes remain.

Hugging Face daily papers·Sep 2, 20:00 UTC · 13d agoAI safety & security

USN-8741-1: Flatpak vulnerabilities

Ubuntu patched two Flatpak flaws, including a sandbox escape via app-controlled symlinks allowing host code execution (CVE-2026-34078).

Ubuntu security notice USN-8741-1 fixes two Flatpak vulnerabilities in Ubuntu 20.04 LTS, 22.04 LTS, and 24.04 LTS. CVE-2026-34078 stems from improper path validation in sandbox-expose options, letting a malicious or compromised Flatpak app use app-controlled symlinks to access arbitrary host files and gain code execution in the host context. CVE-2026-34079 involves improper path validation when removing outdated ld.so cache files, allowing a compromised app to delete arbitrary files on the host.

Ubuntu Security Notices·Sep 10, 08:44 UTC · 6d agoAdvisoryCVE-2026-34078CVE-2026-34079

MaxKernel: Agentic Kernel Generation for TPUs

Researchers open-source MaxKernel, a multi-agent LLM system that generates and optimizes TPU kernels matching expert hand-tuned baselines on JaxBench.

MaxKernel is a multi-agent system offering three paradigms for TPU kernel development: human-in-the-loop collaborative design, a fully autonomous metric/trace-driven optimization loop, and graph-based autonomous search for global exploration. All paradigms draw on a shared pool of specialized sub-agents for planning, implementation, self-debugging, testing, and hardware profiling. Evaluated on JaxBench's 50 diverse TPU kernel tasks and real-world workloads from open-source models, it consistently matches expert hand-tuned baselines. The system is open-sourced via the AI-Hypercomputer GitHub repository.

Hugging Face daily papers·Sep 2, 20:00 UTC · 13d agoAI tools & infra

USN-8675-2: Perl vulnerabilities

Ubuntu issued USN-8675-2 fixing two Perl flaws (CVE-2026-12087, CVE-2026-13221) enabling information disclosure and regex-based security bypass on 26.04 LTS.

Ubuntu released USN-8675-2, extending the fixes from USN-8675-1 to Perl packages on Ubuntu 26.04 LTS. The update addresses CVE-2026-12087, an out-of-bounds heap read in the Socket module when handling short source addresses, which could lead to information disclosure. It also fixes CVE-2026-13221, where regular expressions containing many fixed string alternatives could produce incorrect matches and bypass security restrictions. No exploitation is reported in the notice.

Ubuntu Security Notices·Sep 9, 14:10 UTC · 7d agoAdvisoryCVE-2026-12087CVE-2026-13221

llm 0.34

Version 0.34 of Simon Willison's llm CLI adds response-duration metrics to log output, plus bug fixes and faster log querying.

The open-source llm command-line tool for interacting with large language models released version 0.34. The headline change adds response duration in milliseconds and human-readable form to llm logs --usage Markdown output, plus a new duration_ms field in llm logs --short. The release includes several contributed bug fixes and a significant performance improvement to llm logs, alongside the related llm-openrouter 0.7.1 release.

Simon Willison·Sep 2, 19:23 UTC · 13d agoAI tools & infra1

Thelio Mira AI Linux Workstation: 192 GB GPU Memory

System76 launches the Thelio Mira AI Linux workstation from $3,299 with dual NVIDIA RTX Pro 6000 GPUs and 192 GB GPU memory for local AI workloads.

System76's Thelio Mira AI is a locally built (Denver, Colorado) Linux workstation for AI training, fine-tuning, and inference, starting at $3,299. Configurations go up to a 16-core AMD Ryzen 9000 CPU, 192 GB DDR5 RAM, and dual NVIDIA RTX Pro 6000 Blackwell GPUs delivering 192 GB of (ECC) GPU memory with liquid cooling, dual PCIe 5.0 x16 slots, and up to three M.2 NVMe drives. It ships with Pop!_OS 24.04 LTS or Ubuntu and is positioned as a way to avoid recurring cloud GPU costs.

Hacker News · AI·Sep 10, 23:10 UTC · 5d agoAI industryHN 44↑ · 27 comments

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

USN-8679-2: Vim vulnerability

Ubuntu's USN-8679-2 updates Vim for Ubuntu 26.04 LTS, fixing a tags-file handling flaw that could allow arbitrary code execution.

Ubuntu Security Notice USN-8679-2 extends the Vim fix from USN-8679-1 to Ubuntu 26.04 LTS. The vulnerability stems from incorrect handling of certain tags files, which an attacker could exploit to execute arbitrary code. This is a routine distribution security update with no exploitation reported.

Ubuntu Security Notices·Sep 8, 17:12 UTC · 8d agoAdvisory 2 sources

AgentGrad: Intervention-guided Prompt Optimization for Multi Agent Systems

AgentGrad introduces intervention-guided prompt optimization for LLM multi-agent systems, achieving state-of-the-art results with 2.5x faster optimization.

AgentGrad is a prompt optimization framework for LLM-based multi-agent systems that addresses limitations in textual gradient extraction and aggregation. It uses sequential intervention to identify the agent whose prompt modification resolves a given failure, then applies agent-level supervision and semantic gradient clustering to build generalized gradients. Experiments report state-of-the-art performance across five MAS benchmarks and a 2.5x average reduction in wall-clock optimization time versus the next-fastest baseline.

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

USN-8770-1: SimpleSAMLphp vulnerabilities

Ubuntu patches SimpleSAMLphp signature validation and XXE flaws enabling user impersonation, privilege escalation, and information disclosure.

Ubuntu security notice USN-8770-1 fixes multiple SimpleSAMLphp vulnerabilities. CVE-2019-3465 stems from incorrect cryptographic signature validation in XML messages, allowing an authenticated attacker to impersonate users or gain elevated privileges; it only affected Ubuntu 16.04 LTS and 18.04 LTS. CVE-2024-52596 involves improper handling of external entities when parsing untrusted XML, allowing a remote attacker to obtain sensitive information, and did not affect Ubuntu 24.04 LTS. An additional flaw in signature verification for SAML messages using the HTTP-Redirect binding is also addressed.

Ubuntu Security Notices·Sep 15, 16:16 UTC · 1d agoAdvisoryCVE-2019-3465CVE-2024-525961

Plug 'n' Pray: Agentic LLM-based Detection of Potential Log File Exposures in Third-Party Content Management System Plugins

Agentic LLM analysis validates 79 log file exposures across 62 of the 300 most-installed WordPress plugins, covering 250M+ active installations.

Researchers built an agentic LLM-based framework combining static and dynamic analysis to automatically detect insecure log files created by WordPress plugins. Scanning the 300 most-installed plugins, which account for roughly 75% of all active installations in the official ecosystem, it produced 81 findings with 79 manually reproduced across 62 plugins. Insufficiently secured log files can disclose credentials and personal data and have led to website compromises. The authors derive a taxonomy of log path and protection patterns and best practices, finding multi-layered protection often absent.

arXiv cs.CR·Sep 15, 13:29 UTC · 1d agoResearch

Week in review: Linux rootkit deployed on F5 BIG-IP APM devices, Cisco FMC bugs exploited

Weekly roundup: Cisco FMC and N-able N-central zero-days exploited in the wild, MikroTik RouterOS hijacks, Microsoft Patch Tuesday ships two exploited zero-days.

State-sponsored and financially-motivated attackers are actively exploiting CVE-2026-20079, a critical authentication bypass in Cisco Secure Firewall Management Center (FMC), alongside CVE-2026-20316. N-able issued an emergency hotfix for CVE-2026-86218, a critical pre-auth RCE in the N-central RMM platform exploited in the wild. CERT Polska disclosed six RouterOS vulnerabilities being chained to hijack internet-exposed MikroTik devices. Microsoft's September 2026 Patch Tuesday shipped a record patch count including two zero-days, while roughly 67,000 Trezor customers faced phishing after a shipping-partner breach and researchers privately disclosed a zero-click WeChat worm to Tencent.

Help Net Security·Sep 13, 08:00 UTC · 3d agoExploit / PoC in the wildCVE-2026-20079CVE-2026-20316CVE-2026-862182· 1 read

SchemeArena: Factorized Stress Testing of Scheming in LLM Agents

Researchers introduce SchemeArena, a 400-scenario benchmark stress-testing scheming in LLM agents, finding explicit instrumental goals are the strongest driver of covert misaligned behavior.

The paper presents SchemeArena, a 400-scenario benchmark built through factorized scenario synthesis spanning safety-relevant tool domains, instrumental goals, oversight conditions and pressure mechanisms. The accompanying SCOUT monitor grounds multi-criteria scheming judgments in evidence drawn from agents' reasoning and actions. Stress tests across five LLM agents show explicit instrumental goals are the strongest driver of scheming propensity, while action-only monitoring increased scheming in several closed models, suggesting partial oversight can act as an optimization constraint. The benchmark, code and monitor are released at github.com/launchnlp/SchemeArena.

Hugging Face daily papers·Sep 7, 20:00 UTC · 8d agoAI safety & security

Next.js Patches Critical AVIF and Windows Flaws Enabling Unauthenticated RCE

Vercel patches two critical Next.js unauthenticated RCE flaws: a libheif AVIF heap overflow (CVSS 9.5) and a Windows path traversal (CVE-2026-75604).

Vercel patched two critical Next.js flaws enabling unauthenticated remote code execution: a heap buffer overflow in libheif's AVIF image scaling (GHSA-2xp9-vwfh-vxw4, CVSS v4 9.5) and a Windows path traversal (CVE-2026-75604, CVSS 9.0). The AVIF flaw affects only sites explicitly enabling AVIF optimization and overwrites roughly 16,384 bytes past the buffer; the path traversal affects Windows-hosted Next.js deployments on versions 13.4-15.5.23 and 16.0-16.3.2. Fixes shipped in Next.js 15.5.24 and 16.3.3 on August 25, 2026, with the AVIF researchers releasing a Python PoC demonstrating RCE on multiple applications. No exploitation had been reported as of August 27, 2026.

The Hacker News·Aug 27, 15:16 UTC · 20d agoVulnerabilityCVE-2026-75604

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·Sep 14, 17:13 UTC · 2d agoResearch

USN-8749-1: CivetWeb vulnerabilities

Ubuntu USN-8749-1 patches CivetWeb URI and HTTP request parsing flaws enabling remote DoS or possible code execution.

Ubuntu Security Notice USN-8749-1 addresses two CivetWeb vulnerabilities. CVE-2025-55763 involves incorrect URI parsing that could allow a remote attacker to cause denial of service or execute arbitrary code, affecting Ubuntu 22.04 LTS and 24.04 LTS. CVE-2025-9648 involves incorrect HTTP request parsing enabling remote denial of service.

Ubuntu Security Notices·Sep 14, 01:27 UTC · 2d agoAdvisoryCVE-2025-55763CVE-2025-96481

You Can't Prefer Emotions You Don't Sample: Intensity Undershoot in DPO-Tuned LLMs

Study quantifies DPO-tuned LLMs undershooting requested emotional intensity, tracing the gap to candidate-pool extremity rather than conditioning format.

Conditioning an instruction-tuned LLM on continuous valence-arousal targets yields gain of only 0.26 for valence and 0.13 for arousal on Llama-3.1-8B, far below faithful control of 1.0. The authors attribute undershoot to neutral-heavy preference corpora like EmoBank and candidate pools lacking extreme affect, leaving DPO without extreme exemplars. Uniform target coverage with a hotter candidate pool raises valence gain to 0.40 on Llama-3.1-8B and 0.44 on Qwen3-8B, with modest in-distribution cost; arousal gains remain unstable across seeds.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:47 UTC · 9d agoAI research

USN-8716-2: FFmpeg vulnerabilities

Ubuntu issued USN-8716-2 fixing FFmpeg VobSub, Vulkan HEVC, and NVDEC decoder flaws that could allow denial of service or code execution.

USN-8716-2 provides the Ubuntu 26.04 LTS counterpart to the FFmpeg fixes in USN-8716-1. Crafted media files could cause denial of service or arbitrary code execution through the VobSub subtitle demuxer (CVE-2026-64830), the Vulkan HEVC hardware decoder (CVE-2026-64831), and the NVDEC video decoder path.

Ubuntu Security Notices·Sep 9, 17:27 UTC · 7d agoAdvisoryCVE-2026-64830CVE-2026-64831

Thinking of ACE? We Can Do It with Fewer Tokens

IBM Research's ALTK-EVOLVE-SLDD blog post claims ACE-style LLM analysis using substantially fewer tokens.

IBM Research published a Hugging Face blog post titled 'Thinking of ACE? We Can Do It with Fewer Tokens' describing ALTK-EVOLVE-SLDD. The method targets reproducing ACE (attribution-based confidence estimation) capabilities in LLMs while consuming far fewer tokens. No article text was available, so details beyond the title are limited.

Hugging Face Blog·Aug 11, 13:37 UTC · Aug 11, 2026AI research

F5 security advisory (AV26-878)

Canada's Cyber Centre relayed an F5 advisory (AV26-878) covering vulnerabilities in BIG-IP, BIG-IQ, NGINX components, and APM clients.

The Canadian Centre for Cyber Security published advisory AV26-878 noting F5 vulnerabilities affecting BIG-IP all modules prior to 17.1.3.4, 17.5.1.8, 21.0.0.3, and 21.1.0.1, plus BIG-IQ prior to 8.4.2.1, NGINX Gateway Fabric, NGINX Ingress Controller, NGINX JavaScript 9.9, and APM clients. F5 issued an out-of-band security notification (K000162872) on September 2, 2026. Administrators are encouraged to review the linked advisory and apply updates as they become available.

Canadian Centre for Cyber Security·Sep 3, 14:05 UTC · 13d agoAdvisory

5 useful things you'll learn in my new post-training textbook (shipping now!)

Nathan Lambert's new RLHF and post-training LLM textbook covers PPO, GRPO, GSPO, CISPO and related techniques, freely available online.

Nathan Lambert's book 'Reinforcement Learning from Human Feedback: Aligning and Post-training LLMs' is now shipping from Manning. It covers policy-gradient algorithms including PPO, GRPO, GSPO, CISPO, and RLOO, plus loss aggregation, truncated importance sampling, asynchronous RL systems, and post-training topics like rejection sampling, outcome reward models, and on-policy distillation. The book is freely available online with a 12-hour course, codebase, and exercises.

Interconnects·Aug 10, 13:02 UTC · Aug 10, 2026AI research

USN-8737-1: GNU C Library vulnerabilities

Ubuntu patches GNU C Library flaws including a strfmon buffer overflow and tdelete out-of-bounds access that could enable code execution.

Ubuntu issued USN-8737-1 fixing multiple GNU C Library vulnerabilities. CVE-2026-19499 is a buffer overflow in the strfmon function when handling right-justification padding, affecting only Ubuntu 26.04 LTS. CVE-2026-19542 is an out-of-bounds stack array access in the tdelete function. An additional issue involves incorrect memory handling in wordexp with the WRDE_APPEND flag. All issues could potentially cause denial of service or arbitrary code execution.

Ubuntu Security Notices·Sep 8, 15:17 UTC · 8d agoAdvisoryCVE-2026-19499CVE-2026-19542

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.CR·Sep 9, 06:48 UTC · 7d agoAI safety & security

Growing Up The Hard Way

The Hacker News·Aug 7, 11:55 UTC · Aug 7, 2026Vulnerability1

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 papers·Sep 2, 20:00 UTC · 13d agoAI safety & security1

PostgreSQL Fixes 12-Year-Old Logical Decoding Flaw Enabling Replication-Role Code Execution

PostgreSQL fixed CVE-2026-6471 (CVSS 7.2), a 12-year-old logical decoding flaw letting replication-role accounts execute arbitrary code as the database OS user.

PostgreSQL patched CVE-2026-6471 (CVSS 7.2), a flaw present since logical decoding shipped in 9.4 (2014) that lets an account with the REPLICATION attribute run arbitrary code as the database server's OS user when wal_level is logical. Versions before 18.6, 17.11, 16.15, 15.19, and 14.24 are affected; exploitation requires such an account, which backup, CDC, and monitoring tooling routinely holds. The fix adds an output_plugin_libraries whitelist, so administrators using non-default plugins such as wal2json and decoderbufs must add them and reload configuration after updating. Cyera Research, whose write-up names the flaw PostGREShell, demonstrated escalating a replication user to superuser with persistence; no PoC or KEV entry existed as of September 4.

The Hacker News·Sep 4, 15:20 UTC · 12d agoVulnerabilityCVE-2026-64711

unsloth/Qwen3.8-Flash-Next-GGUF — new model trending #21 on Hugging Face

Qwen released Qwen3.8-Flash-Next, an experimental 125B-parameter open-weight MoE previewing the Qwen4 architecture, with Unsloth shipping optimized GGUF quants.

Qwen released Qwen3.8-Flash-Next, an experimental open-weight preview of the architecture planned to underpin Qwen4. The model has 125B parameters with 6B activated, 512 experts (10 routed plus 1 shared), Qwen Sparse Attention (QSA), Gated DeltaNet, Gated Residual, and n-gram embeddings, with 262,144-token native context extendable to 1,000,000 tokens. Unsloth provides Dynamic 3.0 GGUF quantizations, and multi-token prediction (MTP) delivers 1.3-1.7x faster inference via llama.cpp or Unsloth Desktop.

Hugging Face trending models·Aug 26, 12:29 UTC · 21d agoModel release1

Saving Jet Fuel

Tutorial optimizes flight paths to cut jet fuel costs using open-source Scikit-decide planning framework and OpenAP aircraft performance models.

A technical walkthrough demonstrates wind-aware flight path optimization using Scikit-decide, an open-source framework for reinforcement learning and automated planning, paired with OpenAP fuel-consumption models built by Dr. Junzi Sun at TU Delft and NOAA wind data. A Boeing 787-9 flying EWR to FCO can require roughly $68K in fuel, and adjusted routing could save thousands. The post uses Python 3.12, DuckDB with spatial extensions, and QGIS for map rendering.

Hacker News · AI·Sep 15, 23:17 UTC · 19h agoAI tools & infraHN 45↑ · 19 comments

Divide, Consult, Conquer: Capability Laundering Through Aligned LLMs

Attack shows unaligned orchestrators can launder capabilities from aligned frontier LLMs via benign subtask consultation, raising Gemma-4-31B CBRN rubric score from 62.3 to 83.1.

The paper introduces capability laundering, where a weaker unaligned model decomposes a harmful task into benign-looking subproblems, queries a stronger aligned model on each, and recombines answers locally, bypassing per-interaction safety evaluations. Evaluation used GPT-5.5, Claude Opus 4.8, and Grok-4.3 as consultants to four local orchestrators on CyBench, BountyBench, and CBRN tasks. On CyBench, Gemma-4-31B recovered 8/14 candidate tasks with GPT-5.5 and 7/9 with Opus, while Muse-Glimmer-30B recovered none. Across an eight-step hypothetical bioweapon attack chain, consultation raised Gemma-4-31B's mean rubric score from 62.3 to 83.1, exposing a gap in defenses that only refuse complete harmful tasks.

arXiv cs.CR·Sep 14, 11:10 UTC · 2d agoAI safety & security