Search: “acl”

[20260809] - Core - Improper ACL checks when injection schema.org contact data

Joomla fixed CVE-2026-73372, an improper ACL check that leaks inaccessible contact items' data into schema.org snippets, affecting CMS 5.1.0-5.4.7 and 6.0.0-6.1.2.

Joomla security advisory 20260809 describes CVE-2026-73372, an incorrect access control issue when injecting schema.org contact data. Improper access checks inject contact information for inaccessible contact items into schema.org snippets, exposing restricted data. Affected versions are 5.1.0-5.4.7 and 6.0.0-6.1.2; fixes ship in 5.4.8 and 6.1.3. The issue was reported by Stefan Wendhausen on 2026-07-31.

Joomla Security Centre·Aug 17, 14:00 UTC · Aug 17, 2026AdvisoryCVE-2026-73372

[20260804] - Core - Improper ACL checks for custom fields webservice endpoints

Joomla patches CVE-2026-72531, an improper ACL check allowing unauthorized custom-field creation via webservice endpoints, in CMS 5.4.8/6.1.3.

Joomla disclosed an incorrect access control issue (CVE-2026-72531) letting unauthorized users create custom fields for inaccessible components through webservice endpoints. It affects Joomla CMS 4.0.0-5.4.7 and 6.0.0-6.1.2, rated moderate impact and severity with low probability. Fixed in Joomla 5.4.8 and 6.1.3; reported by ebadfd on 2026-07-06.

Joomla Security Centre·Aug 17, 14:00 UTC · Aug 17, 2026AdvisoryCVE-2026-72531

[20260808] - Core - Improper ACL checks for batch copy actions

Joomla fixed CVE-2026-73371, an improper ACL check letting unauthorized users batch-copy uneditable items in Joomla CMS 4.0.0-5.4.7 and 6.0.0-6.1.2.

Joomla security advisory 20260808 describes CVE-2026-73371, an incorrect access control issue in batch copy actions. The flaw allows unauthorized users to perform copy batch operations on items they cannot edit. Affected versions are 4.0.0-5.4.7 and 6.0.0-6.1.2; fixes ship in 5.4.8 and 6.1.3. The issue was reported by Sabuhi Mammadov on 2026-07-28.

Joomla Security Centre·Aug 17, 14:00 UTC · Aug 17, 2026AdvisoryCVE-2026-73371

CVE-2026-79993: Apache ZooKeeper: Missing ACL check on deleteContainer opcode allows unauthorized deletion of any empty persistent/container znode

Critical ZooKeeper flaw lets any authenticated client delete arbitrary empty persistent or container znodes by bypassing ACL checks.

CVE-2026-79993 (critical severity) affects Apache ZooKeeper 3.9.0-3.9.5 and 3.8.0-3.8.6. The deleteContainer opcode (0x14/20) is processed without verifying the caller's ACL permissions, allowing any authenticated client to delete specific empty znodes in the data tree regardless of ACL restrictions on the znode or its parent. This can corrupt coordination state for dependent distributed systems like Kafka, HBase, or Solr clusters relying on ZooKeeper.

oss-security·Sep 15, 18:24 UTC · 21h agoVulnerabilityCVE-2026-79993

[20260805] - Core - Improper ACL checks for category webservice endpoints

Joomla fixes CVE-2026-72532, an improper ACL check letting unauthorized users create categories via webservice endpoints, in CMS 5.4.8/6.1.3.

Joomla disclosed an incorrect access control flaw (CVE-2026-72532) in category webservice endpoints, allowing unauthorized users to create categories for inaccessible components. It affects Joomla CMS 4.0.0-5.4.7 and 6.0.0-6.1.2 and is rated moderate impact and severity with low probability. The fix ships in Joomla 5.4.8 and 6.1.3 on 2026-08-18; it was reported by Amin Isayev and Geo.

Joomla Security Centre·Aug 17, 14:00 UTC · Aug 17, 2026AdvisoryCVE-2026-72532

[20260803] - Core - Inconsistent ACL checks for mutating webservice endpoints

Joomla fixes CVE-2026-71574, inconsistent ACL checks letting unauthorized users mutate data via webservice APIs, in CMS 5.4.8/6.1.3.

Joomla disclosed an inconsistent access control flaw (CVE-2026-71574) in mutating webservice endpoints, where unauthorized users could perform mutations restricted in the backend UI; impact is rated high. It affects Joomla CMS 4.0.0-5.4.7 and 6.0.0-6.1.2, with moderate severity and low probability. The fix ships in Joomla 5.4.8 and 6.1.3 on 2026-08-18.

Joomla Security Centre·Aug 17, 14:00 UTC · Aug 17, 2026AdvisoryCVE-2026-71574

CVE-2026-59739: Apache ZooKeeper: Information disclosure via SetWatches reconnect replay

CVE-2026-59739: Apache ZooKeeper missing ACL check in SetWatches reconnect replay lets attackers discover ACL-restricted znode paths.

Apache ZooKeeper versions 3.8.0-3.8.6 and 3.9.0-3.9.5 contain a critical information disclosure (CVE-2026-59739) caused by a missing ACL check during SetWatches reconnect replay. An attacker can register exists-watches on non-existent paths and reconnect after those paths are created, revealing the existence of ACL-restricted paths. The issue is fixed in patched releases.

oss-security·Sep 15, 18:21 UTC · 21h agoVulnerabilityCVE-2026-59739

The Router Within: Eliciting Native Skill Routing from a Frozen LLM

Gavel reads native skill-routing signals from a frozen LLM's forward passes with two linear maps, beating retrieve-and-rerank pipelines by up to 21.9 points on Qwen3-32B.

Gavel (Glance And Verdict from a frozen LLM) elicits skill routing from a frozen agent LLM using two trained linear maps that read mid-layer states, keeping all skill text out of context. A glance step scores the full library against compact per-skill banks built in one forward pass at installation; a verdict step resumes shortlisted skills' forward passes and fuses likelihood and yes/no judgments as a product of experts. It transfers zero-shot to three public benchmarks plus SkillTraj, a new benchmark of 372 simulated agent trajectories. On Qwen3-32B it beats progressive disclosure and retrieve-and-rerank pipelines adding 1.2B–16B external parameters by up to 13.4 points on written tasks and 21.9 when skills are needed mid-rollout.

Hugging Face daily papers·updated Sep 14, 17:58 UTC · 1d ago·first Sep 13, 20:00 UTC · 2d agoAI research 2 sources

Cisco Secure Firewall Adaptive Security Appliance and Secure Firewall Threat Defense Software Object Group Access Control List Bypass Vulnerabilitiesnew

Cisco patched ACL Object Group Search bypass flaws in ASA and FTD firewall software that let unauthenticated attackers reach protected networks.

Cisco disclosed multiple vulnerabilities in the ACL Object Group Search implementation of Secure Firewall ASA and FTD Software, caused by a logic error in populating group access control policies. An unauthenticated remote attacker could send traffic that should be blocked through the device, bypassing configured access controls. Cisco has released software updates; no exploitation is mentioned.

Cisco Security Advisories·16:00 UTC · 19m agoAdvisory 9 sources

Training a coding model to paint watercolours with TRL and OpenEnv

Hugging Face tutorial trains a coding model with TRL and OpenEnv to paint watercolours through generated code.

A Hugging Face blog walkthrough uses the TRL reinforcement learning library and the OpenEnv environment framework to train a coding model. The target task is generating code that produces watercolour-style drawings, serving as a hands-on reinforcement learning training example. No article body was available in the feed, so specifics are limited to the title.

Hugging Face Blog·Sep 3, 00:00 UTC · 13d agoAI tools & infra1

Multiple Vulnerabilities Discovered in a SCADA System

Unit 42 details five vulnerabilities (CVSS 7.0-7.8) in ICONICS Suite SCADA software enabling privilege escalation and DoS.

Unit 42 discovered five vulnerabilities (CVE-2024-1182, CVE-2024-7587, CVE-2024-8299, CVE-2024-8300, CVE-2024-9852) in ICONICS Suite versions 10.97.2 and earlier for Windows during a 2024 security assessment. The flaws, rated CVSS 7.0-7.8, allow DLL hijacking, privilege escalation, information disclosure, denial-of-service and potentially full system compromise. ICONICS Suite is a SCADA solution with hundreds of thousands of installations in over 100 countries, widely used in critical infrastructure, and several dozen servers are internet-exposed per Unit 42 telemetry. ICONICS released patches and advisories with workarounds in 2024.

Palo Alto Unit 42·Aug 19, 12:04 UTC · 28d agoVulnerabilityCVE-2024-1182CVE-2024-7587CVE-2024-8299+2 CVEs1

Your Cloud Security Checklist Doesn't Work the Way You Think It Does

Intruder's 2026 Cloud Security Index found misconfiguration risk profiles differ sharply across AWS, Azure, and Google Cloud across 3,000 organizations.

Intruder analyzed misconfiguration data from 3,000 organizations across AWS, Azure, and Google Cloud for its 2026 Cloud Security Index. Weak IAM controls and missing logging affected 80-98% of accounts regardless of provider, while exposed services ranged from 76% on AWS to just 8% on Google Cloud. Top issues included S3 buckets without HTTPS enforcement (87% of AWS accounts), Entra ID users without MFA (55% of Azure accounts), and missing OS Login MFA (77% of Google Cloud accounts). Weak IAM prevalence rose with organization size, from 87% at SMEs to 98% at large enterprises, and midmarket organizations took the longest to remediate at 35 days on average.

The Hacker News·Sep 7, 11:45 UTC · 9d agoResearch

MInTRL: Off-policy Intervention can boost On-policy RL

MInTRL injects sparse judge corrections into on-policy RL rollouts, expanding exploration beyond on-policy sampling while preserving learnability on math and code benchmarks.

Minimal Intervention Reinforcement Learning periodically has a judge-intervention policy replace erroneous suffixes of the current policy's output with short corrections, then returns control, keeping trajectories largely on-policy. Training uses a sequence-level advantage-regression objective that removes the need for importance sampling. Across math and code benchmarks it consistently beats standard on-policy and off-policy baselines, remains effective with self-intervention, and performs best at moderate intervention intensity.

Hugging Face daily papers·Sep 10, 20:00 UTC · 5d agoAI research

CVE-2026-82437: Apache Storm Logviewer: Log Access Controls Not Enforced by Logviewer

Apache Storm Logviewer ignores logs.users and logs.groups ACLs for daemon logs, letting unauthorized users read sensitive logs.

CVE-2026-82437 (severity: moderate) affects Apache Storm Logviewer (storm-webapp) versions 3.0.0 before 3.1.0. The Logviewer offers logs.users and logs.groups settings for operators to control who may read log content, but for daemon logs the access decision combined the "this is a daemon log" flag with the authorizer result in a way that discarded the authorizer's answer. As a result, configured access controls were not enforced and unauthorized users could read daemon log content.

oss-security·Sep 13, 14:37 UTC · 3d agoVulnerabilityCVE-2026-82437

Bank of America impersonators weaponize ScreenConnect, then make it hard to remove

Help Net Security·Aug 5, 00:00 UTC · Aug 5, 2026Vulnerability

SAFIRE: Safety-Critical Benchmark for Fine-grained Fire and Smoke Understanding in Multimodal LLMs

SAFIRE, an 83K-image fire and smoke benchmark, shows open-source multimodal LLMs average only 61.9% accuracy on safety-critical fire reasoning.

SAFIRE is a large-scale benchmark for fire-smoke understanding in multimodal LLMs with 83K captioned images across 20 scenarios and 193K multiple-choice VQA questions spanning 10 evaluation dimensions from perception to higher-order reasoning. Annotations were built via a GPT-5.4-assisted multi-stage pipeline with MLLM majority voting. Ten open-source MLLMs (8B-38B) average 61.9% accuracy, exposing major gaps in safety-critical reasoning. Adapting vision encoders on 7% of the domain data raises fire-scene classification from 20.1% to 64.5%.

arXiv cs.AI / cs.LG / cs.CL·Sep 7, 17:59 UTC · 8d agoAI research

CVE-2026-82431: Apache Storm Client: Authorization Bypass When nimbus.groups Is Configured Without nimbus.users

Apache Storm Client 3.0.0 ACL bypass lets every authenticated principal bypass restrictions when nimbus.groups is set without nimbus.users.

CVE-2026-82431 affects Apache Storm Client (org.apache.storm:storm-client) versions 3.0.0 before 3.1.0. SimpleACLAuthorizer returned early when nimbus.users was empty, before evaluating nimbus.groups. Operators restricting cluster access by group alone received no restriction, so any authenticated principal could execute user-level commands. Fixed in version 3.1.0.

oss-security·Sep 13, 14:28 UTC · 3d agoVulnerabilityCVE-2026-82431

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 · 22h agoAI research

Cisco bundles fixes for multiple vulnerabilities, some critical, into one patch

Cisco patched seven IOS XR vulnerabilities, two rated CVSS 9.8, allowing unauthenticated remote code execution and root access on carrier routers; no exploitation observed.

Cisco released fixes for seven internally discovered vulnerabilities in IOS XR, its Linux-based network operating system for carrier-grade routers. Two flaws, CVE-2026-20274 and CVE-2026-20279, are rated CVSS 9.8 (critical) and involve lifetime resource control issues that can enable unauthenticated remote code execution with root access; the other five are rated 8.2-8.8 and cover buffer overflows, access control failures, and out-of-bounds access. All IOS XR releases including IOS XR7 are affected regardless of configuration, no workarounds exist, and remediation requires software maintenance upgrades (SMUs) or fixed releases 26.2.2/26.3.1. Cisco says the flaws are not known to be actively exploited, but experts urge immediate patching of internet-facing and core routing systems, citing parallels with Salt Typhoon tradecraft.

CSO Online·Sep 9, 02:22 UTC · 7d agoVulnerabilityCVE-2026-20274CVE-2026-20279

[AINews] Fal’s H3 Max Live breaks the infinite videogen barrier

Fal post-trained MiniMax H3 into a 'Max' variant with 35x-faster inference, enabling faster-than-realtime AI video generation and infinite streams.

Fal post-trained MiniMax's H3 model into a 'Max' variant and optimized it for its in-house inference engine, achieving roughly 35x the speed of the official endpoint. The optimization enables faster-than-realtime video generation, demonstrated by an infinite interactive AI-generated stream productized by levels.io. The roundup also notes Meta Muse Code's general availability with an SDK, open DeepSeek-V4-Flash-Vision-Exp weights, GLM-5.3-Flash's strong agentic cost/performance rankings, and Tencent's 770B-parameter Hy4 Preview MoE with 49B active parameters.

Latent Space·Sep 1, 04:36 UTC · 15d agoAI industry

Massive Redis Cryptojacking Campaign Hijacks Thousands of Linux Servers

RedisRaider cryptojacking campaign compromises thousands of exposed, unauthenticated Linux Redis servers using cron persistence to deploy XMRig Monero miners.

Hunt.io researchers track a large-scale cryptojacking operation named RedisRaider that scans IPv4 ranges for Redis services exposed on TCP port 6379 and targets instances accepting unauthenticated connections. From a master list of 12,966 candidate hosts, 2,342 were confirmed to accept commands without authentication. Attackers abuse Redis commands (CONFIG SET dir/dbfilename, SET, BGSAVE) to write malicious cron entries into /etc/cron.d or /var/spool/cron, launching XMRig-based Monero miners, with branches including SSH authorized-key injection, Lua probing, and WordPress spraying. The article also references CVE-2026-81934, a Redis TLS use-after-free allowing unauthenticated command execution, fixed in releases including 8.2.9, 8.4.6, 8.6.6, 8.8.2, and 8.10.1.

GBHackers·Sep 9, 05:28 UTC · 7d agoExploit / PoC in the wildCVE-2026-81934

Augustinian BabyLM: What Ostensive Definition Can and Cannot Teach a Small Language Model

Study shows visually grounded token embeddings in a small masked LM persist through training and improve object-property knowledge, but escape standard BabyLM benchmarks.

The paper implements ostensive definition for a small DeBERTa masked language model trained on 10M words, seeding visually grounded tokens with embeddings derived from labeled image regions before training. Visual initialization leaves a persistent, seed-replicated advantage on object-property knowledge (COMPS) and a corpus-tailored Visual-Property Swap benchmark covering color, material, size, and shape, but has no effect on most BabyLM grammar benchmarks. Synthetic grounding of previously unseeded words causally transfers the advantage to exactly those words.

arXiv cs.AI / cs.LG / cs.CL·Sep 10, 17:43 UTC · 5d agoAI research

Securing the unpatchable in an age of AI-driven vulnerabilities

Cisco Talos argues AI-driven vulnerability discovery leaves unpatchable OT systems exposed, recommending virtual patching via NGFW/IPS and micro-segmentation.

AI-assisted code analysis is uncovering vulnerabilities faster than organizations can patch, leaving certified or end-of-life OT systems with unmitigated known flaws. Talos recommends virtual patching with next-generation firewalls and IPS, micro-segmentation using VLANs and ACLs, and building visibility-based inventories of legacy systems. The article cites WannaCry's impact on the NHS and 2023 exploitation of end-of-life software in government systems, and warns that air gaps and data diodes are routinely circumvented by operational shortcuts.

Cisco Talos·10:00 UTC · 6h agoResearch

MobileVLA-R1 2.0: RL-Enhanced Reasoning for Mobile Robot Control

MobileVLA-R1 2.0 couples chain-of-thought reasoning with RL for mobile robot control, gaining 10 points on real Unitree G1 tasks.

MobileVLA-R1 2.0 is an RL-enhanced vision-language-action framework that explicitly couples structured embodied reasoning with executable mobile robot control via supervised Chain-of-Thought alignment and reinforcement learning. A reasoning-conditioned action decoder maps multimodal reasoning representations to task-level action targets, decoupling high-level action generation from robot-specific actuation for both locomotion and manipulation. It achieves an average 1.6 point SR improvement on VLN-CE and a 10.0 point improvement in full-task success on real-world Unitree G1 mobile manipulation, with evaluations covering navigation, quadruped control, and real deployments on Unitree Go2 and G1 robots.

Hugging Face daily papers·Sep 4, 20:00 UTC · 11d agoAI research

Train Smarter, Not Harder: Switching Signal-Guided Training in Active Learning

HybridAL is an active-learning training schedule that switches from retraining to fine-tuning on stabilization signals, saving up to 49% time.

Researchers find that choosing between retraining from scratch and fine-tuning is an exploitable decision variable in active learning: retraining helps in early rounds while fine-tuning is safer once the model trajectory stabilizes. HybridAL monitors an online stabilization signal using spectral exponent change and accuracy change, switching from retraining to fine-tuning after sustained stabilization. Across three encoder backbones and six text-classification tasks with five seeds each, HybridAL keeps endpoint macro-F1 non-inferior within a 0.010 margin, saves up to 49% of retraining time, and improves the time-calibration trade-off measured by negative log-likelihood.

Hugging Face daily papers·Sep 5, 20:00 UTC · 10d agoAI research

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 · 1d agoResearch

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

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

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

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

arXiv cs.AI / cs.LG / cs.CL·Sep 11, 17:01 UTC · 4d agoAI research1

CVE-2026-86219: Authen::SASL::Perl::DIGEST_MD5 versions before 2.2100 for Perl accept replayed authentication responses via unverified nonce in server_step

Authen::SASL::Perl::DIGEST_MD5 before 2.2100 for Perl accepts replayed DIGEST-MD5 authentication responses via unverified nonce handling (CVE-2026-86219).

CVE-2026-86219 affects Authen::SASL::Perl::DIGEST_MD5 versions before 2.2100 for Perl. The server_step function does not verify the nonce, allowing replayed authentication responses to be accepted in DIGEST-MD5 SASL exchanges. The fix is available in version 2.2100 of the perl-authen-sasl distribution.

oss-security·Sep 6, 18:55 UTC · 9d agoVulnerabilityCVE-2026-862191

Right Tool, Right Job: Native-Language Evaluation, Tokenizer Sensitivity, and Methodological Findings from a French-Only BabyLM

French BabyLM entry METRON-FR (125M GPT-2, 92.47M words) shows tokenizer artifacts dominate child-scale zero-shot evaluation; proposes standard diagnostics.

METRON-FR is a 125M-parameter GPT-2 pretrained on 92.47M French words, submitted to the BabyLM 2026 Strict track, scoring 85.97% on the native Quebec-French QFrBLiMP benchmark and 62.80% on the BabyLM-weighted leaderboard. A cross-lingual GLUE protocol combining French task-data translation with rank-16 LoRA shows relational tasks gain while world-knowledge tasks regress. Bilingual Lexicon Induction reaches p@1 of 68.84%, 18x above chance, and ablations show single-token zero-shot scoring is dominated by tokenizer and template artifacts at child scale.

arXiv cs.AI / cs.LG / cs.CL·Sep 15, 16:43 UTC · 23h agoAI research

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

CARDEA: Auditable Reasoning Grounded in Spatial Evidence for End-to-End Coronary Angiography Interpretation

CARDEA, a vision-language model trained only on public data, matches cardiologists on coronary angiography complexity assessment while exposing auditable bounding-box evidence.

CARDEA is a unified large vision-language model serving as the inference core of an end-to-end coronary angiography pipeline from multi-view videos to study-level diagnosis. It was trained on public datasets through visual alignment, self-distilled Chain-of-Box cold start, and reinforcement learning with verifiable rewards encouraging bounding-box reasoning. It reached 0.91 accuracy on dominance classification under domain shift and 0.90 on complexity assessment, comparable to two interventional cardiologists. RLVR raised zero-shot report generation vessel-severity macro-F1 from 0.513 to 0.686, while supervised imitation alone did not.

Hugging Face daily papers·Sep 6, 20:00 UTC · 9d agoAI research

Safe Meta-Reinforcement Learning via Information Space Reachability

Safe meta-RL framework reasons about safety in information space, learning a safety value function used for safety filtering and constrained policy optimization.

The paper proposes safe meta-RL that reasons about safety in information space, capturing both physical state and the agent's belief over the underlying task. A safety value function measures the probability of avoiding unsafe regions indefinitely and satisfies a self-consistency condition and Bellman equation, making it learnable via meta-RL. The resulting algorithm uses the learned function for safety filtering and constrained policy optimization, with effectiveness demonstrated on meta-RL benchmarks.

arXiv cs.AI / cs.LG / cs.CL·Sep 14, 17:27 UTC · 1d agoAI research

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·Aug 25, 20:00 UTC · 21d agoAI research

AuK Technical Report: An Open-Source Foundational Model for Speech Generation and Editing

Open-source speech foundation model AuK unifies generation and editing, trained on 1.95 million hours, with distilled AuK-Flash achieving 4.5x speedup.

AuK is an open-source foundational model that unifies speech generation and editing through natural-language instructions and audio context, trained on approximately 3.03 billion instruction-audio instances and 1.95 million hours of supervision across five task families including generation, content editing, and acoustic editing. It combines a multimodal LLM for semantic conditioning, a VAE jointly trained on speech, general audio, and music, and a hybrid rectified-flow Transformer using dual-stream MMDiT blocks followed by unified single-stream DiT blocks. Post-training applies human-feedback preference optimization for editing and reward-based reinforcement learning for generation, and the distilled AuK-Flash performs 4-step inference without classifier-free guidance at a 4.5x wall-clock speedup. Source code and model weights are released.

Hugging Face daily papers·Sep 7, 20:00 UTC · 8d agoModel release2

Opaque recurrence, and other AI terms that you should probably know

TechCrunch updates its plain-English glossary defining common AI terms from AGI and agents to chain-of-thought reasoning.

TechCrunch maintains a regularly updated glossary of AI terminology, defining terms such as AGI, AI agents, API endpoints, chain of thought, coding agents, compute, deep learning, and diffusion. It highlights 'opaque recurrence', the reasoning technique in OpenAI's new Astra model that has drawn attention from AI safety researchers. The piece is an educational living document rather than new research or a product announcement.

TechCrunch · AI·Sep 7, 19:24 UTC · 8d agoAI industry1

DataFlex-RL: An Evaluation Platform for RLVR Data Policies

DataFlex-RL benchmark of 13 RLVR data policies on Qwen2.5-7B finds none reproducibly beats uniform sampling under matched GRPO training.

DataFlex-RL is an evaluation platform comparing rollout-selection, reweighting, and mixture data policies for RLVR under a common GRPO recipe. Across 13 configurations and 12 matched seeds with Qwen2.5-7B-Base on 12 math, logic, and science benchmarks, uniform GRPO improved domain-balanced accuracy by 7.76 points, but no alternative policy achieved a statistically significant improvement. A corrected 12-seed Llama-3.1-8B-Base extension found no consistent winner, and math-heavy evaluation summaries were negatively correlated (-0.33) with domain-balanced summaries.

Hugging Face daily papers·Sep 4, 20:00 UTC · 11d agoAI research

ENCP: Episode-Normalized Conformal Prediction for Vision-and-Language Navigation

ENCP calibrates conformal prediction per navigation episode, giving step-level coverage guarantees for vision-language navigation agents despite within-episode dependence.

Episode-Normalized Conformal Prediction (ENCP) rescales a nonconformity score by a VLN policy's residual confidence and calibrates one maximum score per episode, preserving step-level coverage of at least 1−α despite dependence among steps within an episode. Across four VLN policies and three nonconformity scores on R2R and REVERIE, ENCP meets all reported empirical step-coverage targets in seen-to-unseen evaluation. The model-agnostic uncertainty estimates can signal when an agent should defer to a stronger predictor or human assistance.

arXiv cs.AI / cs.LG / cs.CL·Sep 15, 17:42 UTC · 22h agoAI research