Tous 676 🔐 Cybersécurité 417 🤖 Intelligence Artificielle 253 💻 Tech & Transformation Digitale 6
Chargement…
🧠
Intelligence Artificielle

Beyond Imitation: Filtering On-Policy Distillation by Reasoning Progress

arXiv:2608.19408v1 Announce Type: new Abstract: On-policy distillation (OPD) has emerged as an effective framework for post-training language models by pairing student-generated trajectories with dense token-level supervision from a teacher. However, OPD implicitly assumes that teacher-derived rewards are an appropriate proxy for reasoning progress, and therefore treats all…

Lire l'article
🧠
Intelligence Artificielle

Bounded Sovereignty and the Control Tax: Pricing AI Oversight When the Deployer Does Not Own the Model

arXiv:2608.19216v1 Announce Type: new Abstract: AI control research asks how to deploy models safely even when they may be misaligned, but many control protocols assume that the deployer can instrument the model and its surrounding pipeline. That assumption often fails for regulated organisations using frontier models through APIs or managed endpoints, where the deployer may…

Lire l'article
🔐
Cybersécurité

Critical NetScaler Flaw Can Bypass Authentication on Certain Gateway and AAA Servers

Citrix has released updates to address two security flaws impacting NetScaler ADC and NetScaler Gateway deployments, including a critical-severity authentication bypass vulnerability. According to the cloud computing and virtualization technology company, the issues affect customer-managed NetScaler ADC and NetScaler Gateway, including certain FIPS and NDcPP builds, as well…

Lire l'article
🧠
Intelligence Artificielle

Nine Emotion Centroids: A Label-Free Valence Axis That Transfers Across Four Modalities

arXiv:2608.18090v1 Announce Type: new Abstract: Inside a modern language model sits a single internal direction that tracks how positive or negative a sentence feels. We show how to find this valence axis (V-axis) from just 9 emotion category names plus 50 short narrative paragraphs per emotion -- about 1,500 fewer labels than the usual supervised approach -- and that the…

Lire l'article
🔐
Cybersécurité

US charges Iranian hackers over $3.4 billion intellectual property theft

The U.S. has charged 17 Iranians, alleged members of a hacking-for-hire company called Mabna Institute, involved in years-long operations that stole data from American organizations. [...]

Lire l'article
🔐
Cybersécurité

StopAndProtect Uses Nearly 2,000 Hacked WordPress Sites to Spread Malware and Steal Data

Cybersecurity researchers have flagged a global cybercrime operation that abuses thousands of hacked WordPress websites as infrastructure to disseminate malware, commandeer infected hosts, store stolen documents, screenshots, and activity logs created to track the status of the activity. "The operation doesn't rely on a single piece of malware, but on a whole toolkit of…

Lire l'article