Aggiornato quotidianamente1797 storie verificate
Copertura controllata dall'IA su lanci di prodotti, cambiamenti di politiche, ricerche sulla sicurezza e mosse del settore, spiegata in un linguaggio semplice da un team educativo non profit.
Ogni articolo è collegato alle prove più forti disponibili: fonti originali quando disponibili, altrimenti segnalazioni chiaramente attribuite.
Cosa è successo, perché è importante e cosa guardare, senza usare termini tecnici.
Quando il segnale è debole, non pubblichiamo nulla invece di riempire il feed.
Storie di intelligenza artificiale controllate alla fonte, prima le più recenti, per le persone che hanno bisogno di comprendere l'intelligenza artificiale senza inseguire l'hype.
A new benchmark reports that eight large language models frequently fail to detect conflicting evidence or preserve correct analyses when spreadsheets and tables are altered.
A new arXiv preprint reports that automated metrics and LLM judges often fail to match human assessments of creativity in short stories, with judges systematically favoring AI-generated writing styles.
An arXiv preprint reports that representation, structural screening and ranking choices can substantially alter which AI-generated Big Five items reach psychometrician review, even when global summaries appear stable.
A preprint reports that four multilingual eighth-grade students challenged LLM classifications of their math discussions, arguing that adult annotations and standard scores can miss students’ own interpretations.
A new arXiv study reports that giving language models access to stored examples improves their sensitivity to syntactic contrasts involving rare words, although the gap is not eliminated.
An arXiv study proposes a two-layer detector for package-name hallucinations in local coding language models, reporting that hallucination rates rose from 0–10% on routine prompts to 40–73% on adversarial slopsquatting prompts.
A new arXiv paper introduces AgentSpec, a speculative-decoding method designed to reduce response-time degradation when LLM agents run in large batches. The authors evaluate it across five workloads and four models from four LLM families in vLLM.
A new arXiv preprint proposes measuring the geometry of model inference to identify deceptive behavior that conventional linear probes may miss.
A new benchmark reports that vision-language models often changed a correct chest X-ray answer after receiving conflicting text or prior-image context, with misleading text exerting a stronger pull than misleading visual context.
A new EMNLP 2026 paper finds that large language models often overproduce long responses in counseling settings, where brief acknowledgments can better support attentive listening and continued disclosure.
A new arXiv paper describes PARTAB, a framework that selects relevant table regions before an AI model answers questions, reporting improved results on several table-reasoning benchmarks and reduced reasoning context.
The Decoder reports that a UK-Ukraine defense partnership gives approved British firms access to Ukraine’s manually labeled combat imagery and sensor data for military AI development. The arrangement is described as a secure, Ukraine-controlled collaboration involving drone detection, acoustic sensing and low-power…
Un briefing utile ogni settimana
Ricevi le notizie verificate sull'IA della settimana, dati originali, strumenti utili, scelte per l'apprendimento e nuovi lavori in IA.
Assumere un professionista dell'IA o lanciare un prodotto AI utile? Mettilo davanti a chi è venuto qui per imparare e agire.
Pubblica un lavoro AI Invia uno strumento AI