Актуализира се ежедневно2293 проверени истории
AI новини. Без шума.
Проверено от източници AI покритие на пускането на продукти, промените в политиките, изследванията на безопасността и индустриалните ходове, обяснено на ясен език от образователен екип на неправителствена организация.
Потвърден източник
Всяка история се свързва с най-силните налични доказателства: оригинални източници, когато са налични, иначе ясно приписани доклади.
Обикновен английски
Какво се случи, защо има значение и какво да гледаме - без жаргон.
Без пълнител
Когато сигналът е слаб, ние не публикуваме нищо, вместо да допълваме емисията.
Още истории
9 историиИновация
Preprint describes a multi-agent AI framework for analyzing Chinese herbal formulas
A new arXiv preprint describes DeepTCM1.0, a system using 11 specialized AI agents and iterative review to analyze possible mechanisms of traditional Chinese medicine formulas. It reports an evaluation design but no numerical results, clinical validation, or evidence of improved patient care.arxiv.orgСигурност
Ph.D. dissertation examines backdoor attacks in language and vision-language models
An arXiv-listed Ph.D. dissertation studies how backdoor attacks can affect language models and vision-language models, including methods for analysis, detection and attack design.arxiv.orgИновация
Preprint reports higher LLM candidate ratings for prestigious institutions
An arXiv preprint reports that four large language models rated candidate profiles more favorably when linked to higher-prestige institutions and journals. The authors say institutional and geographic cues can shape evaluation, while the tested models and professional domains remain unspecified in the supplied source.arxiv.orgИновация
Study finds inference context can change LLM outputs in medical allocation scenarios
An arXiv paper reports that three of four tested language models shifted resource-allocation probabilities differently when the same clinical scenario was evaluated with or without the model’s prior response in context.arxiv.orgИновация
Preprint reports AI decoding of natural sentences from non-invasive brain recordings
A research team describes Brain2Qwerty v2, a model that uses real-time MEG recordings to decode typed natural sentences, reporting a 39% average word error rate across nine subjects.arxiv.orgИновация
Study maps 46 language models built for Portuguese
A systematic mapping study catalogs 46 Portuguese language models and compares architectures, training resources, licensing, code, data, and weights. The authors say the field is growing but difficult to assess because information is spread across papers, technical reports, repositories, and project documentation.arxiv.orgИновация
Interpretability study links Qwen3-4B’s overconfident answers to a certainty-biased mechanism
An arXiv study reports that Qwen3-4B favors certainty over uncertainty in controlled reasoning tasks and identifies model features that may drive the imbalance. The authors say targeted interventions reduced overconfident errors, but the abstract does not disclose effect sizes or testing details.arxiv.orgИновация
The Deontic Gap: Large Language Models and the Modal Language of Obligation
Modal auxiliaries such as must, should, and have to mark necessity and obligation within the contexts of speaker authority and interpersonal stance.arxiv.orgСигурност
Paper proposes obscuring refusal signals to resist abliteration attacks
An arXiv preprint introduces a weight-editing method intended to make safety refusals harder to extract and remove. The paper reports stronger post-abliteration refusal scores on two open models, with different tradeoffs in general-purpose performance.arxiv.org
Един полезен брифинг всяка седмица
Поддържайте изкуствения интелект, без да живеете в потока.
Получете потвърдените новини за AI за седмицата, оригинални данни, полезни инструменти, учебни предложения и нови AI работни места.
Достигнете до хора, които учат AI
Наемане на AI професионалист или пускане на полезен AI продукт? Представете го пред хора, които са дошли тук, за да учат и действат.
Публикувайте работа за AIИзпратете AI инструмент