All published intelligence, sovereign stack research, ecosystem news, and AI foundation model reviews.
Газпромбанк ищет решения от научных команд в финтех-акселератор Трек «Финтех.Наука» рассчитан на исследователей из вузов, НИИ и R&D-центров, у которых есть научный задел в IT и желание довести его до продакшена в банке. Готовый прототип не обязателен. К научной экспертизе…
Here are Google’s latest AI updates from September 2026
🏢 OpenClaw при поддержке Nvidia и Red Hat запустила открытую платформу для управления корпоративными ИИ-агентами OpenClaw Enterprise позволяет компаниям централизованно контролировать постоянно работающих агентов, которым нужен доступ к внутренним данным и инструментам.…
В Meta предложили мета-рассуждение для агентов Это способ управлять самим ходом работы агентов, а не только решать задачу. По мере того, как ИИ-агенты берутся за сложные и долгосрочные задачи, возникает проблема не только что делать, но и как управлять всем процессом работы…
✔️ OpenAI заблокировала атаку дистилляции скрытых CoT своих моделей Компания пресекла кампанию по состязательной дистилляции, направленную на извлечение скрытых цепочек рассуждений. Ключевой кластер активности компания связала с Moonshot AI. В атаке участвовало 4 тысячи учетных…
Про новую макроэкономику ИИ Один из крупных разметчиков данных для ИИ-лабораторий Mercor сравнил работу бухгалтеров с ИИ-моделями на реальных задачах и предлагает иначе мерить момент, когда модель обходит человека. Вообще сейчас пошел тренд на ИИ-бухгалтерию, тут кейсы Компания…
🦠 OpenAI обнаружила новый тип атак на ИИ-агентов — вредоносные промпты научились «размножаться» Атака получила название self-replicating prompt injection. Скрытая инструкция заставляет агента не только выполнить действия злоумышленника, но и скопировать саму себя в другие…
✔️ Anthropic раздаёт бесплатные плюшевые игрушки и стикеры На сайте для разработчиков claude.dev появилась раздача фирменных подарков. В разделе Терминал нужно ввести команду /plushies, чтобы оформить заявку на плюшевого Clawd, маскота Claude Code, или /stickers, чтобы получить…
DeepSeek выложила в открытый доступ библиотеку, которая ускоряет вычисления на GPU, лежащие в основе ИИ-моделей. Она называется DeepGEMM. Библиотека оптимизирует матричные операции, на которых работают большие языковые модели, и позволяет эффективнее использовать NVIDIA Hopper и…
🎓 ИИ-репетиторы уже могут давать такой же быстрый прирост знаний, как занятия с экспертами Новое исследование проверило ИИ на подготовке студентов к GRE. По его результатам, обучение с помощью моделей давало сопоставимый с экспертами прирост результатов, а лучше всего среди…
How should an embodied agent respond when a person's correction may be wrong? We formulate grounded correction arbitration as a choice among accepting, rejecting, inspecting the world, and asking the speaker. GAVA implements this interface with observation-bounded evidence,…
Role-specialized QA pipelines increasingly pass rationales from a reasoner to a verifier, but it is unclear what this message actually buys: better answers, stronger support assessment, or a new failure surface. We introduce a message-intervention diagnostic that fixes the…
A persistent agent must decide both what to retain as information arrives and what to surface once a query appears, yet memory evaluations can confound these decisions by comparing methods that differ in both retention and selection. We build a streaming-recall benchmark…
A sound judgment applies the law to established facts and weighs the circumstances in which they arose. However, existing methods swing between rigid statute matching and ungrounded discretion, benchmarks score a label or a rubric, and the experience that would supply the…
Public-service recommendations require evidence that matches the requested service, scope, and date. Yet treating every missing detail as decisive can withhold useful recommendations. We introduce EviGraph, which distinguishes critical decision requirements from information that…
Detailed profession-specific system prompts raise token use and estimated cost per response without a consistent accuracy gain. We evaluate Scientific Agents, an open-source corpus of 503 profession-specific AGENTS.md profiles, with Gemini 3.8 Flash via OpenRouter in the Pi…
K-Dense BYOK (bring your own keys) is a free, open-source AI research assistant for scientists in any field that runs on the researcher's own computer. The researcher supplies access to a model of their choice, hosted or running locally, and the application supplies everything…
Personalizing large language models (LLMs) requires aligning generation behavior with user-specific preferences rather than aggregate quality. While Direct Preference Optimization (DPO) provides a stable framework for preference learning, its effectiveness in personalized…
When selecting mathematical training data for LLMs, a natural organizing principle is topic: probability examples for probability targets. An alternative is reasoning approach: worked solutions that share a solution method with the target, even when the mathematical domain…
We investigate automated rewards for training language models in conversational humor, focusing on reward exploits and countermeasures. Two approaches aim to capture understandable surprise and predicted audience amusement. Controlled tests show that an embedding-based surprise…
Resource limits are becoming an authorization boundary for AI agents that delegate work across concurrent and failure-prone workers. Parent-child allocation constraints, affine objects, and distributed escrow do not by themselves prevent overspend when replies are lost, effects…
Evaluating large language models becomes increasingly challenging as their capabilities advance: benchmarks can saturate, public test sets risk contamination, and assessing harder tasks can require expensive grading or execution infrastructure. These challenges are amplified in…
Large language model agents rely on external harnesses to pass information between the model and its environment and to recover from execution errors. Yet recovery is usually judged only by average task success. This hides an important tension. The same operation can rescue a…
Diversity collapse in parallel chain-of-thought has motivated inference-time interventions built on a natural design: when a process reward model (PRM) prunes a chain, its high-PRM prefix is extracted and grafted verbatim as an in-context demonstration into a still-decoding…
When an informed adversary shares the audience of a constrained signalling channel, the signal that best protects the truth is the signal that best describes it. On 108 confirmatory items, the adversary-robust optimum aligns exactly with the salience pole from prior work. Across…
Structured tool calls often fail after only a small number of fields violate a schema or an execution contract. Regenerating the complete object enlarges the action surface and makes repeated repair difficult to audit. We introduce ContractRL, a contract-constrained sequential…
Long-horizon language agents increasingly rely on external memory as a frozen world model, yet current memory systems are usually judged only by task success or token cost. We argue that the missing object is the shape of memory use: under finite context and repeated retrieval,…
LLM agents generate intermediate reasoning and actions token by token, making extended interactions slow and computationally expensive. Jev-style models offer fast probabilistic predictions over finite fields, but require those fields to be specified in advance. This requirement…
LLM agents increasingly act through modular systems, such as order, payment, inventory, and shipment services, where actions in one module change which transitions are valid in another. Standard world models usually fit observational traces, but this is not the quantity needed…
Research agents explain planned experiments. We measure predictive credit with paired forecasts sharing an intervention, forecaster, and outcome while varying description, matched explanation, and donor context. Five checks track commitment, delivery, predictive gain, alignment,…