Барлық жарияланған талдаулар, егеменді стек зерттеулері, экожүйе жаңалықтары және ЖИ модельдерінің шолулары.
"Train a model on the same problems written under two incompatible conventions, both correct, and ask what the ordering of that data writes into the parameters. The learning-rate schedule is not a background condition for that question. It is the averaging operator, and it…
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,…
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,…
Scientific coding agents receive equations, boundary conditions, and output requirements in writing, then must assess the programs they revise. Rules to Tools (R2T) supplies prepared executable checks of public scientific requirements. Matched SciCode repair groups share written…
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…
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…
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…
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…
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…
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 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…
Large-language-model agents can propose and execute actions, but proposal, authority, dispatch, verified external effect, and serving promotion are different claims. We present Praxa, an agent harness that represents these states explicitly through deterministic admission,…
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…
Agent harnesses increasingly want to run small language models (SLMs) on the microtasks around a frontier large language model (LLM) planner: auto-approving shell commands, writing memory, selecting tools, ranking past turns. We ask whether off-the-shelf SLMs meet…
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…
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…
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…
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…
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…
The book “Disciplined Entrepreneurship” by Bill Aulet, managing director of the Martin Trust Center for MIT Entrepreneurship and the Ethernet Inventors Professor of the Practice at the MIT Sloan School of Management, walks readers through the 24 steps of starting a venture. With…
Supabase just launched their own eval, and what's crazy is that Opus 5.5 almost lost to an open-source model. But, the point of this benchmark is not to find out which model is smartest. Instead, they drop each agent into a Supabase project and give it actual jobs. Build a…
Hello everyone, welcome to another episode of AI Builder. Today we will see what's the difference between Copilot Chat and Copilot Co-work and when to use which. Think of Copilot Chat as your AI conversation partner. You want to summarize email or you want to draft a response,…
NVIDIA сообщает о запуске GPT-6 Astra Ultrafast с GPU Blackwell, ускоряющего генерацию токенов.
Chatham Financial uses Codex and GPT-5.6 to build technology and redesign workflows, cutting trade validation from 30 minutes to under 4.
MIT представил InstructMesh — новый инструмент для 3D-дизайна и печати.
Asking AI companies to self-regulate is a great way to pretend like you’ve accomplished something.
“What you see is what you get” is a guiding principle for many software engineers — create programs where the content you’re editing looks the same as the final product. But when you’re using generative artificial intelligence systems to 3D print, say, a mug, you’ll likely get a…
OpenAI представила новые функции покупок в ChatGPT, включая виртуальную примерку.
President Trump reportedly asked for Grok's opinion before invading Venezuela and capturing Nicolás Maduro.
September 2026 was a packed month for the robotics industry. Live events, including IMTS, ROSCon, and iROS, brought a flood of developments around automation and AI. This month, our top 10 is filled with new products, company launches, and major changes for some robotics…