每天自动聚合 AI 领域最新动态
World generative models are typically used through what they produce: a rendered future, a video-conditioned action, ...
Test-time scaling is often implemented by spending more compute along one axis: sampling more solutions, extending a ...
LLM inference accounts for over 90% of AI operational energy, scaling directly with input token count---a critical in...
Long-term memory enables language agents to reuse past facts, preferences, and task experience. Persistence also crea...
Diffusion Large Language Models (DLLMs) replace autoregressive next-token prediction with iterative parallel denoisin...
Vision-language models (VLMs) are improving rapidly, but benchmark development lags behind, making weaknesses hard to...
Under the standard split, Muon gets hidden matrices and AdamW embeddings/output head. Muon groks modular addition fas...
Reliable hypothesis testing is the foundation of many empirical scientific claims. Large language model (LLM) agents ...
Human-like cognition does not select past experience by topical similarity alone: affective significance and unresolv...
Agentic coding faces growing problems of affordability and wasted tokens. We introduce Blast Radius, a predictive mem...
Rapid adoption of large language models (LLMs) in enterprise settings has introduced operational, security, and gover...
LLM agents increasingly adapt to recurring tasks by accumulating procedural knowledge in skills. These skills are lig...