Presents TRACE, a training-free framework for GUI agents that reduces inference latency and memory from accumulated high-resolution screenshots via visual…
Describes Benchmark Radar, an infrastructure system that aggregates data from 37 sources into a living database of 1,283 benchmark records…
Introduces DataFlex-RL, a platform for comparing rollout-selection and reweighting data policies used in reinforcement learning with verifiable rewards (RLVR) under…
Proposes Posterior Label Correction DPO (PLC-DPO), which routes each preference pair as clean, flipped, or tied using the calibrated policy-reference…
Details a data-centric training framework built from five complementary systems (Choulea, SkyReal, Hongzwang, PSBreakup, and Kreator) used to construct specialized…
Studies adaptive routing of prompts across multiple LLM “experts” to maximize response quality under limited feedback, formulated as a K-armed…
Formalizes task-agnostic environment preprocessing, where an LLM-agent “studying system” explores a new environment under a budget and builds reusable artifacts…
Technical report on StepAudio 3 Gen, a unified audio-generation system covering zero-shot TTS, voice design, vocal and sound-effect generation, and…
During a panel at the All-In Summit in Los Angeles on September 14, 2026, Nvidia CEO Jensen Huang took a…
Microsoft published its first formal Code of Conduct for its in-house MAI models on September 14, 2026, establishing outright bans…
Apple announced the next generation of Apple Intelligence on September 14, 2026, centered on Siri AI, a rebuilt assistant with…
Shanghai Artificial Intelligence Laboratory released Atria Dawn Preview, a 744-billion-parameter open-weight agentic AI model, under an MIT license with no…
