Tencent Hunyuan maps 549 papers on self-evolving AI agents into a five-level taxonomy, and introduces a reliability ladder to ensure updates can actually be trusted
Ant Group's InclusionAI releases a 7.9B MoE reasoning model that matches GPT-class performance with just 1.3B active parameters, free and MIT-licensed.
Microsoft Research's Dion3 optimizer slashes Muon's orthogonalization overhead by up to 6x through smarter math, custom kernels, and selective row updates.
A Salesforce and UIUC team shows a strong builder model can hand-craft inference-time scaffolds that nearly double a weaker model's accuracy without touching its weights.
Artificial Analysis launches AA-AnalystAgent, a new benchmark that tests AI agents on real spreadsheets — and even the best model only passes 54% of tasks every time.
Google's AMIE now conducts live video consultations using a three-agent architecture, matching primary care physicians across all core clinical metrics in a 300-consultation randomized trial.
SGLang's Unified Radix Cache replaces a growing matrix of specialized cache classes with one composable tree, unlocking proper prefix caching for hybrid models like DeepSeek-V4 and Inkling.
Databricks acquires ElectricSQL to embed a full Postgres database inside every AI agent sandbox, syncing local state back to its Lakebase cloud platform
NVIDIA's open 30B MoE model with only 3B active parameters targets the high-volume execution layer of always-on AI agents, delivering 4x faster output speed than comparable models