NIRC is based on the National Key Laboratory of Network and Switching Technology at Beijing University of Posts and Telecommunications. It has built a technology matrix across four AI domains—intelligent cloud networking, natural language processing, computer vision, and machine learning systems—dedicated to solving real‑world problems, creating top‑tier systems, publishing high‑impact papers, and contributing significantly to the rapid advancement of China's network technology.
The article demonstrates how Nano Banana Pro, a generative AI drawing tool, can rapidly prototype research figures and transform textual descriptions into clear logical diagrams, dramatically improving visualization efficiency while adhering to academic aesthetic standards.
This guide walks graduate researchers through a step‑by‑step workflow—collecting papers with Zotero, translating them via an LLM plugin, generating structured markdown notes, and then using Obsidian’s bidirectional links and Canvas to build a local, visual knowledge graph that ties individual citations into a coherent research map.
TileLang is a Python‑style DSL built on TVM that separates algorithm logic from hardware scheduling, offers beginner to expert interfaces, supports multiple GPU and CPU backends, and delivers performance on par with or better than existing AI kernels, as demonstrated with GEMM, FlashAttention and other benchmarks.
The article explains OpenTelemetry’s three core components—SDKs, Collector, and Operator—detailing how the Operator’s automatic injection simplifies Kubernetes deployments and how the modular Collector can export telemetry to any backend such as Jaeger.
The article defines mechanistic interpretability as reverse‑engineering LLMs to reveal how they represent knowledge and make decisions, explains its importance for transparency, risk mitigation, and model improvement, and surveys key techniques such as causal tracing, zero‑making, noise‑making, and logit‑lens methods with illustrative examples.
This article explains why modern LLM‑based applications need agent capabilities, introduces LangGraph’s core features such as stateful execution, graph‑based orchestration, tool integration, human‑in‑the‑loop and multi‑agent support, and provides a step‑by‑step Python example that builds a simple chat‑bot agent.
SEAgent introduces a self‑evolving framework that enables a GUI agent to master unfamiliar software through autonomous exploration and experience learning, leveraging a curriculum generator, a world‑state model, and GRPO‑based reinforcement with adversarial imitation, achieving state‑of‑the‑art performance on OSWorld.
AutoCCL analyzes NCCL’s six key performance parameters, uses coordinate‑descent and an online leader‑worker architecture to automatically adjust them during training, overcoming state‑space explosion and compute‑communication interference, and achieves 1.07‑1.32× faster iteration times on models such as Phi‑2, Llama‑3.1‑8B and VGG‑19.
At Beijing University of Posts and Telecommunications' 70th anniversary, the Network Intelligence Research Center showcased a micro‑gesture‑driven VR system that captures millimeter‑scale finger motions with high‑precision, low‑latency hand tracking, delivering efficient, fatigue‑reducing interactions and earning strong audience approval.
Researchers present LucaOne, a Transformer‑based foundation model that unifies DNA/RNA and protein sequences using a 39‑token vocabulary, rotary positional encoding, and molecule‑type embeddings, and demonstrate through extensive multi‑task benchmarks that it outperforms domain‑specific models across seven biological tasks.
