DeepSeek Upgrades DeepSeek-V4-Flash-0731 with Major Agentic and Coding Gains
8 月 1, 2026admin NUAI,Committee,新闻,Uncategorized(0)DeepSeek published DeepSeek-V4-Flash-0731 on Hugging Face and moved the official V4-Flash API into public beta...
DeepSeek Researchers Apply a 1967 Matrix Normalization Algorithm to Fix Instability in Hyper Connections
1 月 4, 2026admin NUAI,Committee,新闻,Uncategorized(0)DeepSeek researchers are trying to solve a precise issue in large language model training. Residual...
DeepSeek Releases DSpark, a Speculative Decoding Framework That Accelerates DeepSeek-V4 Per-User Generation 60–85% Over MTP-1
6 月 27, 2026admin NUAI,Committee,新闻,Uncategorized(0)DeepSeek released DSpark, a speculative decoding framework, with open-source checkpoints and training code. It is...
DeepSeek R1-0528 arrives in powerful open source challenge to OpenAI o3 and Google Gemini 2.5 Pro
5 月 30, 2025admin NUAI,Committee,新闻,Uncategorized(0)Additionally, the model’s hallucination rate has been reduced, contributing to more reliable and consistent output.Read...
DeepSeek AI Researchers Introduce Engram: A Conditional Memory Axis For Sparse LLMs
1 月 15, 2026admin NUAI,Committee,新闻,Uncategorized(0)Transformers use attention and Mixture-of-Experts to scale computation, but they still lack a native way...
DeepSeek AI Releases DeepSeek-OCR 2 with Causal Visual Flow Encoder for Layout Aware Document Understanding
1 月 30, 2026admin NUAI,Committee,新闻,Uncategorized(0)DeepSeek AI released DeepSeek-OCR 2, an open source document OCR and understanding system that restructures...
DeepResearch Bench: A Comprehensive Benchmark for Deep Research Agents
6 月 16, 2025admin NUAI,Committee,新闻,Uncategorized(0)arXiv:2506.11763v1 Announce Type: new Abstract: Deep Research Agents are a prominent category of LLM-based agents...
DeepReinforce Team Introduces CUDA-L1: An Automated Reinforcement Learning (RL) Framework for CUDA Optimization Unlocking 3x More Power from GPUs
8 月 3, 2025admin NUAI,Committee,新闻,Uncategorized(0)Estimated reading time: 6 minutes Table of contents The Breakthrough: Contrastive Reinforcement Learning (Contrastive-RL) How...
DeepReinforce Releases Ornith-1.0: An Open-Source Coding Model Family That Learns Its Own RL Scaffolds
6 月 25, 2026admin NUAI,Committee,新闻,Uncategorized(0)DeepReinforce has released Ornith-1.0, an open-source model family built for agentic coding. The lineup spans...
