最新
Open Standard for Tool Use and Data Access Gains Momentum OpenAI adopts Model Context Protocol to boost LLM tool integration
<!--kg-card-begin: html-->Loading the Elevenlabs Text to Speech AudioNative Player...<!--kg-card-end: html-->OpenAI embraced Model Context Protocol, providing powerful support for an o
Open Standard for Tool Use and Data Access Gains Momentum
OpenAI adopts Model Context Protocol to boost LLM tool integration
Toward LLMs That Understand Misspellings New byte-based model beats Llama 3 on spelling, noise, and translation
<!--kg-card-begin: html-->Loading the Elevenlabs Text to Speech AudioNative Player...<!--kg-card-end: html-->Researchers built a model that’s more robust to noisy inputs like misspelli
Toward LLMs That Understand Misspellings
New byte-based model beats Llama 3 on spelling, noise, and translation
Linear-MoE统一序列建模，长文本处理迎来革命性突破
学术界与产业界联合研发的Linear-MoE框架，通过统一线性序列建模与混合专家系统，实现百万token长文本处理速度提升3倍，或定义下一代模型架构标准。
OCTS算法突破LLM推理瓶颈，逆长尾问题终得解法
新型OCTS算法通过答案聚类与动态停止机制，有效缓解LLM推理中的逆长尾效应，在复杂任务中提升45%响应速度，降低30%计算资源消耗。
Dynamic Tanh革新Transformer架构，MetaAI重塑归一化技术标准
MetaAI推出Dynamic Tanh（DyT）技术，通过tanh函数替代传统LayerNorm，在H100s显卡上实现训练加速与成本降低，推动多模态Transformer高效化部署，或成下一代模型标准。