Attention Mechanism

技术

Attention Mechanism，一种使模型在处理数据时动态聚焦于关键部分的神经网络技术，典型应用于 Transformer 等大语言模型

197 次提及156 个连接首次出现: 2026-03-30最近出现: 2026-07-29

关系图谱

关系 (161)

使用技术 (123)

AA-sLSTM Transformer GATv2 GAT MSA CubeComposer DINO LongCat-AudioDiT Gemma 4 Kimi Moonshot emojiGPT TouchAnything GEN-1 HISA PMDformer LaSM PG-VTON InternVL2-8B SEMF LoRA ERA-42 Claude 3.5 Sonnet Claude Opus 4.6 OneStory CARE Transformer LLaVA-OneVision Agent Being-H0.7 BioMD FreeLOC KV Cache 3D World Policy LingBot-Map GPT-2 月之暗面 Google DeepMind Transformer QuatRoPE MoE Vector Database Claude Mythos HedgeMamba Mamba AtomVLA STCast Fast-WAM softmax LitePT DeepSeek V4 DeepSeek V4-Pro DeepSeek-V4-Flash DeepSeek V3 深度求索 DeepSeek ELTDF-Net Xiaomi-Robotics-0 LTX-2 MTSS LDA-1B Qwen3 通义千问 Max 神经计算机 RecBERT OneTrackerV2 STCast 深度学习 MiniCPM-V 4.6 Claude VGGT-Ω ViT³ OFA-TAD o1 BinaryAttention FlexAttention INTRA OSCAR NaLaFormer 线性注意力 AutoMoT Gamma-World 上海交大 GPT-4o 稀疏注意力 UniCM HorizonStream DiffusionGemma 腾讯 Q-K=V MiniMax M3 GuidedVLA TranscriptFormer GLM-5.2 Claude Fable 5 Hybrid Linear Diffusion Transformer HappyOyster 1.0 CameraSquad Unlimited OCR Spatial-TTT NEO-ov Artic REViT MOSS-Audio MedP-CLIP VLX系列 Flux.1-Kontext-dev QwenImage-Edit-2511 OpenDecoder FlashBlock EchoAttention Afford-VLA Claude 3.5 Haiku LingBot-World 2.0 MagicBokeh RankMixer Inkling Kimi K3 SenseNova U1 Pro SoftDelta Harness NVIDIA FireAttention AnyStyle LPU CorrAdapter

应用于 (31)

基于 (4)

Token Dictionary Cross-attention GCA MoE Transformer

创建 (2)

Noam Shazeer Geoffrey Hinton

发布 (1)

相关文章 (197)

下滑加载更多...（已显示 30 / 197）