Discover latest generative AI models, prompt engineering papers, and MCP servers

Menu Close

首页
data
training
models
inference
context
prompt
mcp
agents
关于

嗯，目前共计420篇文章

2024

4月 29, 2024

DPO Meets PPO: Token级强化优化统一RLHF范式

4月 02, 2024

长上下文 LLM 在长上下文学习中的挑战

Search for:

标签云

0528-update 2-bit 量化 3-bit-4-bit 3d-parallelism 4-bit-quantization 405b 671b 70b 8-bit 训练 AB-MCTS ACE框架 AGI AI AI Agent AI Agents AI 代理 AI 加速器 AI-agents AI安全 Agentic RAG Agent架构 Agent系统 Alignment Anthropic Attention优化 Blackwell GPU Bradley-Terry 模型 COCONUT COSTAR CPU卸载 CUDA kernel Chain-of-Thought ChatGPT Claude Claude-4 CoT-limitations Common Crawl Consumer-GPU Context Engineering Continuous Batching D-最优设计 DES DMS DPO DSPy DeepSeek DeepSeek R1 DeepSpeed Development Direct Preference Optimization Dueling Bandit EM-LLM Expected-Attention FLOPs 优化 FP4 量化 FP8 FlashAttention FlashInfer GEMM 优化 GPT-5 GPTQ GPU加速 GRPO GSPO GTPO GUI导航 GVote Gemini Gemini-2.5 Google Graph of Thoughts GraphRAG Grok Grok-4 H20 GPU Hadamard 变换 Hadamard 旋转 ICL ICLR 2025 (Poster) ICLR-2026 Instruction-Tuning K-means聚类 KV Cache KV Cache 量化 KV Cache量化 KV-cache KVPress KV缓存 KV缓存优化 KV缓存压缩 Kernel优化 LCNC LLM Alignment LLM Serving LLM serving LLM 局限性 LLM 评分 LLM 预训练 LLM-as-judge LLM-serving LLM加速 LMS3方法 Llama Llama-4 LoRA LoRR LongICLBench MCTS MIPRO MLA MMS MOSS Megatron-LM Meta MoE MoE训练 Monte-Carlo Multi-Head Latent Attention NSA NVIDIA NeurIPS 2024 NeurIPS-2025 ORM Official OmegaPRM OpenAI OpenAI o1 PAMU PEFT PLENA PPO PRM PagedAttention Preference Learning Preference Optimization Productivity Profiling Prompt Engineering QLoRA Qwen Qwen3 RAG RAG综述 RLHF RTX-4060 ReAct Reasoning Reward-Model SFT SGLang SageAttention Saguaro Search SmoothQuant Speculative Decoding Survey Tensor Core Test-Time Compute Token-wise-Optimization Token优化 Tree-of-Thoughts Triton W8A8 量化 WGMMA WebScraping ZeRO-Offload accessibility acting adaptive-alignment adaptive-budget adaptive-reasoning african-languages agent agent-architecture agent-evaluation agent-harness agent-memory agent-model agent-optimization agentcore agentic-ai agentic-reasoning agi-research ai ai-agents ai-automation ai-coding ai-evaluation ai-feedback ai-gateway ai-integration ai-search ai4bharat aleph-alpha alerting alibaba alignment-methods alternative-paths analogical-prompting analogical-reasoning anthropic any-to-any apache-2.0 ape api apm arithmetic-reasoning assessment async-inference atlas atlassian atmospheric-science attention 优化 attention-budget attention-mechanism audio audio-model audio-visual automate-cot automated-annotation automatic-cot automatic-optimization automatic-prompt-engineering automatic-prompt-generation automatic-prompt-optimization automatic-prompting automation autorag autoregressive aws azure baas backtracking banking batched-generation beam-search bedrock benchmark benchmark-design best-practices billing biology biomedical-ai biosurveillance brave browser browser-automation budget forcing cache category-theory cdn chain-of-thought chat chat-model chemistry ci-cd classifier-free-guidance claude claude-3.5 claude-4 claude-4.5 claude-ai claude-api claude-capabilities claude-code claude-sonnet climate clip cloud cloud-services cloudflare cluster-management code-agent code-editing code-intelligence code-management code-quality code-reasoning code-review code-search coding coding-agents coding-model cognitive-psychology collaboration commercial communication comparative-study compass compositional-generalization comprehensive comprehensive-review compression computational-efficiency compute-scaling computer-use confluence consensus container container-management container-registry containers content-extraction context-engineering context-management context-window contrastive-filtering controlled-experiments conversation-continuity convertapi convolution coordination copilot cost-aware-optimization cost-effective cost-efficiency cost-efficient cot cpu-offload critical-analysis cross-modal-generation cultural-ai cursor custom kernel customer-management cypher damo data-access data-analysis data-commons data-distribution data-parallel data-parallelism data-processing database datadog debugging decoding decoding-strategies decoding-strategy decomposed-prompting decomposition deep-reasoning deepmind deepseek deepseek-r1 deepseek-v3 deepspeed deliberate-reasoning demonstration-selection dependency-parsing deployment derivative-free-optimization design-to-code developer-tools development devops devtools diagnostic-imaging diffusion discrete-flow distributed-training dns docker docker-compose docker-hub document-database document-processing documentation domain-adaptation domain-specific double-quantization drug-discovery dueling-bandit dynamic-prompt-refinement e-commerce e-learning earth-observation ecmwf ecommerce edge-ai edge-computing edge-deployment education educational-content educational-data efficiency efficient elasticsearch embeddings emergent-abilities engineering engineering-practice ensemble-method enterprise entities entity-extraction equall error-correction error-tracking european-ai evaluation-methods evo-cot evolutionary-merge evolutionary-methods experimental extended-thinking external-memory factual-accuracy factual-generation fast-inference few-shot few-shot-learning few-shot-reasoning figma file-converter file-system finance fine-tuning fintech flagship flan flash-attention flava formal-verification format-conversion foundation-model foundation-models fp8 free-tier french frontend fsdp full-text-search function-calling gates gdpr-compliant gemini gemini-2.0 gemma gemma-3 gemma-3n generalization genomic-ai german git github github-copilot glm gnn go google got gpt-3 gpt-4o gpt-5 gpt-5-pro gpu-kernel gpu-optimization gradient-based-optimization grading-systems gradio grafana granite graph-database graph-memory graph-of-thoughts graph-rag graph-retrieval graph-structured-knowledge greedy-vs-exploration grok grpo gui-navigation h100 hardware-aligned hashicorp hausa healthcare helm hessian-ai hessian-guided heuristic-sampling high-entropy-tokens high-throughput hindi html-parser http-client huggingface human-feedback hunyuan hybrid-architecture hybrid-retrieval hyena hyper-scaling iac ibm idefics image-conversion image-editing image-generation image-management image-text-pairs image-to-image image-understanding imo-gold implicit-reasoning in-context-learning inclusionai indic-languages indonesian inference-optimization inference-time-compute infinite-context influence-functions influxdb information-theory infrastructure instruct instruct-model instruction-guided instruction-optimization instruction-tuning integration interactive-reasoning interleaved-comprehension intermediate-computation internlm intrinsic-reasoning invoice iot iterative-refinement janus japanese java jest jira joint-optimization k8s knowledge-base knowledge-graph knowledge-intensive knowledge-management kubectl kubernetes kv-cache label-free language-model-programs language-models large-scale-corpora latency-optimization latency-reduction latent-capabilities latent-reasoning law leap learning-from-mistakes least-to-most-prompting legal legal-ai lfm2 ling liquid-ai liquidai llama llama-3.1 llama-3.3 llama-4 llm llm-agents llm-as-optimizer llm-evaluation llm-generation llm-judge llm-optimization lmarena lms loki long-context long-context-processing long-cot-reasoning long-running-agents long-term-memory low-code low-latency low-resource magistral mamba markdown markdown-conversion matformer math mathematical-reasoning mathematics matrix-decomposition mcp mcp-server mechanistic-interpretability media-conversion medical-ai megatron-lm meilisearch memory memory-management memory-optimization memory-system messaging meta meta-ai meta-learning meta-policy-optimization meta-prompt meta-prompting metadata metagenomic methodology metrics microsoft mistral mistral ai mistral-based mistral-small mit-license mixtral mixture-of-experts ml model-compression model-confidence model-merging model-scaling modular-prompting moe molecular-science mongodb monitoring monte-carlo-estimation mosaicml multi-agent multi-cluster multi-hop-reasoning multi-llm multi-scale-support multi-stage-optimization multi-step-reasoning multi-step-tasks multi-task-learning multiagent-systems multilingual multimodal multimodal-prompting mysql n8n nasa natural-language natural-language-optimization needle-in-haystack nemotron neo4j news-search nf4 nlp-tasks non-transformer nosql notion nvidia o1 o1-mini oauth observability observations official omni-modal omnimodal on-device oncall one-shot-quantization online-learning open-source open-source-models open-weights openai openhathi openshift operational-ai opro optimization-theory opus out-of-distribution outlier 抑制 overthinking paged-optimizers pandemic-monitoring parallel-compute parallel-decoding parallel-thinking parameter-driven-methods parameter-free-methods passage-retrieval pattern-matching payment payments pe2 performance-engineering persistence phd-level phi-4 physics pipeline-parallelism planning playwright pods policy-evolution post-training-quantization postgres postgresql ppo pre-training preference-learning principle-learning problem-solving process reward model process-management process-reward-models process-supervision production-ready productivity programmatic-tool-use progress-tracking progressive-quantization project-management project-organization prometheus prompt-engineering prompt-engineering-survey prompt-optimization prompt-optimization-toolkit prompt-selection prompt-self-optimization prompt-space prompting-methods proprietary protocol prover pseudo-demonstrations public-data public-health pubmed python pytorch qdrant qlora quantization quantization-finetuning question-answering qwen qwen3 radar rag rag-benchmarks rag-datasets rational-chain-generation react real-time real-time-search reasoning reasoning-efficiency reasoning-emergence reasoning-evaluation reasoning-limitations reasoning-model reasoning-optimization reasoning-scaling reasoning-uncertainty reasoning-without-prompting recurrent-architecture redis reference reflection reinforcement-learning relations remote-mcp repository representation-learning research retail retrieval-augmented-generation reward-hacking-prevention reward-model-optimization reward-modeling rlaif rlhf rust s1 safety sakana-ai scaling-laws scientific-ai scientific-computing scratchpad screenshot-html search search-agent search-algorithms search-engine security selective-guidance self-adaptive-prompting self-consistency self-feedback self-generation self-improvement self-refine self-rubrics semantic-search sentry sequential-thinking serper-api serverless session-management sitemap skeleton-of-thought slack small-but-capable small-language-model small-model smaller-models social-media software-engineering sonarqube sonnet southeast-asian-languages sovereign-ai sparse-attention sparsification speculative-decoding speech-generation speech-recognition sql sqlite sqlserver stagehand stanford state-space-model static-analysis statistics step-by-step strategy-elicitation stripe structural-prompting structure-aware-generation structured-input structured-thinking sub-task-delegation subgraph-extraction subscription supabase supervised-fine-tuning survey sustainability swahili swe-bench symbolic-reasoning synthetic-data tailwind task-decomposition task-loss-gradients taxonomy team-collaboration telegram telethon tencent tensor-cores tensor-parallelism terminal terraform test-time compute test-time-compute test-time-intervention test-time-reasoning testing testing-framework text-feedback-optimization text-to-image thai theorem-proving theoretical-analysis theoretical-framework thinking-mode thinking-model thompson-sampling thought-generation time time-series timezone together-ai token 效率 token-optimization token-reduction tool-calling tool-design tool-discovery tool-use top-k-decoding training-free transformer-llms tree-of-thoughts trillion-parameter tts type-theory typescript ui-design unified-framework unified-model unified-system upstash usp utility vLLM vLLM优化 vcs vector-search verifiable-rewards version-3.2 version-control video video-understanding vietnamese vision vision-language vision-language-models visual-prompting voice-ai voxtral vscode w4a8kv4 weather weather-forecasting web-agent web-research web-scraping web-search web-task web-ui webscraping woocommerce wordpress workers workflow-automation workspace x-platform xAI xai xhosa yoruba zero zero-offload zero-shot zero-shot-learning zettelkasten zhipu-ai zulu 上下文一致性上下文优化上下文学习上下文工程上下文窗口上下文管理专家级测试专家选择个性化中文支持主动学习事件分割事实准确性产品更新，AI 人类反馈人类可读提示人类认知模型代理架构代码代码生成任务分解任务无关企业服务企业转型，零售优化局限性低代码平台低比特量化低精度推理信息提取偏好优化偏好学习元提示内存优化内存管理分布式训练动态结构构建动态路由动态适应博客汇总即时编译去污染参数高效微调反思机制合成数据合成边界案例后训练向量量化吞吐量优化吞吐量提升图RAG 图像描述图检索图检索增强生成图神经网络在线学习块稀疏格式基于AI反馈的强化学习基准测试多Agent系统多专家多任务学习多任务理解多模态偏好学习多模态理解多语言多跳推理多适配器多重记忆系统大型语言模型大型语言模型提示优化大语言模型奖励模型实际能力宪法AI 宪法AI对齐对比学习对话对话对齐对齐小模型少样本学习层级索引复用工作负载均衡工具增强工具调用工具链幻觉缓解幻觉问题序列级优化开放式任务开源异步更新弹性并行强化学习微缩放微调思维链思维链微调性能优化性能监控性能评估性能边界情节记忆成本高效技术分享投机解码拒绝机制指令微调指令遵循指向定位推测解码推理推理优化推理加速推理增强推理局限性推理时扩展推理时计算推理模型推理迁移提示优化提示优化分析提示工程提示框架提示模式提示重写教育内容合成数学数学推理数据中心部署数据质量文本生成文档，最佳实践无参考优化无标注学习无监督学习无限上下文显存优化智能体查询中心树搜索校准过程样本效率样本选择梯度优化梯度冲突梯度检查点梯度重要性检索增强生成模块化框架模型压缩模型对齐模型架构步骤级监督注意力优化流水线并行测试时优化测试时扩展测试时计算混合专家模型混合推理混合精度训练激活平滑熵控制理论分析用户建模用户意图对齐直接偏好优化知识图谱知识密集型任务知识蒸馏硬件加速示例选择离线学习稀疏注意力端到端优化符号提取策略优化粒度优化系统架构线性校正结构化提示结果奖励模型综述论文网络可塑性网页搜索，动态过滤能力标签体系脚手架自主检索自优化自动化优化自动提示优化自动提示工程自动缩放自我批评自然语言理解自监督学习自适应学习自适应推理自适应计算行为优化视觉-语言对齐视觉问答视角理解计数任务计算scaling 计算效率认知架构认知模型训练免费训练加速训练可视化训练后量化记忆更新记忆检索记忆管理记忆系统记忆膨胀记忆质量设计空间探索评估基准语义压缩语义块语言建模负样本学习负载均衡转置优化软件模式轻量级模型边缘设备过程奖励模型过程监督连续空间推理迭代优化配对比较采样策略重置策略量化训练金融服务金融服务，插件错误分析长上下文长上下文推理长上下文训练长序列推理长文本推理长文本理解长期记忆长期运行问答预训练

归档

© 2026 Generative AI Discovery All Rights Reserved.

Theme by hiero