<?xml version="1.0" encoding="UTF-8"?><?xml-stylesheet href="/rss/styles.xsl" type="text/xsl"?><rss version="2.0"><channel><title>HelloAI · 系统化学 AI</title><description>中文 AI 学习与资讯门户 · 学习路径 + 论文精读 + 最新资讯</description><link>https://ai.xwebgame.com/</link><language>zh-cn</language><item><title>监控与可观测性：LLM 应用的生产管理</title><link>https://ai.xwebgame.com/learn/l7-07-monitoring/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l7-07-monitoring/</guid><description>一个 LLM 产品上线后，怎么知道它&quot;行不行&quot;？慢了、错了、贵了、被攻击了？这一篇讲监控栈。</description><pubDate>Mon, 31 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L7</category><category>监控</category><category>Observability</category><category>LLMOps</category><category>L7</category></item><item><title>AI 安全研究入门：怎么进入这个方向</title><link>https://ai.xwebgame.com/learn/l6-07-safety-research/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l6-07-safety-research/</guid><description>想做 AI 安全研究？这一篇讲方向、机构、起步项目、推荐阅读——这是 2026 年最稀缺的人才方向之一。</description><pubDate>Sun, 30 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L6</category><category>AI 安全</category><category>研究入门</category><category>职业</category><category>L6</category></item><item><title>视频生成：从 Sora 到现代视频 AI</title><link>https://ai.xwebgame.com/learn/l5-06-video-generation/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l5-06-video-generation/</guid><description>Sora、Runway Gen-3、Veo、Kling……2024-2026 视频生成爆发。这一篇讲技术原理 + 工程细节 + 商业格局。</description><pubDate>Sat, 29 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L5</category><category>视频生成</category><category>Sora</category><category>Diffusion</category><category>L5</category></item><item><title>📑 Scaling Laws for Neural Language Models</title><link>https://ai.xwebgame.com/papers/scaling-laws/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/scaling-laws/</guid><description>OpenAI 2020 年的奠基性发现——&quot;模型损失随参数、数据、算力呈幂律下降&quot;。这条曲线是 GPT-3、GPT-4 等大模型投资的理论基础。</description><pubDate>Fri, 28 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Scaling Laws</category><category>理论</category><category>OpenAI</category><category>必读</category></item><item><title>📑 LoRA: Low-Rank Adaptation of Large Language Models</title><link>https://ai.xwebgame.com/papers/lora/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/lora/</guid><description>Microsoft 提出 LoRA—只训 0.01% 参数 + 不损失性能 = 让&quot;消费级 GPU 微调大模型&quot;成为可能。开源 LLM 微调生态的关键技术。</description><pubDate>Thu, 27 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>LoRA</category><category>微调</category><category>PEFT</category><category>必读</category></item><item><title>📑 Training Compute-Optimal Large Language Models (Chinchilla)</title><link>https://ai.xwebgame.com/papers/chinchilla/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/chinchilla/</guid><description>DeepMind 证明 GPT-3 等大模型&quot;参数太多、数据太少&quot;。给出了&quot;算力如何在参数和数据间最优分配&quot;的新法则——重塑了大模型训练。</description><pubDate>Wed, 26 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Chinchilla</category><category>Scaling Laws</category><category>训练</category><category>必读</category></item><item><title>📑 Direct Preference Optimization (DPO)</title><link>https://ai.xwebgame.com/papers/dpo/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/dpo/</guid><description>把 RLHF 简化成一个简单的损失函数——跳过奖励模型和 PPO，效果接近，工程简单 10 倍。开源 LLM 对齐的事实标准。</description><pubDate>Tue, 25 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>DPO</category><category>RLHF</category><category>对齐</category><category>必读</category></item><item><title>LLM 成本优化：从 prompt 到部署的 10× 省钱方法</title><link>https://ai.xwebgame.com/learn/l4-12-cost-optimization/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-12-cost-optimization/</guid><description>LLM 调用又贵又快——一不小心就把 AWS 账单炸了。这一篇讲实战中的成本管控技巧。</description><pubDate>Mon, 24 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>成本优化</category><category>LLM 工程</category><category>L4</category></item><item><title>MCP 工具开发：手把手做一个自己的 MCP server</title><link>https://ai.xwebgame.com/learn/l4-11-mcp-development/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-11-mcp-development/</guid><description>L4-07 讲了 MCP 是什么。这一篇手把手教你写一个能让 Claude / Cursor 直接用的 MCP server。</description><pubDate>Sun, 23 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>MCP</category><category>工程实战</category><category>L4</category></item><item><title>Multi-Agent 系统：让 AI 团队协作</title><link>https://ai.xwebgame.com/learn/l4-10-multi-agent/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-10-multi-agent/</guid><description>一个 Agent 够强了吗？很多场景下不够——需要多个 Agent 分工。AutoGen、CrewAI 等框架开启了&quot;AI 团队&quot;时代。</description><pubDate>Sat, 22 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>Multi-Agent</category><category>AutoGen</category><category>L4</category></item><item><title>📑 The Pile: An 800GB Dataset of Diverse Text for Language Modeling</title><link>https://ai.xwebgame.com/papers/the-pile/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/the-pile/</guid><description>EleutherAI 开源的 800GB 训练数据集——第一个真正可用的&quot;GPT-3 级别&quot;开源训练数据。开源 LLM 革命的&quot;砖头&quot;。</description><pubDate>Fri, 21 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>The Pile</category><category>数据集</category><category>开源</category><category>基础</category></item><item><title>📑 Visual Instruction Tuning (LLaVA)</title><link>https://ai.xwebgame.com/papers/llava/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/llava/</guid><description>把 CLIP + LLaMA + 指令微调 缝合起来——开源多模态指令模型的起点。让&quot;图像+对话&quot;AI 进入开源社区。</description><pubDate>Thu, 20 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>LLaVA</category><category>多模态</category><category>开源</category><category>指令微调</category></item><item><title>📑 Transformers are SSMs: Generalized Models and Efficient Algorithms (Mamba 2)</title><link>https://ai.xwebgame.com/papers/mamba-2/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/mamba-2/</guid><description>Mamba 团队的反击——证明 Transformer 和 SSM 在数学上等价，并提出更快的 Mamba 2 架构。SSM 路线的关键升级。</description><pubDate>Wed, 19 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Mamba</category><category>SSM</category><category>架构</category><category>前沿</category></item><item><title>📑 TruthfulQA: Measuring How Models Mimic Human Falsehoods</title><link>https://ai.xwebgame.com/papers/truthfulqa/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/truthfulqa/</guid><description>一个测 LLM &quot;是否真实&quot;的 benchmark。第一次系统揭示：模型越大，反而在某些常见误区上越错。</description><pubDate>Tue, 18 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>TruthfulQA</category><category>评估</category><category>Benchmark</category><category>幻觉</category></item><item><title>训练优化进阶：让大模型训得动</title><link>https://ai.xwebgame.com/learn/l7-06-training-optimization/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l7-06-training-optimization/</guid><description>梯度检查点 / 混合精度 / Activation Recomputation / ZeRO Offload——这些工程技巧让 70B 模型在单卡上能微调。</description><pubDate>Mon, 17 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L7</category><category>训练优化</category><category>Mixed Precision</category><category>L7</category></item><item><title>AI 政策与监管：EU AI Act / 美国 EO / 中国办法对照</title><link>https://ai.xwebgame.com/learn/l6-06-policy/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l6-06-policy/</guid><description>AI 公司和应用面临的法律义务正在快速演化。这一篇给你全球三大法域的对照表 + 影响。</description><pubDate>Sun, 16 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L6</category><category>政策</category><category>监管</category><category>EU AI Act</category><category>AI Safety</category><category>L6</category></item><item><title>TTS 语音合成：从拼接到神经合成</title><link>https://ai.xwebgame.com/learn/l5-05-tts/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l5-05-tts/</guid><description>从机器人腔到逼真人声——TTS 走过的路。VALL-E 让 3 秒声音克隆任何人——这个能力的两面性。</description><pubDate>Sat, 15 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L5</category><category>TTS</category><category>语音合成</category><category>L5</category></item><item><title>📑 Gemini: A Family of Highly Capable Multimodal Models</title><link>https://ai.xwebgame.com/papers/gemini/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/gemini/</guid><description>Google 用 6 年时间 + 1 万张 TPU 训出的&quot;原生多模态&quot;大模型。1M+ 上下文窗口，是 GPT-4 的最大挑战者之一。</description><pubDate>Fri, 14 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Gemini</category><category>Google</category><category>多模态</category><category>大模型</category></item><item><title>📑 Segment Anything (SAM)</title><link>https://ai.xwebgame.com/papers/sam/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/sam/</guid><description>Meta 的&quot;图像分割基础模型&quot;——点一下就能分割任何物体。开源 + 1100 万张图 + 1 亿 mask，让&quot;通用分割&quot;成为现实。</description><pubDate>Thu, 13 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>SAM</category><category>分割</category><category>视觉</category><category>基础模型</category><category>必读</category></item><item><title>📑 Robust Speech Recognition via Large-Scale Weak Supervision (Whisper)</title><link>https://ai.xwebgame.com/papers/whisper/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/whisper/</guid><description>OpenAI 用 68 万小时弱监督音频训出最强 ASR。开源后统治整个开源语音识别市场。99 种语言通吃。</description><pubDate>Wed, 12 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Whisper</category><category>ASR</category><category>语音</category><category>开源</category><category>必读</category></item><item><title>📑 DeepSeek-V3 / R1：开源推理模型的革命</title><link>https://ai.xwebgame.com/papers/deepseek/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/deepseek/</guid><description>DeepSeek 用 $5.6M 训出接近 GPT-4 的开源模型——震动了整个行业。证明&quot;开源 + 高效工程 + 创新算法&quot; 能挑战美国巨头。</description><pubDate>Tue, 11 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>DeepSeek</category><category>开源</category><category>推理</category><category>前沿</category><category>必读</category></item><item><title>Tool Use 工程实战：让 LLM 真正会用工具</title><link>https://ai.xwebgame.com/learn/l4-09-tool-use/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-09-tool-use/</guid><description>LLM 能用工具——但要让它&quot;用得稳、用得对、用得省&quot;是另一门工程艺术。这一篇讲实战。</description><pubDate>Mon, 10 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>Tool Use</category><category>Function Calling</category><category>L4</category></item><item><title>LLM 评估：从 MMLU 到真实业务</title><link>https://ai.xwebgame.com/learn/l4-08-llm-evaluation/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-08-llm-evaluation/</guid><description>怎么知道你的 LLM 应用&quot;好不好&quot;？这一篇讲学术 benchmark + 工程评估的完整工具栈。</description><pubDate>Sun, 09 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>评估</category><category>Evaluation</category><category>Benchmark</category><category>L4</category></item><item><title>MCP 协议详解：AI 世界的 USB 接口</title><link>https://ai.xwebgame.com/learn/l4-07-mcp/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-07-mcp/</guid><description>Anthropic 2024 提出的 MCP——让任何工具能接入任何 LLM。一年时间成为 AI Agent 生态的事实标准。</description><pubDate>Sat, 08 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>MCP</category><category>Tool Use</category><category>Agent</category><category>L4</category></item><item><title>📑 Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone</title><link>https://ai.xwebgame.com/papers/phi-3/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/phi-3/</guid><description>Phi-3 mini 仅 3.8B 参数——但在多项 benchmark 上接近 GPT-3.5。证明了&quot;小模型 + 极致数据质量&quot;是另一条路。</description><pubDate>Fri, 07 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Phi-3</category><category>小模型</category><category>数据质量</category><category>前沿</category></item><item><title>📑 The Llama 3 Herd of Models</title><link>https://ai.xwebgame.com/papers/llama-3/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/llama-3/</guid><description>Meta 公开了 Llama 3 405B 的完整训练细节——开源模型首次达到 GPT-4 级别。92 页技术报告揭秘大模型训练的工程实战。</description><pubDate>Thu, 06 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Llama</category><category>开源</category><category>大模型</category><category>必读</category></item><item><title>📑 Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training</title><link>https://ai.xwebgame.com/papers/sleeper-agents/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/sleeper-agents/</guid><description>Anthropic 证明：可以训练一个&quot;装好的&quot;AI——表面对齐，遇到特定触发词激活恶意行为。而且当前所有对齐方法都检测不出来。</description><pubDate>Wed, 05 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Sleeper Agents</category><category>对齐</category><category>AI 安全</category><category>警告</category></item><item><title>📑 Learning to Reason with LLMs (OpenAI o1)</title><link>https://ai.xwebgame.com/papers/openai-o1/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/openai-o1/</guid><description>推理时计算的范式转变——让 LLM 在回答前花更多时间&quot;思考&quot;，复杂问题准确率从 20% 升到 80%。开启了&quot;推理模型&quot;时代。</description><pubDate>Tue, 04 Aug 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>o1</category><category>Reasoning</category><category>CoT</category><category>前沿</category><category>必读</category></item><item><title>模型部署与服务化：从训完到上线</title><link>https://ai.xwebgame.com/learn/l7-05-deployment/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l7-05-deployment/</guid><description>训出来一个模型只是开始。怎么把它变成一个 24/7 稳定、便宜、可扩展的服务？这一篇讲生产工程。</description><pubDate>Mon, 03 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L7</category><category>部署</category><category>MLOps</category><category>推理服务</category><category>L7</category></item><item><title>Whisper：让 AI 听懂 99 种语言</title><link>https://ai.xwebgame.com/learn/l5-04-whisper/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l5-04-whisper/</guid><description>OpenAI 开源的 Whisper 是当下最强语音识别。手机录音转文字、会议纪要、字幕生成——背后几乎都是它。</description><pubDate>Sun, 02 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L5</category><category>Whisper</category><category>ASR</category><category>语音识别</category><category>L5</category></item><item><title>In-Context Learning：为什么 LLM 看几个例子就会</title><link>https://ai.xwebgame.com/learn/l4-06-in-context-learning/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-06-in-context-learning/</guid><description>GPT-3 让所有人惊讶的能力——不微调，只给几个例子就能学。它是怎么工作的？这一篇用研究解读。</description><pubDate>Sat, 01 Aug 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>ICL</category><category>Few-shot</category><category>L4</category></item><item><title>📑 Video generation models as world simulators (Sora)</title><link>https://ai.xwebgame.com/papers/sora/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/sora/</guid><description>OpenAI 的视频生成模型 Sora——把视频切成&quot;时空 patch&quot;用 Transformer 做扩散。1 分钟高质量视频成为可能，&quot;AI 世界模拟器&quot;露端倪。</description><pubDate>Fri, 31 Jul 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Sora</category><category>视频生成</category><category>Diffusion</category><category>Transformer</category><category>前沿</category></item><item><title>📑 FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness</title><link>https://ai.xwebgame.com/papers/flash-attention/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/flash-attention/</guid><description>通过感知 GPU 内存层级，让注意力计算快 2-4 倍 + 显存少 10 倍——而且数学上完全相同。所有现代 LLM 都用它。</description><pubDate>Thu, 30 Jul 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>FlashAttention</category><category>GPU</category><category>系统优化</category><category>必读</category></item><item><title>📑 Constitutional AI: Harmlessness from AI Feedback</title><link>https://ai.xwebgame.com/papers/constitutional-ai/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/constitutional-ai/</guid><description>Anthropic 提出的对齐新方法——让 AI 用&quot;宪法原则&quot;自评自改，跳过大量人类标注。Claude 的核心训练秘密。</description><pubDate>Wed, 29 Jul 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Constitutional AI</category><category>对齐</category><category>Anthropic</category><category>必读</category></item><item><title>📑 Mamba: Linear-Time Sequence Modeling with Selective State Spaces</title><link>https://ai.xwebgame.com/papers/mamba/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/mamba/</guid><description>挑战 Transformer 霸权的&quot;选择性状态空间模型&quot;——线性复杂度处理超长序列，理论上能取代 Transformer。2024 年最热的架构研究之一。</description><pubDate>Tue, 28 Jul 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Mamba</category><category>SSM</category><category>架构</category><category>前沿</category></item><item><title>量化深度解析：GPTQ / AWQ / FP8 / GGUF 全谱</title><link>https://ai.xwebgame.com/learn/l7-04-quantization/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l7-04-quantization/</guid><description>让 70B 模型塞进 24GB 显存——量化是消费级硬件跑大模型的关键。这一篇详解各家方案。</description><pubDate>Mon, 27 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L7</category><category>量化</category><category>GPTQ</category><category>AWQ</category><category>FP8</category><category>L7</category></item><item><title>偏见与公平：AI 学到的不止是规则，还有人类的&quot;暗面&quot;</title><link>https://ai.xwebgame.com/learn/l6-05-bias-fairness/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l6-05-bias-fairness/</guid><description>训练数据是人类社会的镜像——AI 学到的&quot;模式&quot;包含了所有偏见、刻板印象、不公平。这一篇直面这个问题。</description><pubDate>Sun, 26 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L6</category><category>偏见</category><category>公平</category><category>Fairness</category><category>L6</category></item><item><title>Agent 构建详解：让 LLM 自己干活</title><link>https://ai.xwebgame.com/learn/l4-04-agent/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-04-agent/</guid><description>从&quot;问一句答一句的对话&quot;到&quot;能连续工作 4 小时的同事&quot;——Agent 是 LLM 应用工程的下一步。</description><pubDate>Sat, 25 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>Agent</category><category>Tool Use</category><category>MCP</category><category>L4</category></item><item><title>📑 Training language models to follow instructions with human feedback (InstructGPT)</title><link>https://ai.xwebgame.com/papers/instruct-gpt/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/instruct-gpt/</guid><description>从 GPT-3 到 ChatGPT 的&quot;桥梁&quot;。提出 SFT + RLHF 三阶段训练让 LLM &quot;听话&quot;——这套流程定义了之后所有商业 LLM 的训练范式。</description><pubDate>Fri, 24 Jul 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>RLHF</category><category>InstructGPT</category><category>对齐</category><category>ChatGPT</category><category>必读</category></item><item><title>📑 Highly Accurate Protein Structure Prediction with AlphaFold</title><link>https://ai.xwebgame.com/papers/alphafold2/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/alphafold2/</guid><description>DeepMind 用 Transformer 解决了 50 年的&quot;蛋白质折叠&quot;问题。预测了所有已知生物的 2 亿个蛋白质结构。2024 年诺贝尔化学奖。</description><pubDate>Thu, 23 Jul 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>AlphaFold</category><category>AI for Science</category><category>蛋白质</category><category>诺贝尔奖</category><category>必读</category></item><item><title>📑 BERT: Pre-training of Deep Bidirectional Transformers</title><link>https://ai.xwebgame.com/papers/bert/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/bert/</guid><description>2018 年的 NLP 核爆。提出 Masked Language Modeling + 双向 Transformer，让&quot;预训练 + 微调&quot;成为 NLP 主流范式。</description><pubDate>Wed, 22 Jul 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>BERT</category><category>NLP</category><category>预训练</category><category>必读</category></item><item><title>📑 Learning Transferable Visual Models From Natural Language Supervision (CLIP)</title><link>https://ai.xwebgame.com/papers/clip/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/clip/</guid><description>用 4 亿张&quot;图 + 描述&quot;对训练——让图像 encoder 和文本 encoder 在同一向量空间对齐。从此 AI 能&quot;看图说话&quot;，&quot;看图作画&quot;。</description><pubDate>Tue, 21 Jul 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>CLIP</category><category>多模态</category><category>对比学习</category><category>必读</category></item><item><title>机制可解释性：看见神经元在想什么</title><link>https://ai.xwebgame.com/learn/l6-04-interpretability/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l6-04-interpretability/</guid><description>LLM 是黑盒——但研究者已经能从几百亿参数里&quot;读出&quot;具体的概念了。这一篇带你认识 AI 内部的&quot;心理学&quot;研究。</description><pubDate>Mon, 20 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L6</category><category>可解释性</category><category>Interpretability</category><category>L6</category></item><item><title>ViT 与 CLIP：让 Transformer 看图</title><link>https://ai.xwebgame.com/learn/l5-03-vit-clip/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l5-03-vit-clip/</guid><description>把图像切成 patch，喂给 Transformer——视觉领域 2020 年最大的范式转变。</description><pubDate>Sun, 19 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L5</category><category>ViT</category><category>CLIP</category><category>视觉</category><category>L5</category></item><item><title>Prompt 进阶技巧：CoT / Self-Consistency / Tree of Thoughts / Reflexion</title><link>https://ai.xwebgame.com/learn/l4-02-advanced-prompting/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-02-advanced-prompting/</guid><description>L0-05 教你 10 个基础招。这一篇讲学术研究里的&quot;硬核&quot; prompt 工程——能让 GPT-4 在数学题上从 50% 升到 85%。</description><pubDate>Sat, 18 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>Prompt</category><category>CoT</category><category>L4</category></item><item><title>📑 Attention Is All You Need</title><link>https://ai.xwebgame.com/papers/attention-is-all-you-need/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/attention-is-all-you-need/</guid><description>提出 Transformer 架构——完全抛弃 RNN，只用注意力机制。这篇 8 页的论文催生了今天所有大模型。被引 12 万+。</description><pubDate>Fri, 17 Jul 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Transformer</category><category>Attention</category><category>NLP</category><category>必读</category></item><item><title>📑 Denoising Diffusion Probabilistic Models (DDPM)</title><link>https://ai.xwebgame.com/papers/ddpm/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/ddpm/</guid><description>提出 DDPM —— 用&quot;加噪 → 去噪&quot;的范式做图像生成。Stable Diffusion、Sora 都基于这个思路。</description><pubDate>Fri, 17 Jul 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>Diffusion</category><category>生成模型</category><category>图像</category><category>必读</category></item><item><title>📑 Language Models are Few-Shot Learners (GPT-3)</title><link>https://ai.xwebgame.com/papers/gpt-3/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/gpt-3/</guid><description>175B 参数的 GPT-3 展示了&quot;in-context learning&quot;——不微调，只给几个例子就能学会新任务。这篇论文重新定义了人们对 LLM 的预期。</description><pubDate>Fri, 17 Jul 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>GPT-3</category><category>LLM</category><category>Few-shot</category><category>必读</category></item><item><title>📑 Deep Residual Learning for Image Recognition</title><link>https://ai.xwebgame.com/papers/resnet/</link><guid isPermaLink="true">https://ai.xwebgame.com/papers/resnet/</guid><description>提出残差连接（skip connection），让神经网络能训到 100+ 层。CVPR 2016 最佳论文，引用 25 万+，至今所有大模型仍在用这个技巧。</description><pubDate>Fri, 17 Jul 2026 00:00:00 GMT</pubDate><category>论文精读</category><category>CNN</category><category>ResNet</category><category>残差连接</category><category>视觉</category><category>必读</category></item><item><title>推理优化：vLLM / 量化 / 投机解码 / KV Cache</title><link>https://ai.xwebgame.com/learn/l7-03-inference/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l7-03-inference/</guid><description>训完模型只是开始。让 LLM 在生产环境跑快、跑省、跑稳，是另一套工程艺术。</description><pubDate>Thu, 16 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L7</category><category>推理</category><category>vLLM</category><category>量化</category><category>L7</category></item><item><title>分布式训练：DP / DDP / FSDP / Tensor Parallel 怎么选</title><link>https://ai.xwebgame.com/learn/l7-02-distributed/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l7-02-distributed/</guid><description>一张 H100 装不下 70B 模型。怎么把训练任务分给几千张卡？这一篇梳理 4 种主流并行策略。</description><pubDate>Wed, 15 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L7</category><category>分布式训练</category><category>FSDP</category><category>DDP</category><category>L7</category></item><item><title>GPU 速览：为什么 AI 离不开它</title><link>https://ai.xwebgame.com/learn/l7-01-gpu/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l7-01-gpu/</guid><description>A100、H100、B200 是什么？为啥 AI 用 GPU 不用 CPU？这一篇打开硬件的黑盒。</description><pubDate>Tue, 14 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L7</category><category>GPU</category><category>CUDA</category><category>硬件</category><category>L7</category></item><item><title>红队与越狱：攻击 LLM 的方法与防御</title><link>https://ai.xwebgame.com/learn/l6-03-red-team/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l6-03-red-team/</guid><description>AI 安全的&quot;反派&quot;视角——专门找模型漏洞的人在做什么。理解攻击才能防御。</description><pubDate>Mon, 13 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L6</category><category>Red Team</category><category>越狱</category><category>Jailbreak</category><category>L6</category></item><item><title>RLHF 与 Constitutional AI：两大对齐方法对比</title><link>https://ai.xwebgame.com/learn/l6-02-rlhf-cai/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l6-02-rlhf-cai/</guid><description>OpenAI 用 RLHF，Anthropic 用 CAI。它们都让 LLM &quot;听话&quot;，但思路完全不同。</description><pubDate>Sun, 12 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L6</category><category>RLHF</category><category>Constitutional AI</category><category>对齐</category><category>L6</category></item><item><title>为什么需要 AI 对齐：从对错到价值观</title><link>https://ai.xwebgame.com/learn/l6-01-why-alignment/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l6-01-why-alignment/</guid><description>能力越强的 AI，对齐越关键。这一篇讲清楚&quot;对齐&quot;到底是什么，为什么它是 AI 时代最重要的研究方向之一。</description><pubDate>Sat, 11 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L6</category><category>对齐</category><category>AI 安全</category><category>L6</category></item><item><title>BERT vs GPT：两大 LLM 流派的本质区别</title><link>https://ai.xwebgame.com/learn/l3-09-bert-vs-gpt/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l3-09-bert-vs-gpt/</guid><description>都是 Transformer，为啥一个做&quot;理解&quot;一个做&quot;生成&quot;？两者的训练目标差一个根本性的设计选择。</description><pubDate>Fri, 10 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L3</category><category>BERT</category><category>GPT</category><category>Transformer</category><category>L3</category></item><item><title>SVM 支持向量机：经典 ML 里的几何派</title><link>https://ai.xwebgame.com/learn/l2-08-svm/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l2-08-svm/</guid><description>深度学习火之前，SVM 统治了 2000-2010 整整一个时代。今天它仍然是小数据集和文本分类的最优选之一。</description><pubDate>Thu, 09 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L2</category><category>SVM</category><category>分类</category><category>L2</category></item><item><title>Diffusion 数学：从加噪到生成</title><link>https://ai.xwebgame.com/learn/l5-02-diffusion-math/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l5-02-diffusion-math/</guid><description>Stable Diffusion、DALL·E 3、Sora 都基于扩散模型。这一篇讲清楚它的核心数学——用最少的公式。</description><pubDate>Wed, 08 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L5</category><category>Diffusion</category><category>Stable Diffusion</category><category>生成模型</category><category>L5</category></item><item><title>多模态总览：AI 如何同时&quot;看、听、读&quot;</title><link>https://ai.xwebgame.com/learn/l5-01-multimodal-overview/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l5-01-multimodal-overview/</guid><description>GPT-4o 能识别你画的草图、Sora 能生成视频——这些&quot;多模态&quot;AI 是怎么做到的？这一篇打开全景图。</description><pubDate>Tue, 07 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L5</category><category>多模态</category><category>CLIP</category><category>视觉</category><category>语音</category><category>L5</category></item><item><title>LoRA 微调入门：让大模型&quot;特化&quot;成你需要的样子</title><link>https://ai.xwebgame.com/learn/l4-05-lora/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-05-lora/</guid><description>不用重训整个 70B 模型，只调几百万个参数，一张消费级显卡就能微调 LLM。</description><pubDate>Mon, 06 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>LoRA</category><category>微调</category><category>Fine-tuning</category><category>L4</category></item><item><title>RAG 从 0 到 1：让 LLM 基于你的数据回答</title><link>https://ai.xwebgame.com/learn/l4-03-rag/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-03-rag/</guid><description>企业 AI 应用 90% 都是 RAG。这一篇带你搭一个能跑的 RAG 系统——从分块到部署。</description><pubDate>Sun, 05 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>RAG</category><category>检索</category><category>LLM</category><category>工程实战</category><category>L4</category></item><item><title>LLM 是怎么炼成的：Pretrain → SFT → RLHF 全流程</title><link>https://ai.xwebgame.com/learn/l4-01-llm-training/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-01-llm-training/</guid><description>&quot;训练一个 ChatGPT&quot; 不是一步，是三步。每一步用完全不同的数据和目标。看完你能跟工程师对话。</description><pubDate>Sat, 04 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>LLM</category><category>Pretrain</category><category>RLHF</category><category>L4</category></item><item><title>完整 Transformer 架构：把所有积木组合起来</title><link>https://ai.xwebgame.com/learn/l3-08-transformer-architecture/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l3-08-transformer-architecture/</guid><description>注意力是核心，但 Transformer 还有位置编码、FFN、残差连接、LayerNorm。这一篇把它们组合成一个完整可跑的模型。</description><pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L3</category><category>Transformer</category><category>BERT</category><category>GPT</category><category>L3</category></item><item><title>RNN / LSTM 兴衰：序列模型简史</title><link>https://ai.xwebgame.com/learn/l3-04-rnn/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l3-04-rnn/</guid><description>2017 年 Transformer 出现前，所有处理语言/语音的 AI 都靠 RNN 和它的衍生品。理解它的兴衰，才理解为什么 attention 是革命。</description><pubDate>Thu, 02 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L3</category><category>RNN</category><category>LSTM</category><category>序列建模</category><category>L3</category></item><item><title>从感知机到多层神经网络：深度学习的起点</title><link>https://ai.xwebgame.com/learn/l3-01-perceptron-to-mlp/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l3-01-perceptron-to-mlp/</guid><description>所有神经网络都是从一个 1958 年的简单模型扩展来的。这一篇讲清楚&quot;神经元&quot;到底是个啥。</description><pubDate>Wed, 01 Jul 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L3</category><category>感知机</category><category>神经网络</category><category>激活函数</category><category>L3</category></item><item><title>评估指标 + 过拟合 + 正则化：让模型不犯傻的工程实战</title><link>https://ai.xwebgame.com/learn/l2-07-evaluation/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l2-07-evaluation/</guid><description>会训模型不算啥。会判断模型&quot;行不行&quot;、知道为啥不行、怎么修——才是真正的 ML 工程师。</description><pubDate>Tue, 30 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L2</category><category>评估</category><category>过拟合</category><category>正则化</category><category>交叉验证</category><category>L2</category></item><item><title>K-Means 聚类：最经典的无监督算法</title><link>https://ai.xwebgame.com/learn/l2-06-kmeans/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l2-06-kmeans/</guid><description>没标签也能学。K-Means 是用户分群、图像压缩、市场分析的瑞士军刀——20 行代码就能写。</description><pubDate>Mon, 29 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L2</category><category>K-Means</category><category>聚类</category><category>无监督</category><category>L2</category></item><item><title>随机森林 + Boosting：让一群弱学习器变成超人</title><link>https://ai.xwebgame.com/learn/l2-05-random-forest/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l2-05-random-forest/</guid><description>单棵树平庸，一群树投票就能逆天。Kaggle 十年霸主 XGBoost 的来历，就在这一篇。</description><pubDate>Sun, 28 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L2</category><category>随机森林</category><category>XGBoost</category><category>集成学习</category><category>L2</category></item><item><title>决策树：最直观可解释的 ML 算法</title><link>https://ai.xwebgame.com/learn/l2-04-decision-trees/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l2-04-decision-trees/</guid><description>一连串&quot;是不是&quot;问题，组成一棵决策的树。它是 XGBoost、LightGBM 等竞赛王者算法的基石。</description><pubDate>Sat, 27 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L2</category><category>决策树</category><category>分类</category><category>L2</category><category>可解释 ML</category></item><item><title>逻辑回归与分类：从回归到决策</title><link>https://ai.xwebgame.com/learn/l2-03-logistic-regression/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l2-03-logistic-regression/</guid><description>名字叫&quot;回归&quot;其实是分类。在神经网络出来之前，逻辑回归是工业界分类的事实标准——今天仍然是 baseline。</description><pubDate>Fri, 26 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L2</category><category>逻辑回归</category><category>分类</category><category>L2</category><category>Sigmoid</category></item><item><title>线性回归：最简单也最深刻的 ML 模型</title><link>https://ai.xwebgame.com/learn/l2-02-linear-regression/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l2-02-linear-regression/</guid><description>你以为线性回归很简单？神经网络的最后一层、几乎所有 ML 的起点——都是它。</description><pubDate>Thu, 25 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L2</category><category>线性回归</category><category>监督学习</category><category>L2</category></item><item><title>监督 / 无监督 / 强化：机器学习的三大世界观</title><link>https://ai.xwebgame.com/learn/l2-01-ml-paradigms/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l2-01-ml-paradigms/</guid><description>所有机器学习算法都属于这三类之一。理解这个分类，你立刻能给任何算法&quot;找它的家&quot;。</description><pubDate>Wed, 24 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L2</category><category>ML</category><category>监督学习</category><category>无监督学习</category><category>强化学习</category><category>L2</category></item><item><title>PyTorch 基础：张量、自动求导、你的第一个神经网络</title><link>https://ai.xwebgame.com/learn/l1-10-pytorch/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l1-10-pytorch/</guid><description>把 L1 全部的数学和 Python 武器组合起来——这是你的&quot;AI 工程师&quot;入门仪式。</description><pubDate>Tue, 23 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L1</category><category>PyTorch</category><category>神经网络</category><category>L1</category><category>深度学习</category></item><item><title>Pandas 数据处理：DataFrame 是 ML 数据的标准载体</title><link>https://ai.xwebgame.com/learn/l1-09-pandas/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l1-09-pandas/</guid><description>看到 CSV、Excel 这些&quot;表格数据&quot;，先用 Pandas 装起来再说。这是 ML 工程师每天用的工具。</description><pubDate>Mon, 22 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L1</category><category>Pandas</category><category>数据处理</category><category>L1</category><category>工具</category></item><item><title>NumPy 数组运算：让 Python 像 MATLAB 一样会做矩阵</title><link>https://ai.xwebgame.com/learn/l1-08-numpy/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l1-08-numpy/</guid><description>NumPy 是所有 ML 代码的地基。这一篇让你的&quot;循环式思维&quot;升级为&quot;向量化思维&quot;——快 100 倍的那种。</description><pubDate>Sun, 21 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L1</category><category>NumPy</category><category>向量化</category><category>L1</category><category>工具</category></item><item><title>Python 速成（一）：30 分钟从零到能写函数</title><link>https://ai.xwebgame.com/learn/l1-07-python-basics/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l1-07-python-basics/</guid><description>只讲 ML 用得到的 Python 子集——不教 OOP，不教装饰器，不教 generator。够用就好。</description><pubDate>Sat, 20 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L1</category><category>Python</category><category>入门</category><category>L1</category><category>编程基础</category></item><item><title>链式法则与反向传播：手算一个 3 层网络</title><link>https://ai.xwebgame.com/learn/l1-06-backprop/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l1-06-backprop/</guid><description>把 L1 前 5 篇的所有数学武器组合起来——让我们手算一遍真实神经网络的梯度。</description><pubDate>Fri, 19 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L1</category><category>反向传播</category><category>链式法则</category><category>L1</category><category>神经网络</category></item><item><title>信息论：熵、交叉熵、KL 散度的直觉</title><link>https://ai.xwebgame.com/learn/l1-05-information-theory/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l1-05-information-theory/</guid><description>损失函数为什么长 -log P 这样？这一篇从信息论角度回答。理解了，你看到任何损失函数都不再陌生。</description><pubDate>Thu, 18 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L1</category><category>信息论</category><category>熵</category><category>交叉熵</category><category>L1</category></item><item><title>Tokenizer 与 BPE：LLM 看到的不是字，是 token</title><link>https://ai.xwebgame.com/learn/l4-02-tokenizer-bpe/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l4-02-tokenizer-bpe/</guid><description>为什么 GPT 把&quot;strawberry&quot;切成 4 个 token？为什么数错&quot;strawberry 有几个 r&quot;？这些都和 tokenizer 有关。</description><pubDate>Wed, 17 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L4</category><category>Tokenizer</category><category>BPE</category><category>LLM</category><category>L4</category></item><item><title>CNN 卷积原理：从滤镜到 ResNet</title><link>https://ai.xwebgame.com/learn/l3-06-cnn/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l3-06-cnn/</guid><description>在 Transformer 称霸前，CNN 是计算机视觉的统治者。它今天仍然是处理图像的标配。这一篇讲清楚&quot;卷积&quot;到底是什么。</description><pubDate>Tue, 16 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L3</category><category>CNN</category><category>卷积</category><category>视觉</category><category>L3</category></item><item><title>注意力机制详解：从直觉到完整推导</title><link>https://ai.xwebgame.com/learn/l3-05-attention/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l3-05-attention/</guid><description>Attention is all you need. 这一篇带你从&quot;它到底在干嘛&quot;到&quot;它的每行公式&quot;，一次性吃透 Transformer 的核心。</description><pubDate>Mon, 15 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L3</category><category>Transformer</category><category>Attention</category><category>Q-K-V</category><category>L3</category><category>深度学习</category></item><item><title>优化器深度解析：SGD / Momentum / Adam 为什么是这样</title><link>https://ai.xwebgame.com/learn/l2-09-optimizers/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l2-09-optimizers/</guid><description>你以为梯度下降只是&quot;沿梯度反方向走&quot;——其实有 10 多种花式走法。Adam 凭什么成为深度学习默认款？</description><pubDate>Sun, 14 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L2</category><category>优化器</category><category>SGD</category><category>Adam</category><category>L2</category></item><item><title>概率与最大似然：为什么 ML 是个概率问题</title><link>https://ai.xwebgame.com/learn/l1-04-probability/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l1-04-probability/</guid><description>所有 ML 模型其实都在干同一件事：寻找让&quot;看到这些数据&quot;的概率最大的参数。这个观点一旦你 get 了，整个 ML 都通透了。</description><pubDate>Sat, 13 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L1</category><category>概率</category><category>最大似然</category><category>L1</category><category>统计</category></item><item><title>导数与梯度：&quot;学习&quot;的数学定义</title><link>https://ai.xwebgame.com/learn/l1-03-derivatives/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l1-03-derivatives/</guid><description>&quot;训练神经网络&quot; 的本质就是导数。把这个词搞懂，AI 的训练流程在你眼里就再不神秘了。</description><pubDate>Fri, 12 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L1</category><category>微积分</category><category>导数</category><category>梯度</category><category>L1</category></item><item><title>线性代数：用图片和位置讲明白向量和矩阵</title><link>https://ai.xwebgame.com/learn/l1-02-linear-algebra/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l1-02-linear-algebra/</guid><description>所有 AI 的本质都是矩阵运算。这一篇不讲行列式不讲特征值，只让你&quot;看见&quot;向量在干什么。</description><pubDate>Thu, 11 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L1</category><category>线性代数</category><category>向量</category><category>矩阵</category><category>L1</category></item><item><title>数学不发愁：这本来该是数学课，但我们换个学法</title><link>https://ai.xwebgame.com/learn/l1-01-math-manifesto/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l1-01-math-manifesto/</guid><description>被高中数学伤过没关系。我们用图、动画、Python 代码把 AI 数学讲清楚——不会的不要先怕，是教材写得不行而已。</description><pubDate>Wed, 10 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L1</category><category>数学</category><category>入门</category><category>L1</category><category>学习方法</category></item><item><title>L0 毕业了：下一步学什么？</title><link>https://ai.xwebgame.com/learn/l0-12-next-steps/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l0-12-next-steps/</guid><description>恭喜读完启蒙路径。这一篇帮你规划下一步——是继续往深里学，还是横着拓宽，还是直接动手做项目。</description><pubDate>Sun, 07 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L0</category><category>学习路径</category><category>入门毕业</category><category>规划</category></item><item><title>AI 词汇表：写给文科生的 30 个核心术语</title><link>https://ai.xwebgame.com/learn/l0-11-ai-glossary/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l0-11-ai-glossary/</guid><description>Token、Embedding、Transformer、RAG、Fine-tuning……这些每天在群里飞的黑话，一次性给你讲完。</description><pubDate>Sat, 06 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L0</category><category>术语</category><category>词汇表</category><category>入门</category></item><item><title>AI 安全：你和 AI 聊的话，去哪了？</title><link>https://ai.xwebgame.com/learn/l0-10-ai-data-safety/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l0-10-ai-data-safety/</guid><description>你贴进 ChatGPT 的合同、病历、产品机密——它们怎么处理？谁能看到？这些事知道了，你才能放心用。</description><pubDate>Fri, 05 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L0</category><category>AI 安全</category><category>隐私</category><category>数据</category><category>入门</category></item><item><title>ChatGPT / Claude / Gemini 怎么选？——一份不站队的横评</title><link>https://ai.xwebgame.com/learn/l0-09-llm-comparison/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l0-09-llm-comparison/</guid><description>不吹也不黑，把三大模型的优劣摆在桌面上。哪个最适合你，取决于你要干什么。</description><pubDate>Thu, 04 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L0</category><category>大模型对比</category><category>ChatGPT</category><category>Claude</category><category>Gemini</category><category>工具选择</category></item><item><title>AI 会取代我的工作吗？——一个分类框架</title><link>https://ai.xwebgame.com/learn/l0-08-jobs-and-ai/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l0-08-jobs-and-ai/</guid><description>不要问&quot;会不会取代你&quot;，问&quot;会取代你工作里的哪些部分&quot;。这一篇给你一个能套用一辈子的判断框架。</description><pubDate>Wed, 03 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L0</category><category>职业影响</category><category>入门</category><category>社会变革</category></item><item><title>5 分钟跑通你的第一个 AI 对话</title><link>https://ai.xwebgame.com/learn/l0-07-first-ai-chat/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l0-07-first-ai-chat/</guid><description>理论讲完了，上手最快。手把手注册、登录、第一句话怎么说。</description><pubDate>Tue, 02 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L0</category><category>入门</category><category>上手</category><category>ChatGPT</category><category>Claude</category></item><item><title>&quot;智能&quot;到底是什么？——从图灵测试到今天的争论</title><link>https://ai.xwebgame.com/learn/l0-06-what-is-intelligence/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l0-06-what-is-intelligence/</guid><description>一个我们以为很清楚、其实从来没说清楚过的词。当你下次纠结&quot;AI 算不算真的智能&quot;，先看完这篇。</description><pubDate>Mon, 01 Jun 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L0</category><category>哲学</category><category>图灵测试</category><category>入门</category></item><item><title>提示词入门 10 招：让 AI 干活立刻精准 10 倍</title><link>https://ai.xwebgame.com/learn/l0-05-prompt-basics/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l0-05-prompt-basics/</guid><description>不用学黑话，10 个动作就够。这一篇让你的 AI 使用水平直接超过 80% 的人。</description><pubDate>Sun, 31 May 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L0</category><category>Prompt</category><category>提示词</category><category>入门</category><category>实用技巧</category></item><item><title>&quot;幻觉&quot;是什么、为什么会发生、怎么减轻</title><link>https://ai.xwebgame.com/learn/l0-04-hallucination/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l0-04-hallucination/</guid><description>为什么 AI 会一本正经地胡说八道？这不是 bug，这是它本来就这样工作。理解了机制，你就知道怎么用。</description><pubDate>Sat, 30 May 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L0</category><category>幻觉</category><category>可靠性</category><category>入门</category><category>RAG</category></item><item><title>AI 现在到底能做什么、不能做什么——用真实例子说话</title><link>https://ai.xwebgame.com/learn/l0-03-what-can-ai-do/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l0-03-what-can-ai-do/</guid><description>不要再听吹牛和黑稿了，我们一条一条核对：哪些事 AI 已经做得比你好，哪些事它彻底做不了。</description><pubDate>Fri, 29 May 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L0</category><category>能力边界</category><category>入门</category><category>应用场景</category></item><item><title>一篇文章读完 AI 简史：从图灵到 ChatGPT</title><link>https://ai.xwebgame.com/learn/l0-02-ai-history/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l0-02-ai-history/</guid><description>70 年压成 15 分钟。不讲流水账，只挑改变了游戏规则的 11 个时刻。</description><pubDate>Thu, 28 May 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L0</category><category>AI 简史</category><category>入门</category><category>人物</category><category>里程碑</category></item><item><title>📰 Claude Opus 4.6 发布：上下文跃迁至 2M，编码超越 GPT-5</title><link>https://ai.xwebgame.com/news/2026-05-27-claude-opus-46/</link><guid isPermaLink="true">https://ai.xwebgame.com/news/2026-05-27-claude-opus-46/</guid><description>Anthropic 今晨发布 Opus 4.6，首次将 Claude 模型上下文窗口扩展至 2M token，SWE-bench 78.4%。本文带你看完技术亮点与定价变化。</description><pubDate>Wed, 27 May 2026 08:00:00 GMT</pubDate><category>资讯</category><category>Claude</category><category>Opus</category><category>上下文窗口</category><category>编码能力</category><category>Anthropic</category></item><item><title>AI、机器学习、深度学习、大模型，到底什么关系？</title><link>https://ai.xwebgame.com/learn/l0-01-ai-ml-dl-llm-relationship/</link><guid isPermaLink="true">https://ai.xwebgame.com/learn/l0-01-ai-ml-dl-llm-relationship/</guid><description>一篇文章拆清楚四个被搞混最多的词，从此你不会再被任何吹牛和黑话唬住。</description><pubDate>Wed, 27 May 2026 00:00:00 GMT</pubDate><category>学习路径</category><category>L0</category><category>入门</category><category>概念辨析</category><category>AI 基础</category></item><item><title>📰 Mistral 开源 8x22B-V2：MoE 效率提升 2.3 倍</title><link>https://ai.xwebgame.com/news/2026-05-26-mistral-8x22b-v2/</link><guid isPermaLink="true">https://ai.xwebgame.com/news/2026-05-26-mistral-8x22b-v2/</guid><description>Mistral 发布新版 8x22B-V2 模型，每 token 仅激活 39B 参数，推理速度比上代提升 2.3 倍，可在单张 H100 上跑。</description><pubDate>Tue, 26 May 2026 15:00:00 GMT</pubDate><category>资讯</category><category>Mistral</category><category>MoE</category><category>开源</category><category>推理优化</category></item><item><title>📰 DeepMind 提出 Titan-Mem 架构：把无穷记忆装进 Transformer</title><link>https://ai.xwebgame.com/news/2026-05-25-deepmind-titan-mem/</link><guid isPermaLink="true">https://ai.xwebgame.com/news/2026-05-25-deepmind-titan-mem/</guid><description>Google DeepMind 发布新架构 Titan-Mem，通过显式记忆模块在不重训的情况下持续学习，可视为对 RAG 的根本性替代。</description><pubDate>Mon, 25 May 2026 12:00:00 GMT</pubDate><category>资讯</category><category>DeepMind</category><category>Transformer</category><category>长上下文</category><category>记忆机制</category><category>持续学习</category></item><item><title>📰 欧盟 AI Act 实施细则落地：5 类高风险模型需备案</title><link>https://ai.xwebgame.com/news/2026-05-24-eu-ai-act-rollout/</link><guid isPermaLink="true">https://ai.xwebgame.com/news/2026-05-24-eu-ai-act-rollout/</guid><description>2026 年 7 月 1 日起，10B 参数以上模型在欧盟商用必须先备案，包括能力评估、训练数据来源说明、版权合规证明等。</description><pubDate>Sun, 24 May 2026 18:00:00 GMT</pubDate><category>资讯</category><category>EU AI Act</category><category>监管</category><category>合规</category><category>政策</category></item><item><title>📰 Cursor 4.0 发布：Agent 模式可连续自主工作 4 小时</title><link>https://ai.xwebgame.com/news/2026-05-23-cursor-4-released/</link><guid isPermaLink="true">https://ai.xwebgame.com/news/2026-05-23-cursor-4-released/</guid><description>Cursor 4.0 基于 Claude Opus 4.6 推出长任务 Agent 模式，能自主规划、执行、自我纠错，处理跨多个文件的复杂 refactor。</description><pubDate>Sat, 23 May 2026 20:00:00 GMT</pubDate><category>资讯</category><category>Cursor</category><category>Agent</category><category>编程工具</category><category>Claude</category></item></channel></rss>