<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>LIZI STACK — Notes</title><description>Notes on recommender systems, LLMs, agent engineering, experiments, and long-term thinking.</description><link>https://lizistack.cn/</link><item><title>Evaluating RAG Beyond Plausible Answers</title><link>https://lizistack.cn/en/notes/rag-evaluation-from-retrieval-to-answer/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/rag-evaluation-from-retrieval-to-answer/</guid><description>Decompose RAG failures across retrieval, context construction, and generation so each result points to an actionable change.</description><pubDate>Sun, 01 Nov 2026 00:00:00 GMT</pubDate></item><item><title>Agents Need Explainable Traces, Not Just More Logs</title><link>https://lizistack.cn/en/notes/agent-observability/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/agent-observability/</guid><description>Once an agent calls tools, changes state, and replans, conventional request logs are no longer enough to explain success or failure.</description><pubDate>Sat, 31 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Why Better Offline Ranking Does Not Guarantee Better Online Results</title><link>https://lizistack.cn/en/notes/ranking-model-offline-evaluation/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/ranking-model-offline-evaluation/</guid><description>Offline metrics are evidence, not a conclusion; data splits, exposure bias, and user segments determine whether that evidence is trustworthy.</description><pubDate>Fri, 30 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Why This Blog Exists: A Version History for Technical Growth</title><link>https://lizistack.cn/en/notes/why-write/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/why-write/</guid><description>I want a record of learning, engineering judgement, and long-term thinking that can be revisited and tested.</description><pubDate>Thu, 29 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Reading Transformer: Why Attention Could Replace Recurrence</title><link>https://lizistack.cn/en/notes/attention-is-all-you-need-reading/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/attention-is-all-you-need-reading/</guid><description>A close reading of scaled dot-product attention, multi-head attention, positional encoding, and the engineering conditions that made Transformer consequential.</description><pubDate>Wed, 28 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Write the Learning Process, Not Only the Model Conclusion</title><link>https://lizistack.cn/en/notes/learning-in-public/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/learning-in-public/</guid><description>A useful learning note is not a tutorial paraphrase. It keeps the question, the experiment path, and the boundary of the conclusion together.</description><pubDate>Mon, 26 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Keep a Ledger for Every Experiment</title><link>https://lizistack.cn/en/notes/an-experiment-ledger/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/an-experiment-ledger/</guid><description>An experiment note is not a diary; it is a small contract covering the hypothesis, cost, evidence, and next decision.</description><pubDate>Sat, 24 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Reading DeepFM: Learning Low- and High-Order Feature Interactions Together</title><link>https://lizistack.cn/en/notes/deepfm-reading/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/deepfm-reading/</guid><description>How shared embeddings connect the FM and deep branches, and why the architecture remains a useful recommender-systems baseline.</description><pubDate>Wed, 21 Oct 2026 00:00:00 GMT</pubDate></item><item><title>How Deep Should I Learn a New Technology?</title><link>https://lizistack.cn/en/notes/how-deep-should-i-learn/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/how-deep-should-i-learn/</guid><description>Depth should follow frequency, the cost of being wrong, and whether the technology forms part of your core judgment.</description><pubDate>Wed, 21 Oct 2026 00:00:00 GMT</pubDate></item><item><title>What a Recommender-Model Experiment Should Leave Behind</title><link>https://lizistack.cn/en/notes/engineering-note-template/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/engineering-note-template/</guid><description>More useful than remembering that a metric improved is keeping how the problem was defined, how the experiment was designed, and where the conclusion stops.</description><pubDate>Sun, 18 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Reading ESMM: Modeling Post-Click Conversion in the Entire Space</title><link>https://lizistack.cn/en/notes/esmm-reading/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/esmm-reading/</guid><description>How the exposure-click-conversion probability chain helps ESMM address sample-selection bias and sparse supervision in CVR estimation.</description><pubDate>Sun, 18 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Reading ReAct: Closing the Loop Between Reasoning and Action</title><link>https://lizistack.cn/en/notes/react-reading/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/react-reading/</guid><description>How interleaved Thought, Action, and Observation trajectories connect language reasoning to tools—and what production agents still need beyond the prompt.</description><pubDate>Wed, 14 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Leave a Table for an Unfinished Agent Idea</title><link>https://lizistack.cn/en/notes/a-small-lab/</link><guid isPermaLink="true">https://lizistack.cn/en/notes/a-small-lab/</guid><description>An engineering experiment is not a product plan. It records hypotheses, prototypes, and explorations that do not yet have an answer.</description><pubDate>Fri, 09 Oct 2026 00:00:00 GMT</pubDate></item></channel></rss>