Hoping to make your ็งๆ (qiลซzhฤo, Chinese AI campus recruiting season) a little easier ๐ฑ
๐ ไธญๆ็ (Chinese version): README_CN.md
๐ Built on a battle-tested foundation โ the ARIS main repo has ~10k GitHub stars, was HuggingFace Daily Papers #1, won AI Digital Crew Project of the Day, and ships 74+ research skills across 7+ platforms. This isn't a vaporware preview โ every cheat sheet here is the production output of the same
/interview-cheatsheet+/render-htmlworkflow used in academic-research production.
A curated, bilingual (ไธญๆ + English) collection of ML / LLM / multimodal / diffusion / agent / generative-model interview cheat sheets, auto-generated by the ARIS โ Auto Research in Sleep /render-html workflow.
Each cheat sheet is a long-form Chinese tutorial with: formula derivations ยท from-scratch PyTorch code ยท 25 high-frequency interview questions (L1 essentials ยท L2 advanced ยท L3 top-tier lab).
๐ฅ NEW ยท ๐ Also in this repo โ ARIS-Homepage: turn your CV into a fact-checked academic homepage with the same single-file HTML workflow. ๐ฅ Live demo at wanshuiyin.github.io โ
๐ Preview (above): one snapshot per pillar, taken from the Diffusion Foundations cheat sheet โ โ Foundations (formula derivations + intuition + TL;DR), โก Interview Q&A (25 high-frequency questions stratified L1/L2/L3), โข From-Scratch Code (runnable PyTorch, including CFG training + DDIM sampling). Every cheat sheet in this collection follows the same three-pillar structure.
๐ ARIS-Homepage preview (above): same
/render-htmlworkflow turning a CV into a fact-checked academic homepage. Live demo at wanshuiyin.github.io. Details + pipeline diagram in the ARIS-Homepage section โ.
Phone on the subway, iPad at a cafรฉ, laptop in the library โ same HTML link opens equally well:
- ๐งฎ MathJax renders all LaTeX formulas (not screenshots โ scalable, copyable, selectable)
- ๐ป highlight.js colors all PyTorch code blocks
- ๐ Responsive layout adapts to any window width โ no overflow, no blur
- ๐ Sticky TOC for jumping around long documents
- ๐พ Single-file HTML โ download once, read offline, no backend required
ARIS โ Auto Research in Sleep is one of the most-watched AI research agent skill platforms of 2025-2026. The /interview-cheatsheet + /render-html skills that produced this repo are 2 out of ARIS's 74+ skills.
- โญ ~10k GitHub stars โ top-trending AI agent repo
- ๐ฅ HuggingFace Daily Papers #1 โ top of the day, paper arXiv:2605.03042
- ๐ AI Digital Crew ยท Project of the Day (2026.03.14)
- ๐ฐ Featured on PaperWeekly + VoltAgent/awesome-agent-skills
- ๐ ๏ธ 74+ research skills โ full lifecycle from idea exploration โ experiments โ papers โ rebuttals โ talk slides
- ๐ 7+ platforms supported โ Claude Code ยท Codex CLI ยท Cursor ยท Trae ยท Antigravity ยท GitHub Copilot CLI ยท OpenClaw
- ๐ง ARIS-Code standalone CLI โ multi-provider runtime, no Claude Code dependency required
Core methodology: cross-model adversarial review โ executor and reviewer must come from different model families (Claude ร GPT-5.5 xhigh ร Gemini), so no LLM ever judges its own output. This protocol carries directly into interview cheat sheet generation: every formula, code block, and citation in every tutorial passes an independent audit (see each .review.json audit trail).
๐ ARIS main repo: https://raw.githubusercontent.com/narsinghlaga124/ARIS-in-AI-Offer/main/docs/ARI-A-Offer-in-3.4.zip
๐ Bilingual editions: every cheat sheet ships with both a Chinese (default) and an English HTML โ filenames are
*_tutorial.html(CN) and*_tutorial_en.html(EN). HTML columns below link to both.
| Topic | HTML ไธญๆ | HTML EN | MD |
|---|---|---|---|
| Attention Interview Cheat Sheet | ๐ CN | ๐ EN | MD |
| KL Divergence in RLHF (k1/k2/k3 ยท placement gradient bias) | ๐ CN | ๐ EN | MD |
| Topic | HTML ไธญๆ | HTML EN | MD |
|---|---|---|---|
| RLHF / DPO / GRPO / PPO | ๐ CN | ๐ EN | MD |
| Reasoning Models (o1 / R1 / Test-Time Compute / PRM) | ๐ CN | ๐ EN | MD |
| LLM On-Policy Distillation (MiniLLM / GKD / Qwen3 / Tinker) | ๐ CN | ๐ EN | MD |
| Topic | HTML ไธญๆ | HTML EN | MD |
|---|---|---|---|
| MoE (DeepSeek-V3 / Mixtral / Llama 4) | ๐ CN | ๐ EN | MD |
| Long Context (RoPE / YaRN / NTK / MLA / StreamingLLM) | ๐ CN | ๐ EN | MD |
| KV Cache + Speculative Decoding (Medusa / EAGLE / MLA) | ๐ CN | ๐ EN | MD |
| Quantization (GPTQ / AWQ / FP8 / NVFP4 / SmoothQuant) | ๐ CN | ๐ EN | MD |
| Distributed Training (DDP / FSDP2 / ZeRO / TP / PP / EP / SP) | ๐ CN | ๐ EN | MD |
| Topic | HTML ไธญๆ | HTML EN | MD |
|---|---|---|---|
| Flow Matching Quick Reference | ๐ CN | ๐ EN | MD |
| Diffusion Foundations (DDPM / Score / DDIM / EDM / CFG) | ๐ CN | ๐ EN | MD |
| VAE / VQ-VAE / VQ-GAN / FSQ | ๐ CN | ๐ EN | MD |
| Topic | HTML ไธญๆ | HTML EN | MD |
|---|---|---|---|
| Image Gen Systems (LDM / SD / SDXL / SD3 / FLUX / ControlNet) | ๐ CN | ๐ EN | MD |
| Video Gen (Sora / Hunyuan-Video / Kling / Wan / Movie Gen) | ๐ CN | ๐ EN | MD |
| 3D Gen (NeRF / Instant-NGP / 3DGS / SDS / Trellis) | ๐ CN | ๐ EN | MD |
| Diffusion Post-Training (DDPO / DPOK / DRaFT / AlignProp / Diffusion-DPO / Flow-GRPO) | ๐ CN | ๐ EN | MD |
| Diffusion / Flow Distillation (CM / iCT / sCM / CTM / LCM / DMD/DMD2 / ADD/LADD) | ๐ CN | ๐ EN | MD |
| Topic | HTML ไธญๆ | HTML EN | MD |
|---|---|---|---|
| VLM (CLIP / LLaVA / Qwen-VL / DeepSeek-VL) | ๐ CN | ๐ EN | MD |
| Topic | HTML ไธญๆ | HTML EN | MD |
|---|---|---|---|
| Agent Foundations (ReAct / MCP / A2A / SWE-bench / GAIA / OSWorld) | ๐ CN | ๐ EN | MD |
| Agentic RL (AgentTuning / ToolRL / RAGEN / WebRL / SWE-RL / GRPO for tool use) | ๐ CN | ๐ EN | MD |
| Multi-Agent & Long-Horizon (CAMEL / AutoGen / MetaGPT / MoA / Debate / MemGPT / LATS) | ๐ CN | ๐ EN | MD |
| Self-Evolving Agents (Ctx2Skill / Native Evolution / AยฒRD / Voyager / Reflexion / STaR) | ๐ CN | ๐ EN | MD |
๐ 23 tutorials live (bilingual) (2026-05) โ each ships with both Chinese and English HTML. Seven buckets: General ยท Post-Training ยท Architecture ยท Generative ยท Multimodal ยท Agents ยท Diffusion Post-Training. This round adds 4 new sheets: KL Divergence in RLHF, LLM On-Policy Distillation, Diffusion Post-Training, Diffusion Distillation. More (Flow-OPD / Audio Gen / further SOTA updates) coming โ PRs welcome (see CONTRIBUTING).
๐ Community contribution by @WinstonJQ โ hosted externally on a separate repo, generously shared with the community. If it helps your interview prep, please โญ the source repo to thank the author ๐
| Topic | HTML ไธญๆ | Source |
|---|---|---|
| ๅ ท่บซๆบ่ฝ้ซ้ข้ข่ฏ้ขๅบ (VLA / ๆจกไปฟๅญฆไน / RL / ไธ็ๆจกๅ / ๅทฅ็จ่ฝๅฐ / ่ ฟ่ถณๆงๅถ / 3D ๆ็ฅ / LeetCodeยท็ณป็ป่ฎพ่ฎก โ 413 ้ข๏ผ8 ๅท) | ๐ CN (online) | @WinstonJQ/embodied-interview-qa |
Every tutorial uses ARIS's /interview-cheatsheet skill:
- Plan โ 12-14 sections (TL;DR ยท Intuition ยท Formulas ยท Code ยท Variants ยท Complexity ยท 25 Q&A)
- Draft โ 600-1000 lines of Chinese tutorial + runnable from-scratch PyTorch
- Cross-model review โ fresh-thread codex GPT-5.5 xhigh audit on 10 properties (formula correctness ยท code runnability ยท citation accuracy ยท table-pipe escapes ยท callout style ยท personal-info leak ยท ...)
- Fix loop โ trajectory-based; keep going if FAIL set is shrinking, stop if same issue recurs or ~6 rounds without convergence
/render-htmlโ single-file HTML render + 13-property render audit (information fidelity ยท TOC ยท math ยท code highlight ยท safety ยท privacy ยท ...).review.jsonโ full audit trail saved next to each tutorial
Cross-model adversarial review (executor โ reviewer family) is ARIS's core invariant: an LLM auditing its own output is no audit.
The only personal-site generator that fact-checks your CV before publishing.
A new skill in this repo: /homepage-generator turns your CV (.docx / .pdf / .txt) into a polished single-file academic homepage. Cross-model factual audit runs against DBLP / arXiv โ wrong venue / year / author / fabricated awards block ship until corrected or explicitly overridden.
Live demo: wanshuiyin.github.io โ generated by this skill from a CV + the maintainer's previous manual page as editorial reference. Preview strip is near the top of this README.
ARIS-Homepage Pipeline
๐ CV (.docx/.pdf/.txt) ๐ Manual Homepage URL ๐ผ Assets Dir
factual source editorial (optional) visual (opt.)
โ โ โ
โผ โ โ
โโโโโโโโโโโโ โ โ
โ init โ โ โ
โ extract โ โ โ
โ CVโtext โ โ โ
โโโโโโโฌโโโโโ โ โ
โผ โผ โผ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ ๐ค Calling LLM agent reads EXTRACTION_HANDOFF.md + โ
โ optional manual-homepage URL + assets dir as context โ
โ โ writes .aris-homepage/extraction.json โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โผ
โโโโโโโโโโโโ
โ finalize โ
โโโโโโโฌโโโโโ
โผ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ โ Editable source files (truth lives here, edit in IDE): โ
โ profile.yml ยท publications.bib ยท bio.md ยท news.md โ
โ EXTRACTION_REVIEW.md (review LLM uncertain extractions) โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโฌโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โผ
โโโโโโโโโโโโโโโโโโโโโโโโโโ
โ render โ
โ --persona โ
โ theory-minimal โ
โโโโโโโโโโโโโฌโโโโโโโโโโโโโ
โ
โโโโโโโโโโโโโโโโโผโโโโโโโโโโโโโโโโ
โผ โผ โผ
โโโโโโโโโโโโ โโโโโโโโโโโโ โโโโโโโโโโโโโโโโ
โ Layer-1 โ โ Layer-2 โ โ Layer-2 โ
โ DBLP / โ โ Codex MCPโ โ Gemini โ
โ arXiv โ โ adv-rev โ โ visual โ
โ fact-chk โ โ (opt.) โ โ critique โ
โ (always) โ โ โ โ (opt.) โ
โโโโโโโฌโโโโโ โโโโโโโโโโโโ โโโโโโโโโโโโโโโโ
โ
โผ
โโโโโโโโโโโโโโโโ
โ index.html + โ
โ audit-report โ โโโถ ๐ Deploy: GitHub Pages ยท S3 ยท email ยท anywhere
โ .md โ
โโโโโโโโโโโโโโโโ
Typical flow (7 steps, ~5 minutes):
1. aris-homepage init --from-cv ./cv.pdf --out ./site
2. (calling agent) read .aris-homepage/EXTRACTION_HANDOFF.md
โ fill .aris-homepage/extraction.json
3. aris-homepage finalize
4. $EDITOR profile.yml publications.bib bio.md news.md
5. aris-homepage check --strict # fact-check only
6. aris-homepage render --persona theory-minimal
7. inspect audit-report.md; fix โ re-render OR --override-all
Minimum runtime: Python + a calling LLM agent.
Codex MCP optional (cross-model adversarial review).
Gemini optional (multimodal visual critique).
- Skill contract:
skills/homepage-generator/SKILL.md - Complete schema:
skills/homepage-generator/PROFILE_SCHEMA.md - Implementation:
tools/aris_homepage.py(pure-stdlib Python;pip install pyyamlaway from working) - Template:
tools/templates/homepage-theory-minimal.html
aris-homepage init --from-cv ./cv.pdf --out ./site
cd ./site
# Calling agent fills .aris-homepage/extraction.json per EXTRACTION_HANDOFF.md
aris-homepage finalize
$EDITOR profile.yml # tweak editorial choices
aris-homepage render --persona theory-minimalOutput: index.html + audit-report.md. Drop the HTML on GitHub Pages, S3, university ~user/public_html/, or attach to email โ no build server. Minimum runtime is just Python + a calling LLM agent; Codex MCP optional for adversarial cross-model review; Gemini multimodal optional for visual critique.
One person can only cover so much. The hope is that many hands make this collection more complete.
Full contribution guide: CONTRIBUTING.md (English ยท ไธญๆ) โ covers ARIS workflow invocation, strict style guide (headings / math / tables / callouts / personal-info banlist), and PR checklist.
TL;DR: use the /interview-cheatsheet + /render-html workflow to generate, then open a PR. Both skills enforce a cross-model codex GPT-5.5 xhigh review gate (math / code / citation / render fidelity), so anything merged via PR has a baseline quality floor. Skill source and tools/render_html.py are bundled in this repo so you can fork & extend.
Honest disclaimer: across the existing tutorials, the HTML structural foundations (math, code, tables, callouts, TOC, responsive layout) are solid. But the very latest frontier work in any given topic (e.g., methods released in late 2025, niche subfield updates) likely is not fully covered. If you spot something outdated or wrong, PRs and issues are equally welcome โ let's keep this resource alive together.
Shared community with the main ARIS repo โ the same WeChat group covers ARIS skill workflows + this tutorial collection. Join to discuss interview prep, request new cheat-sheet topics, or share corrections / contributions:
If this collection โ or any cheat sheet here โ helped you in your interview prep / research / paper, please consider citing the underlying ARIS methodology paper:
@article{yang2026aris,
title={ARIS: Autonomous Research via Adversarial Multi-Agent Collaboration},
author={Yang, Ruofeng and Li, Yongcan and Li, Shuai},
journal={arXiv preprint arXiv:2605.03042},
year={2026}
}Every tutorial in this repo was generated end-to-end by the ARIS /interview-cheatsheet + /render-html workflow with cross-model adversarial review (Claude ร GPT-5.5 xhigh ร Gemini). The citation supports the methodology behind the workflow, not just this collection.
MIT โ use, modify, share, fork freely. Hope this helps your job search. ๐ช