Merge pull request #5 from wanshuiyin/fix/audit-chain-backlog

B: clear the review-chain backlog — honest backfills, blog honesty, content-bug fixes, 23 EN reviews

Author: wanshuiyin

.github/workflows/review-audit.yml

@@ -0,0 +1,45 @@
+name: review-audit
+
+# Enforces the ARIS cross-model review invariant on every shippable docs/ artifact:
+# each academic-template HTML must be a reviewed view with a traceable Codex thread
+# AND a recorded source hash that still matches its current source. Runs the repo's
+# own tools/verify_reviews.py in strict mode (see that script for the full rule set
+# and the explicit EXEMPTIONS allowlist for hand-authored, non-pipeline artifacts).
+
+on:
+  pull_request:
+    paths:
+      - 'docs/**'
+      - 'tools/verify_reviews.py'
+      - 'tools/render_html.py'
+      - 'tools/templates/**'
+      - 'tools/tutorials_render_manifest.json'
+      - '.github/workflows/review-audit.yml'
+  push:
+    branches: [main]
+    paths:
+      - 'docs/**'
+      - 'tools/verify_reviews.py'
+      - 'tools/render_html.py'
+      - 'tools/templates/**'
+      - 'tools/tutorials_render_manifest.json'
+      - '.github/workflows/review-audit.yml'
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+concurrency:
+  group: review-audit-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  verify:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+      - name: Verify the review audit chain (strict + reproduce)
+        run: python3 tools/verify_reviews.py --mode strict --reproduce

README.md

@@ -42,7 +42,7 @@ Each cheat sheet is a long-form Chinese tutorial with: formula derivations · fr
   </a>
 </p>
 
-> 📝 **Long-form blog preview** (above): same `/render-html` workflow producing technical surveys. *A Survey on Continuous DLM* (2026 H1, 6 papers) — bilingual survey by **Ruofeng Yang** (SJTU), written end-to-end via cross-model discussion (Claude Opus 4.7 + Codex GPT-5.5 xhigh + Gemini auto-gemini-3). [**📖 Read full blog ↗**](https://wanshuiyin.github.io/ARIS-in-AI-Offer/blogs/continuous_dlm_2026h1_survey.html).
+> 📝 **Long-form blog preview** (above): a standalone hand-authored long-form technical survey. *A Survey on Continuous DLM* (2026 H1, 6 papers) — bilingual survey by **Ruofeng Yang** (SJTU), written end-to-end via cross-model discussion (Claude Opus 4.7 + Codex GPT-5.5 xhigh + Gemini auto-gemini-3). [**📖 Read full blog ↗**](https://wanshuiyin.github.io/ARIS-in-AI-Offer/blogs/continuous_dlm_2026h1_survey.html).
 
 ### 📱 HTML reads cleanly everywhere
 
@@ -58,7 +58,7 @@ Phone on the subway, iPad at a café, laptop in the library — same HTML link o
 
 ## 📢 What's New
 
-- **2026-05-28** — ![NEW](https://img.shields.io/badge/NEW-red?style=flat-square) 📝 **First blog shipped: *A Survey on Continuous DLM* (2026 H1, 6 papers)** — long-form bilingual technical survey by **Ruofeng Yang** (SJTU), written end-to-end through the `ARIS-in-AI-Offer` workflow (Claude Opus 4.7 + Codex GPT-5.5 xhigh + Gemini auto-gemini-3 cross-model discussion). Compares **ELF**, ByteDance **Cola-DLM**, and **Flow-Matching family** across discrete-DLM problems, the "known-unknown" continuous space idea, training pipeline, architecture / params / shapes, inference grids + Tab 6/7 numerical results, denoising trajectories, and a Field Landscape against Cola-DLM. Lives at [`docs/blogs/continuous_dlm_2026h1_survey.html`](docs/blogs/continuous_dlm_2026h1_survey.html) (1.7 MB self-contained, no build) — demonstrates the kind of long-form analysis the [`render-html`](skills/render-html/SKILL.md) toolchain can produce. **Preview strip + live link** at the top of this README. ([8475a2d](https://github.com/wanshuiyin/ARIS-in-AI-Offer/commit/8475a2d))
+- **2026-05-28** — ![NEW](https://img.shields.io/badge/NEW-red?style=flat-square) 📝 **First blog shipped: *A Survey on Continuous DLM* (2026 H1, 6 papers)** — long-form bilingual technical survey by **Ruofeng Yang** (SJTU), written end-to-end through cross-model discussion (Claude Opus 4.7 + Codex GPT-5.5 xhigh + Gemini auto-gemini-3). Compares **ELF**, ByteDance **Cola-DLM**, and **Flow-Matching family** across discrete-DLM problems, the "known-unknown" continuous space idea, training pipeline, architecture / params / shapes, inference grids + Tab 6/7 numerical results, denoising trajectories, and a Field Landscape against Cola-DLM. Lives at [`docs/blogs/continuous_dlm_2026h1_survey.html`](docs/blogs/continuous_dlm_2026h1_survey.html) (1.7 MB self-contained, no build) — a hand-authored long-form HTML, not yet part of the audited [`/render-html`](skills/render-html/SKILL.md) pipeline. **Preview strip + live link** at the top of this README. ([8475a2d](https://github.com/wanshuiyin/ARIS-in-AI-Offer/commit/8475a2d))
 - **2026-05-28** — ![POLISH](https://img.shields.io/badge/POLISH-blue?style=flat-square) ⚡ **`render-html` P0 polish + all 23 tutorials regenerated** — academic template gained 7 interactive features: print degradation fix (PDF no longer loses `<details>` content), TOC sidebar scrollspy (current section auto-highlights as you scroll), figure lightbox (native `<dialog>` with focus trap + Esc), long-code auto-collapse (`<pre>` ≥30 lines wrapped in `<details class="code-card">`, per-block override via ` ```python {collapsed}` / `{open}` fence flags), paper citation popover (new `[[key]]` MD syntax + `--papers <papers.json>` sidecar), eyebrow cleanup (marketing uppercase → body-serif gray), `--blog-mode` infrastructure (opt-in `aris-blog` body class). XSS-hardened script injection via `json_for_script()` (escapes `</script>` break-out). All 23 bilingual tutorial pairs (= 46 HTMLs) regenerated to pick up the new template shell — source MDs untouched. Codex GPT-5.5 xhigh 4-round review (design × 2 → code × 1 → spot-check × 1). Try it: scroll [`attention_tutorial.html`](docs/tutorials/attention_tutorial.html) and watch the TOC sidebar follow ([b79c57d](https://github.com/wanshuiyin/ARIS-in-AI-Offer/commit/b79c57d), [8793f40](https://github.com/wanshuiyin/ARIS-in-AI-Offer/commit/8793f40)).
 - **2026-05-26** — ![NEW](https://img.shields.io/badge/NEW-red?style=flat-square) 🐍 **5 runnable PyTorch tutorial scripts** — first runnable-code contribution in [`docs/tutorials/code/`](docs/tutorials/code/): [`mha.py`](docs/tutorials/code/mha.py) (MHA + causal mask) · [`axial_attention.py`](docs/tutorials/code/axial_attention.py) (H/W axial + complexity table) · [`flow_matching.py`](docs/tutorials/code/flow_matching.py) (Rectified Flow on 2D moons) · [`mmdit_block.py`](docs/tutorials/code/mmdit_block.py) (double-stream MMDiT block) · [`toy_mmdit_t2i_pipeline.py`](docs/tutorials/code/toy_mmdit_t2i_pipeline.py) (end-to-end T2I skeleton). Pure PyTorch, CPU-runnable in seconds, every script ships with built-in `assert` sanity checks (shape parity, numerical agreement with `nn.MultiheadAttention` where applicable). Pairs with [`attention_tutorial.md`](docs/tutorials/attention_tutorial.md) / [`flow_matching_tutorial.md`](docs/tutorials/flow_matching_tutorial.md) / [`image_generation_systems_tutorial.md`](docs/tutorials/image_generation_systems_tutorial.md) ([f63f468](https://github.com/wanshuiyin/ARIS-in-AI-Offer/commit/f63f468)).
 - **2026-05-24** — ![NEW](https://img.shields.io/badge/NEW-red?style=flat-square) 🐙 **ARIS-Homepage v1.1: `--from-repos`** — snapshot user-selected `owner/repo` list via `gh` CLI; LLM agent merges repo timelines into homepage News + `featured_projects[].github`. Private repos skipped by default. Closes [#2](https://github.com/wanshuiyin/ARIS-in-AI-Offer/issues/2) by [@Yafei-Liu99](https://github.com/Yafei-Liu99) ([cdcf9a2](https://github.com/wanshuiyin/ARIS-in-AI-Offer/commit/cdcf9a2)).

README_CN.md

@@ -42,7 +42,7 @@
   </a>
 </p>
 
-> 📝 **长文 blog 预览**（上图）：同一套 `/render-html` workflow 产出技术综述。*Continuous DLM 综述*（2026 上半年 6 篇）—— 作者 **杨若峰 (Ruofeng Yang)**（上海交大），通过 ARIS 跨模型工作流端到端撰写（Claude Opus 4.7 + Codex GPT-5.5 xhigh + Gemini auto-gemini-3）。[**📖 阅读完整 blog ↗**](https://wanshuiyin.github.io/ARIS-in-AI-Offer/blogs/continuous_dlm_2026h1_survey.html)。
+> 📝 **长文 blog 预览**（上图）：独立手写的长文技术综述。*Continuous DLM 综述*（2026 上半年 6 篇）—— 作者 **杨若峰 (Ruofeng Yang)**（上海交大），通过跨模型讨论端到端撰写（Claude Opus 4.7 + Codex GPT-5.5 xhigh + Gemini auto-gemini-3）。[**📖 阅读完整 blog ↗**](https://wanshuiyin.github.io/ARIS-in-AI-Offer/blogs/continuous_dlm_2026h1_survey.html)。
 
 ### 📱 HTML 格式哪里都能读，清清楚楚
 
@@ -58,7 +58,7 @@
 
 ## 📢 最新动态
 
-- **2026-05-28** — ![NEW](https://img.shields.io/badge/NEW-red?style=flat-square) 📝 **首篇 blog 上线：*Continuous DLM 综述*（2026 上半年 6 篇）** —— 长篇双语技术综述，作者 **杨若峰 (Ruofeng Yang)**（上海交大），通过 `ARIS-in-AI-Offer` 工作流端到端撰写（Claude Opus 4.7 整合 Codex GPT-5.5 xhigh + Gemini auto-gemini-3 跨模型讨论）。覆盖：离散 DLM 已有问题、ELF "已知未知" 连续空间核心想法、Flow-Matching 家族 5 paper 同期对比、ELF 训练 pipeline、架构/参数量/shape 最详细一节、推理代码+时间网格+Tab 6/7 全数字、去噪轨迹定性效果、与字节 Cola-DLM 的 Field Landscape 对比、Q&A、引用资源。文件在 [`docs/blogs/continuous_dlm_2026h1_survey.html`](docs/blogs/continuous_dlm_2026h1_survey.html)（1.7 MB 自包含单文件，无 build）—— 展示 [`render-html`](skills/render-html/SKILL.md) 工具链能产出的 long-form 分析风格。**预览图 + Live 链接见 README 顶部**。（[8475a2d](https://github.com/wanshuiyin/ARIS-in-AI-Offer/commit/8475a2d)）
+- **2026-05-28** — ![NEW](https://img.shields.io/badge/NEW-red?style=flat-square) 📝 **首篇 blog 上线：*Continuous DLM 综述*（2026 上半年 6 篇）** —— 长篇双语技术综述，作者 **杨若峰 (Ruofeng Yang)**（上海交大），通过跨模型讨论端到端撰写（Claude Opus 4.7 + Codex GPT-5.5 xhigh + Gemini auto-gemini-3）。覆盖：离散 DLM 已有问题、ELF "已知未知" 连续空间核心想法、Flow-Matching 家族 5 paper 同期对比、ELF 训练 pipeline、架构/参数量/shape 最详细一节、推理代码+时间网格+Tab 6/7 全数字、去噪轨迹定性效果、与字节 Cola-DLM 的 Field Landscape 对比、Q&A、引用资源。文件在 [`docs/blogs/continuous_dlm_2026h1_survey.html`](docs/blogs/continuous_dlm_2026h1_survey.html)（1.7 MB 自包含单文件，无 build）—— 一份手写长文 HTML，尚未纳入受审的 [`/render-html`](skills/render-html/SKILL.md) 流程。**预览图 + Live 链接见 README 顶部**。（[8475a2d](https://github.com/wanshuiyin/ARIS-in-AI-Offer/commit/8475a2d)）
 - **2026-05-28** — ![POLISH](https://img.shields.io/badge/POLISH-blue?style=flat-square) ⚡ **`render-html` P0 polish + 23 个 tutorial 全量 regenerate** —— academic 模板加了 7 个交互特性：print degradation 修复（PDF 导出不再丢 `<details>` 内容）、TOC 侧栏 scrollspy（当前 section 跟随滚动高亮）、figure lightbox（原生 `<dialog>` + focus trap + Esc）、长代码自动折叠（`<pre>` ≥30 行包成 `<details class="code-card">`，可用 ` ```python {collapsed}` / `{open}` fence flag 单块 override）、论文引用 popover（新增 `[[key]]` MD 语法 + `--papers <papers.json>` sidecar）、eyebrow 风格清理（marketing 大写 → 正文衬线灰）、`--blog-mode` 基础设施（opt-in `aris-blog` body class）。XSS 加固：`json_for_script()` 把 `<` / `>` / `&` / U+2028 等 escape 成 `\\uXXXX`，paper title 含 `</script>` 不会 break out。23 对双语 tutorial（共 46 个 HTML）全量 regenerate 吃到新模板，源 MD 没动。Codex GPT-5.5 xhigh 4 轮审（设计 ×2 → 代码 ×1 → spot-check ×1）。试试：滚动 [`attention_tutorial.html`](docs/tutorials/attention_tutorial.html) 看左侧 TOC 跟随高亮（[b79c57d](https://github.com/wanshuiyin/ARIS-in-AI-Offer/commit/b79c57d)、[8793f40](https://github.com/wanshuiyin/ARIS-in-AI-Offer/commit/8793f40)）。
 - **2026-05-26** — ![NEW](https://img.shields.io/badge/NEW-red?style=flat-square) 🐍 **5 个可跑的 PyTorch 教学脚本** —— [`docs/tutorials/code/`](docs/tutorials/code/) 是仓库第一批 runnable-code 贡献：[`mha.py`](docs/tutorials/code/mha.py)（MHA + causal mask）、[`axial_attention.py`](docs/tutorials/code/axial_attention.py)（H/W axial + 复杂度表）、[`flow_matching.py`](docs/tutorials/code/flow_matching.py)（2D moons 上的 Rectified Flow）、[`mmdit_block.py`](docs/tutorials/code/mmdit_block.py)（双流 MMDiT block）、[`toy_mmdit_t2i_pipeline.py`](docs/tutorials/code/toy_mmdit_t2i_pipeline.py)（端到端 toy T2I pipeline）。纯 PyTorch，CPU 几秒到几十秒跑完，每个脚本自带 `assert` sanity check（shape 对齐、必要时跟 `nn.MultiheadAttention` 数值校准）。配合 [`attention_tutorial.md`](docs/tutorials/attention_tutorial.md) / [`flow_matching_tutorial.md`](docs/tutorials/flow_matching_tutorial.md) / [`image_generation_systems_tutorial.md`](docs/tutorials/image_generation_systems_tutorial.md) 阅读（[f63f468](https://github.com/wanshuiyin/ARIS-in-AI-Offer/commit/f63f468)）。
 - **2026-05-24** — ![NEW](https://img.shields.io/badge/NEW-red?style=flat-square) 🐙 **ARIS-Homepage v1.1：`--from-repos`** — 用户选定 `owner/repo` 列表通过 `gh` CLI 抓快照，调用方 LLM agent 把 repo 时间线 merge 进主页 News + `featured_projects[].github`。默认拒绝 private repo。Closes [#2](https://github.com/wanshuiyin/ARIS-in-AI-Offer/issues/2) by [@Yafei-Liu99](https://github.com/Yafei-Liu99)（[cdcf9a2](https://github.com/wanshuiyin/ARIS-in-AI-Offer/commit/cdcf9a2)）。