dsh-vibe-math 2.5.0 → 2.6.0

package/docs/release-notes/RELEASE-NOTES-2.6.0.md

@@ -0,0 +1,268 @@
+# dsh-vibe-math 2.6.0 — 发布说明（中文）
+
+> 上一版：2.5.0。本版新增**最终论文（final paper）**功能：四个预设默认在收口时自动撰写一篇只整理
+> 已有证据的完整论文，且论文阶段发生在"标记完成"**之前**。既有参数默认值、数据格式与目录结构未变，
+> DSH 支撑窗口不变，无需迁移。
+
+---
+
+## 概览
+
+- 新增**最终论文**：默认开启（`finalPaper=true`）。收口信号命中后、**run 被标记完成之前**进入 paper 阶段。
+- 产物固定在 `Paper/<id>/`：`paper.md`、`paper.tex`、`paper.pdf`（仅在宿主检测到 LaTeX 引擎并编译成功时）、
+  `paper.meta.json`、`paper.log.md`。
+- v4/v5 为**团队合写**：各自写自己的部分 → 合并 → 交叉互审 → 定稿代表定稿 → 全体明确"可交付"才定稿；
+  v5 的定稿代表默认院士，所办仅在手动路径下可用，且必须先与全所交流并开会。
+- 编译失败**先修复、后降级**：没有引擎或编译不通过时仍交付 tex+md，只记日志告警，不阻塞定稿，
+  也绝不覆盖已有的 `paper.pdf`。
+- 随功能修掉若干问题：v5 自动触发的心跳停摆、v2/v3 写手卡死可能永久阻塞、v3 随包语料的时间戳非确定性、
+  以及一处会被空过的论文判定检查。
+- 契约文档 `docs/final-paper.md` 随包发布；双语 README 补齐参数与用法。
+
+## 变更
+
+### 1. 新功能：最终论文（四套一致，默认开启）
+
+- **触发**：各预设用自己的收口信号（v2 严格终止、v3 完整收口、v4 一致性会议的一致停止票、
+  v5 每次解题票上的"已解决"判定）。命中后进入 paper 阶段，**论文完成或降级结束之后才把这次 run
+  标记为完成**——否则会议/唤醒机制已被禁用，团队合写无法进行。
+- **手动**：`/vibe paper [lang=zh|en] [format=both|md|tex] [force]`（v2/v3）、
+  `/v4 paper [… editor=office|resident:<id>] [force]`、`/v5 paper [… editor=office|academician] [force]`。
+  命令行覆盖只作用于本次（不写回持久参数）；`force` 重写已定稿的论文；参数非法返回 `kind:'error'`；
+  `finalPaper=false` 只关闭自动触发，手动命令仍可用并在返回里说明"自动已关闭"。
+- **幂等**：同一 run 只写一次（`finalizedAt` + 逐产物存在性检查），重复触发只补写缺失的产物。
+- **内容骨架（9 节，四套一致）**：① 标题、作者、日期与摘要 ② 引言与问题背景 ③ 原问题的完整解法
+  ④ 已检验通过的命题（含判定为真的估计值与证据来源）⑤ 已解决的子问题 / 中间成果
+  ⑥ 创造或发现的有价值之物（方法、理论、思想、经验、数学理解）⑦ 规律总结 ⑧ 讨论、局限与展望
+  ⑨ 附录：证据与文件索引。论文只整理已有证据、**不得编造**；未决与被否证的条目必须显式标注。
+
+### 2. 参数
+
+- `finalPaper`（默认 `true`，四套）：收口时自动撰写最终论文；`false` 只关闭自动触发。
+- `paperFormat`（默认 `both`，四套）：`both`（md+tex）/ `md`（只写 md，跳过编译且不报"缺 tex"）/ `tex`。
+- `paperLanguage`（默认 `zh`，四套）：`zh`（LaTeX 用 `ctexart`，引擎优先 `xelatex`）/
+  `en`（`article`，引擎优先 `pdflatex`）。
+- `paperCompilePdf`（默认 `true`，四套）：检测到 LaTeX 时是否编译 `paper.pdf`。
+- `paperLatexCommand`（默认空，四套）：优先尝试指定的引擎可执行文件；解析不到时按"未检测到"降级。
+- `paperEditor`（仅 v4/v5）：v4 默认 `office`（会话根 / 人类侧），可选 `resident:<id>`
+  （指定成员已离职则降级回 office 并在 meta/log 记明）；v5 默认 `academician`，`office` 仅手动路径可用。
+
+### 3. 产物、LaTeX 检测与降级
+
+- **目录与文件**：`Paper/<id>/`，id 经各预设既有的归一化函数处理（含路径穿越防护）；内容为
+  `paper.md`、`paper.tex`、`paper.pdf`、`paper.meta.json`（参数、触发原因、编译状态与引擎、`finalizedAt`、
+  v5 的咨询计数等）、`paper.log.md`（谁写了什么、谁提了什么意见、如何采纳）。
+- **检测顺序**：`xelatex` → `latexmk` → `pdflatex` → `lualatex` → `tectonic`（中文优先 xelatex，
+  英文 pdflatex 优先）；`paperLatexCommand` 非空时优先尝试它，之后回到标准顺序。
+- **编译**：在论文目录内执行、`-interaction=nonstopmode`、跑两遍；失败**先修复**（重跑 → 换引擎 →
+  最小模板），到上限后**降级**：保留 tex+md，`paper.meta.json` 记 `not-detected` / `failed`，
+  记日志并上报，不阻塞定稿、不抛错。
+- **不覆盖既有 PDF**：宿主 fs 只支持文本写，`paper.pdf` 只能由编译器子进程产生，插件只做存在性校验；
+  无引擎、编译持续失败、`format=md force` 三种形态都逐字节保留已有 `paper.pdf` 并在 meta 里记明原因。
+- `paperFormat=md` 时不产出 tex，因此跳过编译，且不得产生"缺 tex 无法编译"的警告。
+
+### 4. 团队合写与定稿代表（v4 / v5）
+
+- **流程**：各成员写自己负责的部分 → 合并（去重、统一术语与记号）→ 至少一轮**交叉互审** →
+  定稿代表梳理成最终稿 → 全体明确"可交付"才定稿；有异议则继续迭代（有轮次上限，超限记警告并把
+  分歧写进附录）。全过程写入 `paper.log.md`。
+- **v4**：`paperEditor` 默认 `office`，可选 `resident:<id>`。
+- **v5**：自动流程**固定用 `academician`**（所办是根会话，没有唤醒路径）；`office` 只能通过手动
+  `/v5 paper editor=office` 选择，此时所办**必须**先与全所交流（至少 1 条所办消息）并至少召开一次会议，
+  把结论写进定稿说明，然后调用 `vibe_v5_finalize_paper`；缺一即被拒（`V5_PAPER_CONSULT_REQUIRED`）。
+
+### 5. 本轮修复
+
+- **v5 自动触发的心跳停摆**：paper 分支在重新武装心跳之前返回，被节流的提问会永久卡住（新增的
+  自动触发守卫当场抓出）。现在 paper 的每个等待步骤都会重新武装心跳。
+- **v2/v3 写手卡死**：新增写手判决与回收——心跳会回收"已停滞或已不在注册表"的写手并自动重派
+  （有上限），随后给出明确的 `/vibe paper force` 提示；`force` 在 10 分钟窗口内返回可操作的错误
+  （含子代理 id），超过窗口则回收并重新派发；被回收的写手若稍后返回会被记录，而不是静默丢弃。
+- **既有 PDF 不被覆盖**（见 §3）：无引擎、编译持续失败、`format=md force` 三种形态都验证了
+  `paper.pdf` 逐字节保留。
+- **"不得编造"契约**：论文唤醒提示词明确要求只整理已有证据、不得编造、未决与被否证的条目必须
+  显式标注；把措辞改成"可以编造"会立刻被守卫挡住。
+- **v3 语料确定性**：paper 摘要嵌入的证据索引含毫秒时间戳文件名，两个语料 scrubber 增加时间戳归一化，
+  连续两次运行的随包语料逐字节一致。
+- **守卫**：自动触发（论文写完之前不得标记完成）、既有 pdf 保留与"不得编造"契约都有可证伪的检查
+  （把对应逻辑改坏会立刻变红）；原先可被空过的论文判定检查一并修正。
+
+### 6. 文档与契约
+
+- 新增并随包发布 `docs/final-paper.md`：合并用户确认的规格与设计审计修订，写明各预设真实的收口信号、
+  触发时序、产物、9 节骨架、LaTeX 检测与降级、团队流程与 v5 咨询门。
+- 双语 README：四套各补 5 个参数行（v4/v5 另有第 6 个 `paperEditor`）与"最终论文 / Final paper"小节。
+- `docs/test-timing.md` 与 `docs/AUDIT-CHECKLIST.md` 的套件/探针计数更正为 **57 = 39 套件 + 18 探针**，
+  人设灵敏度探针更正为 13 条；npm 描述加入最终论文说明。
+
+## 兼容性
+
+- **DSH 支撑窗口不变**：`0.1.2-alpha.4 … 0.2.0-rc.2`；组合行（≥0.1.7）与目录（≤0.1.6）两条线都保留。
+- **既有参数默认值未变**；新增的最终论文参数默认开启（`finalPaper=true`），要关掉自动触发设
+  `finalPaper=false`（手动命令仍可用）。
+- **数据与目录结构不变**：状态文件、项目布局与既有镜像位置照旧；`Paper/<id>/` 只是新增的产物目录，
+  无需迁移。
+- **提示词**：新增最终论文参数与 `/vN paper` 的说明，并写明"论文完成后才算完成"；随包语料已按当前
+  四套重新生成。
+
+## 升级
+
+```sh
+npm i dsh-vibe-math@latest
+```
+
+桌面版在「设置 → 插件」里更新；命令行 profile 用 `dsh plugin --profile <你的 profile> add dsh-vibe-math@latest`，随后重启 DSH。
+
+---
+
+# dsh-vibe-math 2.6.0 — Release Notes (English)
+
+> Previous: 2.5.0. This release adds the **final paper**: each of the four presets writes a complete paper
+> at closure that only organises evidence it already has, and the paper phase runs **before** the run is
+> marked complete. Existing parameter defaults, data formats and the directory layout are unchanged, the
+> supported DSH window is unchanged, and no migration is needed.
+
+---
+
+## Overview
+
+- **Final paper** is new and **on by default** (`finalPaper=true`). Once the closure signal fires and
+  **before the run is marked complete**, the run enters the paper phase.
+- The artifacts live in `Paper/<id>/`: `paper.md`, `paper.tex`, `paper.pdf` (only when the host has a
+  LaTeX engine and the compile succeeds), `paper.meta.json` and `paper.log.md`.
+- v4/v5 co-author the paper **as a team**: each member writes its own part → merge → cross-review →
+  the editor finalises → the paper is finalised only when everyone agrees it is deliverable. In v5 the
+  default editor is the academician; the institute office is available only on the manual path and must
+  first consult the whole institute and convene a meeting.
+- A failed compile is **repaired first, degraded second**: with no engine or a failing compile the tex+md
+  are still delivered with a logged warning, the finalisation is never blocked, and an existing
+  `paper.pdf` is never overwritten.
+- The release also fixes: a v5 automatic-trigger heartbeat stall, v2/v3 writers that could hang and block
+  forever, non-determinism in the v3 shipped corpus, and a paper verdict check that could pass vacuously.
+- The contract document `docs/final-paper.md` ships with the package, and both READMEs now document the
+  parameters and usage.
+
+## Changes
+
+### 1. New feature: the final paper (all four presets, on by default)
+
+- **Trigger**: each preset uses its own closure signal (v2 strict termination, v3 full closure, v4 the
+  closing meeting's unanimous stop vote, v5 the "solved" decision on each solve vote). The run then enters
+  the paper phase, and **the run is marked complete only after the paper is finalised or has degraded** —
+  otherwise the meeting/wake mechanisms are already disabled and team co-authoring cannot run.
+- **Manual**: `/vibe paper [lang=zh|en] [format=both|md|tex] [force]` (v2/v3),
+  `/v4 paper [… editor=office|resident:<id>] [force]`, `/v5 paper [… editor=office|academician] [force]`.
+  Command-line overrides apply to that invocation only (they are not written back to the persisted
+  parameters); `force` rewrites an already-finalised paper; an invalid argument returns `kind:'error'`;
+  `finalPaper=false` disables only the automatic trigger, and the manual command still works and says so.
+- **Idempotent**: one paper per run (`finalizedAt` plus a per-artifact existence check); a repeated
+  trigger only fills in missing artifacts.
+- **The nine-section skeleton (identical in all four presets)**: ① title, authors, date and abstract
+  ② introduction and problem background ③ the complete solution of the original problem ④ the verified
+  propositions (with the estimated truth values and their evidence) ⑤ solved sub-problems / intermediate
+  results ⑥ methods, theories, ideas, useful experience and mathematical understanding created or
+  discovered ⑦ patterns distilled from the above ⑧ discussion, limitations and outlook ⑨ appendix:
+  evidence and file index. The paper only organises existing evidence and **must not fabricate**;
+  unresolved and refuted items must be marked explicitly.
+
+### 2. Parameters
+
+- `finalPaper` (default `true`, all four): write the final paper automatically at closure; `false` disables
+  only the automatic trigger.
+- `paperFormat` (default `both`, all four): `both` (md+tex) / `md` (markdown only; skips compilation and
+  does not warn about a missing tex) / `tex`.
+- `paperLanguage` (default `zh`, all four): `zh` (`ctexart`, engine prefers `xelatex`) /
+  `en` (`article`, engine prefers `pdflatex`).
+- `paperCompilePdf` (default `true`, all four): compile `paper.pdf` when a LaTeX engine is detected.
+- `paperLatexCommand` (default empty, all four): try this engine executable first; if it cannot be
+  resolved, degrade as "not detected".
+- `paperEditor` (v4/v5 only): v4 defaults to `office` (the session root / human side) and also accepts
+  `resident:<id>` (if that member has left, it degrades back to office and the meta/log say so);
+  v5 defaults to `academician`, and `office` is available only on the manual path.
+
+### 3. Artifacts, LaTeX detection and degradation
+
+- **Layout**: `Paper/<id>/`, with the id normalised by each preset's existing helper (including path
+  traversal protection). It contains `paper.md`, `paper.tex`, `paper.pdf`, `paper.meta.json` (parameters,
+  trigger, compile status and engine, `finalizedAt`, the v5 consultation counts) and `paper.log.md`
+  (who wrote what, who raised which review point, and how it was adopted).
+- **Detection order**: `xelatex` → `latexmk` → `pdflatex` → `lualatex` → `tectonic` (`xelatex` first for
+  Chinese, `pdflatex` first for English). A non-empty `paperLatexCommand` is tried first, then the standard
+  order.
+- **Compilation**: run inside the paper directory with `-interaction=nonstopmode`, twice; on failure it is
+  **repaired first** (rerun → switch engine → minimal template) and only then **degraded**: tex+md are
+  kept, `paper.meta.json` records `not-detected` / `failed`, a warning is logged and reported, the
+  finalisation is not blocked and nothing is thrown.
+- **An existing PDF is never overwritten**: the host fs only supports text writes, so `paper.pdf` can only
+  be produced by the compiler subprocess and the plugin merely checks that it exists. With no engine, with
+  a persistently failing compile, and with `format=md force`, an existing `paper.pdf` is kept byte for byte
+  and the meta records why.
+- With `paperFormat=md` no tex is produced, so compilation is skipped and no "cannot compile without tex"
+  warning may appear.
+
+### 4. Team co-authoring and the editor (v4 / v5)
+
+- **Flow**: each member writes its own part → merge (deduplicate, unify terms and notation) → at least one
+  **cross-review** round → the editor turns it into the final draft → the paper is finalised only when
+  everyone agrees it is deliverable; objections keep iterating (bounded, after which a warning is recorded
+  and the disagreement goes into the appendix). The whole process is written to `paper.log.md`.
+- **v4**: `paperEditor` defaults to `office`, and also accepts `resident:<id>`.
+- **v5**: the automatic flow **always uses `academician`** (the office is the root session and has no wake
+  path). `office` is available only through the manual `/v5 paper editor=office`, and that path requires
+  the office to consult the whole institute first (at least one office message) and to convene at least one
+  meeting, to state the conclusion in the finalisation note, and then to call `vibe_v5_finalize_paper`;
+  without both counts it is refused (`V5_PAPER_CONSULT_REQUIRED`).
+
+### 5. Fixes in this release
+
+- **v5 automatic-trigger heartbeat stall**: the paper branch returned before re-arming the heartbeat, so a
+  throttled ask could hang forever (the new automatic-trigger guard caught it). Every waiting step of the
+  paper flow now re-arms the heartbeat.
+- **v2/v3 hung writers**: a new writer verdict and reaper — the heartbeat reaps writers that have stalled
+  or left the registry and re-dispatches them (bounded), after which it points at `/vibe paper force`;
+  `force` returns an actionable error inside the 10-minute window (including the subagent id) and reaps and
+  re-dispatches past it; a reaped writer that returns later is recorded instead of being dropped silently.
+- **An existing PDF is never overwritten** (see §3): with no engine, with a persistently failing compile
+  and with `format=md force`, the file is verified to be kept byte for byte.
+- **The "must not fabricate" contract**: the paper wake-up prompts now require that only existing evidence
+  is organised, that nothing may be fabricated, and that unresolved or refuted items are marked explicitly;
+  rewording that to "may fabricate" is caught by the guards immediately.
+- **v3 corpus determinism**: the paper summary embedded an evidence index with a millisecond-timestamped
+  log name; both corpus scrubbers now normalise the timestamp, so two consecutive runs produce
+  byte-identical shipped corpora.
+- **Guards**: the automatic trigger (the run must not be marked complete before the paper is written), the
+  preservation of an existing pdf and the "must not fabricate" contract all have falsifiable checks (break
+  the corresponding logic and they go red); a paper verdict check that used to pass vacuously was corrected
+  as well.
+
+### 6. Documentation and the contract
+
+- `docs/final-paper.md` is new and ships with the package: it merges the user-approved specification with
+  the design-audit amendments and states each preset's real closure signal, the ordering, the artifacts,
+  the nine-section skeleton, detection and degradation, the team flow and the v5 consultation gate.
+- Both READMEs gain the five parameter rows per preset (plus the sixth, `paperEditor`, on v4/v5) and a
+  "final paper" section.
+- The suite/probe counts in `docs/test-timing.md` and `docs/AUDIT-CHECKLIST.md` are corrected to
+  **57 = 39 suites + 18 probes**, and the persona sensitivity probes to 13; the npm description mentions
+  the final paper.
+
+## Compatibility
+
+- **Supported DSH window unchanged**: `0.1.2-alpha.4 … 0.2.0-rc.2`; both the composition-row (≥ 0.1.7) and
+  directory (≤ 0.1.6) lines are kept.
+- **Existing parameter defaults are unchanged**; the new final-paper parameters default to on
+  (`finalPaper=true`) — set `finalPaper=false` to disable the automatic trigger (the manual command still
+  works).
+- **No data or layout change**: the state file, the project layout and the existing mirror locations are
+  unchanged; `Paper/<id>/` is only a new artifact directory, and no migration is needed.
+- **Prompts**: the final-paper parameters and `/vN paper` are documented, and the prompts state that the
+  run is complete only after the paper. The shipped corpora were regenerated for the current four presets.
+
+## Upgrade
+
+```sh
+npm i dsh-vibe-math@latest
+```
+
+Desktop updates through Settings → Plugins; a CLI profile uses
+`dsh plugin --profile <your profile> add dsh-vibe-math@latest`, then restart DSH.

package/docs/test-timing.md

@@ -26,7 +26,7 @@ node tests/run-tests.mjs --json              # 机器可读汇总（含 skippedN
 
 ### 1.1 随包发布 vs 仅仓库（`npm pack --dry-run --json`）
 
-`package.json` 的 `files` 只发 `tests/` 的一个**子集**（发布物共 102 个文件，其中 `tests/` 18 项），
+`package.json` 的 `files` 只发 `tests/` 的一个**子集**（发布物共 105 个文件，其中 `tests/` 18 项），
 所以安装用户跑 `node tests/run-tests.mjs` 得到的是子集的结果——runner 会把缺失的跳过项打印出来，
 不会静默少跑：
 
@@ -35,7 +35,7 @@ node tests/run-tests.mjs --json              # 机器可读汇总（含 skippedN
   `audit-v5-sensitivity.mjs`、`selfdrive-v5.mjs`；套件 **10** 个：`audit-installer-compat`、
   `audit-installer-policy`、`audit-persona-surface`、`audit-preset-rows`、`e2e-v5-round2`、
   `formal-verify-v2`、`formal-verify-v3`、`formal-verify-v4`、`formal-verify-v5`、`prompt-v5-integrity`。
-- **仅仓库（不发）**：其余 28 个套件（v2/v3/v4 的 e2e、`selfdrive-v3`/`v4`、v2/v3 修复探针、
+- **仅仓库（不发）**：其余 29 个套件（v2/v3/v4 的 e2e、`selfdrive-v3`/`v4`、v2/v3 修复探针、
   `audit-fuzz-helpers` 之外的若干 `audit-*`……）与 7 个脚本（`audit-fuzz-helpers`、
   `audit-registration`、`audit-roundtrip-idempotence`、`audit-tool-exec`、
   `audit-v3-registration-parity`、`selfdrive-v3`、`selfdrive-v4`）。
@@ -46,10 +46,10 @@ node tests/run-tests.mjs --json              # 机器可读汇总（含 skippedN
 
 | 脚本 | 串行（sum） | 并行（wall） | 实测输出 |
 |---|---|---|---|
-| `tests/run-tests.mjs`（**56 项** = 38 套件 + 18 探针/变体） | ≈ 629 s | **≈ 208 s**（并发 4，speed-up x3.02） | `TOTAL 56  PASS 56  FAIL 0  (suites 38 · probes 18)`；关键路径 = `audit-formal-sensitivity` 208 s |
+| `tests/run-tests.mjs`（**57 项** = 39 套件 + 18 探针/变体） | ≈ 629 s | **≈ 208 s**（并发 4，speed-up x3.02） | `TOTAL 57  PASS 57  FAIL 0  (suites 39 · probes 18)`；关键路径 = `audit-formal-sensitivity` 208 s |
 | `tests/audit-formal-sensitivity.mjs`（49 探针，自带并发 4） | 803 s | **≈ 208 s** | 关键路径 = 12 个 v2 探针（每个 ≈42–60 s） |
 | `tests/audit-v5-sensitivity.mjs`（39 探针，串行） | ≈ 101 s | — | 每条 = 一次被测套件重跑 |
-| `tests/audit-persona-sensitivity.mjs`（11 探针） | ≈ 4.4 s | — | 本身很快，不需要并行 |
+| `tests/audit-persona-sensitivity.mjs`（16 探针） | ≈ 4.4 s | — | 本身很快，不需要并行 |
 | `tests/audit-prompt-invariants.mjs`（157 条） | ≈ 1.9 s | — | 静态 |
 | `tests/audit-prompt-invariants.mjs --self-probe`（5 探针） | ≈ 7.4 s | — | 每个探针 = 一次自我重跑 |
 | `tests/audit-spec-traceability.mjs`（94 条） | ≈ 0.2 s | — | 静态 |
@@ -64,6 +64,10 @@ node tests/run-tests.mjs --json              # 机器可读汇总（含 skippedN
 | `tests/v2-path-escape.test.mjs` | ≈ 2.2 s | — | 验证日志路径逃逸（7 条） |
 | `tests/v2-tool-cap.test.mjs` | ≈ 0.4 s | — | 能力表来自真实组合、*MaxToolCalls 只是提示（10 条） |
 
+> 上表 `≈629 s / ≈208 s` 两列是 **56 项（38 套件）** 时的实测；第 39 个套件是随 final-paper 功能加入的
+> `tests/v4-final-paper.test.mjs`，本表尚未单独计时，所以全量的绝对时间会比 208 s 略大——数字**形状**
+> （`TOTAL 57 … suites 39 · probes 18`）以 runner 每次运行的输出为准。
+
 单套件耗时（并行时的关键路径按此排序）：
 
 | 套件 | 耗时 | 备注 |

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "dsh-vibe-math",
-  "description": "Multi-agent mathematical problem-solving & verification frameworks for DeepSeek Harness — FOUR agent presets in one install: vibe-math-v2 (probability-driven, classic: qs.json + Propos knowledge base + explorer→solver→review/debate verdict), vibe-math-v3 (THIRD-generation, classic: paper-style Markdown knowledge base with Problems/Progress/Propos/Methods/Verified + planner-agent scheduling that decides the next N actions + universal theory/method invention library + agents write their own Markdown directly via a per-file write lock), and vibe-math-v4 (FOURTH-generation: persistent self-organizing resident subagents that message & meet to decide all tasks, verify only by unanimous consensus, /compact at a context threshold, and stop only when all agree the problem is solved), and vibe-math-v5 (FIFTH-generation research institute: an academician as the organizational centre who decomposes and ASSIGNS work and chairs meetings; permanent researchers who hold the vote and may hire/fire their own temp workers; temp workers with no vote; a group chat and meetings; a durable per-recipient mailbox; a compare-and-set task DAG; and a boolean m-vote consensus rule where an object enters Verified/ only when at least m voting members agree AND every one of them returns exactly 1 or exactly 0). Installing this bundle auto-installs all four presets (v1 was removed at v2.0.0).",
-  "version": "2.5.0",
+  "description": "Multi-agent mathematical problem-solving & verification frameworks for DeepSeek Harness — FOUR agent presets in one install: vibe-math-v2 (probability-driven, classic: qs.json + Propos knowledge base + explorer→solver→review/debate verdict), vibe-math-v3 (THIRD-generation, classic: paper-style Markdown knowledge base with Problems/Progress/Propos/Methods/Verified + planner-agent scheduling that decides the next N actions + universal theory/method invention library + agents write their own Markdown directly via a per-file write lock), and vibe-math-v4 (FOURTH-generation: persistent self-organizing resident subagents that message & meet to decide all tasks, verify only by unanimous consensus, /compact at a context threshold, and stop only when all agree the problem is solved), and vibe-math-v5 (FIFTH-generation research institute: an academician as the organizational centre who decomposes and ASSIGNS work and chairs meetings; permanent researchers who hold the vote and may hire/fire their own temp workers; temp workers with no vote; a group chat and meetings; a durable per-recipient mailbox; a compare-and-set task DAG; and a boolean m-vote consensus rule where an object enters Verified/ only when at least m voting members agree AND every one of them returns exactly 1 or exactly 0). Installing this bundle auto-installs all four presets (v1 was removed at v2.0.0). Each preset also writes a FINAL PAPER by default when its run concludes — and does so BEFORE the run is marked complete: on v4/v5 the team co-authors its parts, merges them, cross-reviews, and the editor named by paperEditor finalises; the output is Paper/<run or institute id>/{paper.md,paper.tex,paper.pdf when a LaTeX engine is detected,paper.meta.json,paper.log.md} (params finalPaper/paperFormat/paperLanguage/paperCompilePdf/paperLatexCommand, plus paperEditor on v4/v5; see docs/final-paper.md).",
+  "version": "2.6.0",
   "type": "module",
   "engines": {
     "node": "^22.19.0 || >=24.0.0",
@@ -34,6 +34,7 @@
     "cordis.patch.yml",
     "docs/架构图.md",
     "docs/formal-verification.md",
+    "docs/final-paper.md",
     "docs/generate_framework_diagram_v2.py",
     "docs/generate_framework_diagram_v3.py",
     "docs/generate_framework_diagram_v4.mjs",
@@ -96,6 +97,7 @@
     "docs/release-notes/RELEASE-NOTES-2.3.16.md",
     "docs/release-notes/RELEASE-NOTES-2.4.0.md",
     "docs/release-notes/RELEASE-NOTES-2.4.1.md",
+    "docs/release-notes/RELEASE-NOTES-2.6.0.md",
     "docs/release-notes/RELEASE-NOTES-2.5.0.md",
     "docs/release-notes/RELEASE-NOTES-2.3.2.md",
     "docs/release-notes/RELEASE-NOTES-2.3.3.md",
@@ -149,7 +151,7 @@
     },
     "minVersion": "0.1.2-alpha.4",
     "testedVersion": "0.2.0-rc.2",
-    "compatNote": "v2.5.0 深度审计修复版本（提示词措辞与工具描述有增量修正；四套预设的数据格式、目录结构与参数默认值未变，DSH 支撑窗口不变）。① 修复会造成实际损害的缺陷：v5 状态文件不再被空研究所覆盖（按路径锁存、读取未成功即拒绝写入）、v2 验证日志文件名归一化堵住路径逃逸、v4 邮箱投递失败不再让调度器永久停摆。② 修复目标功能不成立的缺陷：v4 真实 /compact 首次生效、v2 判断命题链补种解法入口后真正收口、v3 空闲不再每 tick 空转派生 planner（实测 41→13 并首次出现正常终止）、v3 项目锁改定时器续租、v3 最少 2 票才裁决。③ 语义与表述对齐：v4 验证契约明确「只有 1/0 是绝对真/假，其间为概率，非全体一致取平均写进布尔估计」；v5 m-vote 表述与计票代码一致并删除「更易收敛」的不实说法；四套同级（v2/v3 经典、v4/v5 实验）贯通到 npm 描述、选择器、人设与设计文档；v5 调用者授权收口（isOffice 不再把空身份当所办，生命周期命令仅接受可证明的会话根，新增错误码 V5_NOT_OFFICE 已登记）。④ 安装器与守卫：形态判定改为能力差（npm 包级证据：0.1.6-alpha.2 无 @deepseek-ai/dsh-agent-preset，0.1.7-rc.2 才有），声明行补宿主 group 标记，preset 文件原子写、删除前备份、读不出的文件不再无备份删除，状态文件幂等；run-tests 现收集全部测试与探针（此前两套红探针被静默漏收），修掉一处测试抖动（v4 会议用例）与一处语料非确定性（v3）。本版实测：门禁 56 项全绿（38 套件 + 18 探针）、audit-persona-surface 196 项、并在真实 DSH 0.2.0-rc.2 上验证四个预设注册（order 20–23）且零激活警告、组合行线上不写旧的 .agent-presets 目录。",
+    "compatNote": "v2.6.0 最终论文版本（四套预设新增「最终论文」功能；既有参数默认值、数据格式与目录结构未变，DSH 支撑窗口不变，无需迁移）。① 新功能（默认开启）：收口时自动撰写最终论文，产出 Paper/<id>/{paper.md,paper.tex,paper.pdf,paper.meta.json,paper.log.md}；参数 finalPaper（默认 true）、paperFormat=both|md|tex（默认 both）、paperLanguage=zh|en（默认 zh）、paperCompilePdf（默认 true）、paperLatexCommand（默认空 = 自动探测），v4/v5 另有 paperEditor；手动 /vN paper [lang=] [format=] [editor=] [force] 只覆盖本次，finalPaper=false 仅关闭自动触发。② 时序：论文阶段发生在 run 被标记完成之前（v4 拦截一致性停止票、v5 拦截 checkSolved 判定；v2/v3 在调度器停止前派发写手），论文定稿或降级结束后才真正完成。③ 团队合写（v4/v5）：各自写 → 合并 → 交叉互审 → 定稿代表定稿 → 全体「可交付」一致认同；v4 paperEditor 默认 office（可选 resident:<id>，指定成员已离职则降级并记明），v5 自动默认 academician，office 仅手动 /v5 paper 且必须先与全所交流（至少 1 条所办消息）并至少召开一次会议，否则 V5_PAPER_CONSULT_REQUIRED。④ 编译与降级：LaTeX 检测顺序 xelatex→latexmk→pdflatex→lualatex→tectonic（中文优先），两遍 nonstopmode，失败先修复（重跑 / 换引擎 / 最小模板）再降级为 tex+md 并告警，绝不覆盖已有 paper.pdf；paperFormat=md 时跳过编译且不告警。⑤ 随本功能修掉：v5 自动触发的 paper 心跳停摆、v2/v3 写手卡死可被心跳回收与 force 重派、v4 可空过的判定断言、v3 随包语料的时间戳非确定性；契约文档 docs/final-paper.md 入库并随包发布，双语 README 补齐参数与用法。本版实测：门禁 57 项全绿（39 套件 + 18 探针），并在真实 DSH 0.2.0-rc.2 上验证四个预设注册（order 20–23）、零激活警告、组合行线上不写旧的 .agent-presets 目录。",
     "compatibility": {
       "dshReleases": {
         "0.1.2-alpha.4": "compatible",