Auto-claude-code-research-in-sleep

# Auto-claude-code-research-in-sleep (ARIS ⚔️🌙)

💡 *Use ARIS workflows directly in Claude Code / Cursor / Trae, or get the full experience with the standalone CLI — enjoy any way you like!*

🔥 [**ARIS-Code CLI — 独立安装版**](docs/ARIS-Code-README_CN.md) · [English](docs/ARIS-Code-README_EN.md) | [⬇️ Download](https://github.com/wanshuiyin/Auto-claude-code-research-in-sleep/releases/latest)

<img src="docs/aris-code-banner.png" width="600" alt="ARIS-Code CLI">

![ARIS Logo](docs/aris_logo.svg)

![Hero](docs/hero_combined.svg)

[中文版 README](README_CN.md) | English

![Score Progression](docs/auto_review_score_curve.png)

> 🌙 **Let Claude Code do research while you sleep.** Wake up to find your paper scored, weaknesses identified, experiments run, and narrative rewritten — autonomously.
>
> 🪶 **Radically lightweight — zero dependencies, zero lock-in.** The entire system is plain Markdown files. No framework to learn, no database to maintain, no Docker to configure, no daemon to babysit. Every skill is a single `SKILL.md` readable by any LLM — swap Claude Code for [Codex CLI](skills/skills-codex/), [OpenClaw](docs/OPENCLAW_ADAPTATION.md), [Cursor](docs/CURSOR_ADAPTATION.md), [Trae](docs/TRAE_ARIS_RUNBOOK_EN.md), [Antigravity](docs/ANTIGRAVITY_ADAPTATION.md), Windsurf, or your own agent and the workflows still work. Fork it, rewrite it, adapt it to your stack.
>
> *💡 ARIS is a methodology, not a platform. What matters is the research workflow — take it wherever you go. 🌱*

[![Featured on PaperWeekly](https://img.shields.io/badge/Featured%20on-PaperWeekly-red?style=flat)](https://mp.weixin.qq.com/s/tDniVryVGjDkkkWl-5sTkQ) · [![PaperWeekly — MiniMax-M2.7](https://img.shields.io/badge/PaperWeekly-MiniMax--M2.7-red?style=flat)](https://mp.weixin.qq.com/s/KLFU74lAL2FAIc9K6i1Kqg) · [![Featured in awesome-agent-skills](https://img.shields.io/badge/Featured%20in-awesome--agent--skills-blue?style=flat&logo=github)](https://github.com/VoltAgent/awesome-agent-skills) · [![AI Digital Crew - Project of the Day](https://img.shields.io/badge/AI%20Digital%20Crew-Project%20of%20the%20Day%20(2026.03.14)-orange?style=flat)](https://aidigitalcrew.com) · [💬 Join Community](#-community) · [![Cite](https://img.shields.io/badge/📖_Cite_Us-BibTeX-green?style=flat)](#-citation)

Custom [Claude Code](https://docs.anthropic.com/en/docs/claude-code) skills for autonomous ML research workflows. These skills orchestrate **cross-model collaboration** — Claude Code drives the research while an external LLM (via [Codex MCP](https://github.com/openai/codex)) acts as a critical reviewer. 🔀 **Also supports [alternative model combinations](#-alternative-model-combinations) (Kimi, LongCat, DeepSeek, etc.) — no Claude or OpenAI API required.** For example, [MiniMax-M2.7 + GLM-5 or GLM-5 + MiniMax-M2.7](docs/MiniMax-GLM-Configuration.md). 🤖 **[Codex CLI native](skills/skills-codex/)** — full skill set also available for OpenAI Codex. 🖱️ **[Cursor](docs/CURSOR_ADAPTATION.md)** — works in Cursor too. 🖥️ **[Trae](docs/TRAE_ARIS_RUNBOOK_EN.md)** — ByteDance AI IDE. 🚀 **[Antigravity](docs/ANTIGRAVITY_ADAPTATION.md)** — Google's agent-first IDE. 🆓 **[Free tier via ModelScope](docs/MODELSCOPE_GUIDE.md) — zero cost, zero lock-in.**

> 💭 **Why not self-play with a single model?** Using Claude Code subagents or agent teams for both execution and review is technically possible, but tends to fall into **local minima** — the same model reviewing its own patterns creates blind spots.
>
> *Think of it like adversarial vs. stochastic bandits: a single model self-reviewing is the stochastic case (predictable reward noise), while cross-model review is adversarial (the reviewer actively probes weaknesses the executor didn't anticipate) — and adversarial bandits are fundamentally harder to game.*
>
> 💭 **Why two models, not more?** Two is the minimum needed to break self-play blind spots, and 2-player games converge to Nash equilibrium far more efficiently than n-player ones. Adding more reviewers increases API cost and coordination overhead with diminishing returns — the biggest gain is going from 1→2, not 2→4.
>
> Claude Code's strength is fast, fluid execution; Codex (GPT-5.4 xhigh) is slower but more deliberate and rigorous in critique. These complementary styles — **speed × rigor** — produce better outcomes than either model talking to itself.

## 🎯 More Than Just a Prompt

> These are full pipelines — you can also use each workflow independently. Already have an idea? Skip to Workflow 1.5. Have results? Jump to Workflow 3. Got reviews? Jump to Workflow 4. See [Quick Start](#-quick-start) for all commands and [Workflows](#-workflows) for the full breakdown.

**Basic mode** — give ARIS a research direction, it handles everything:

```
/research-pipeline "factorized gap in discrete diffusion LMs"
```

**🔥 Targeted mode** — got a paper you want to improve? Give ARIS the paper + the code:

```
/research-pipeline "improve method X" — ref paper: https://arxiv.org/abs/2406.04329, base repo: https://github.com/org/project
```

ARIS reads the paper → finds its weaknesses → clones the codebase → generates ideas that specifically fix *those* weaknesses with *that* code → runs experiments → writes your paper. Like telling a research assistant: *"read this paper, use this repo, find what's missing, and fix it."*

> Mix and match: `ref paper` only = "what can be improved?", `base repo` only = "what can I build with this code?", both = "improve *this* paper using *this* code."

**🔥 Rebuttal mode** — reviews just dropped? Don't panic. ARIS reads every concern, builds a strategy, and drafts a rebuttal that's grounded, structured, and under the character limit:

```
/rebuttal "paper/ + reviews" — venue: ICML, character limit: 5000
```

| Parameter | Default | What it does |
|-----------|---------|-------------|
| `venue` | `ICML` | Target venue (ICML/NeurIPS/ICLR/CVPR/ACL/AAAI/ACM) |
| `character limit` | — | **Required.** Hard character limit for rebuttal text |
| `quick mode` | `false` | Stop after parsing + strategy (Phase 0-3). See what reviewers want before drafting |
| `auto experiment` | `false` | Auto-run supplementary experiments via `/experiment-bridge` when reviewers ask for new evidence |
| `max stress test rounds` | `1` | How many times GPT-5.4 xhigh stress-tests the draft |
| `max followup rounds` | `3` | Per-reviewer follow-up round limit |

Three safety gates — rebuttal will NOT finalize if any fails:
- 🔒 **No fabrication** — every claim maps to paper/review/user-confirmed result
- 🔒 **No overpromise** — every promise is user-approved
- 🔒 **Full coverage** — every reviewer concern is tracked

Two outputs: `PASTE_READY.txt` (exact char count, paste to venue) + `REBUTTAL_DRAFT_rich.md` (extended version for manual editing).

**After acceptance** — your paper is in, now prepare the presentation:

```
/paper-slides "paper/"     # → Beamer PDF + PPTX + speaker notes + Q&A prep
/paper-poster "paper/"     # → A0/A1 poster PDF + editable PPTX + SVG
```

> *💡 From idea to paper to podium — one toolchain. 🌱*

## 🏆 Papers Accepted with ARIS

| Paper | Score | Venue | Author | Stack |
|-------|:-----:|-------|--------|-------|
| CS Paper | **8/10** "clear accept" | CS Conference | [@DefanXue](https://github.com/DefanXue) & [@Monglitay](https://github.com/Monglitay) | Claude Code + GPT-5.4 |
| AAAI Paper | **7/10** "good paper, accept" | AAAI 2026 Main Technical | [@xinbo820-web](https://github.com/xinbo820-web) | Pure Codex CLI |

> 🎉 Built entirely with ARIS — from idea to acceptance. [Full details + reviewer screenshots →](#-community-showcase--papers-built-with-aris)

## 📢 What's New

- **2026-03-30** — ![NEW](https://img.shields.io/badge/NEW-red?style=flat-square) 🔥 **Auto-debug & exhaust-before-surrender** — experiment-bridge now auto-diagnoses failures (OOM, import, path, CUDA, NaN) and retries up to 3× before giving up. auto-review-loop must try 2+ solution paths before conceding any reviewer

[truncated…]