AutoResearchClaw：全自动AI研究流水线，从主题到完整学术论文（含LaTeX格式）

autoresearchclaw-autonomous-research by aradotso/trending-skills

437 周安装量

10 GitHub Stars

GitHub

安装命令

npx skills add https://github.com/aradotso/trending-skills --skill autoresearchclaw-autonomous-research

AI/机器学习内容创作自动化

🇨🇳中文介绍

AutoResearchClaw — 自主研究流水线

技能来自 ara.so — Daily 2026 Skills 集合。

AutoResearchClaw 是一个完全自主的 23 阶段研究流水线，它接收一个自然语言主题，并生成一篇完整的学术论文：包含真实的 arXiv/Semantic Scholar 引用、沙盒实验、统计分析、多智能体同行评审以及符合会议要求的 LaTeX（NeurIPS/ICML/ICLR）。无幻觉引用。无需人工干预。

安装

# 克隆并安装
git clone https://github.com/aiming-lab/AutoResearchClaw.git
cd AutoResearchClaw
python3 -m venv .venv && source .venv/bin/activate
pip install -e .

# 验证 CLI 是否可用
researchclaw --help

要求： Python 3.11+

配置

cp config.researchclaw.example.yaml config.arc.yaml

最小配置 (`config.arc.yaml`)

project:
  name: "my-research"

research:
  topic: "在此处填写您的研究主题"

llm:
  provider: "openai"
  base_url: "https://api.openai.com/v1"
  api_key_env: "OPENAI_API_KEY"
  primary_model: "gpt-4o"
  fallback_models: ["gpt-4o-mini"]

experiment:
  mode: "sandbox"
  sandbox:
    python_path: ".venv/bin/python"



export OPENAI_API_KEY="$YOUR_OPENAI_KEY"

广告位招租

在这里展示您的产品或服务

触达数万 AI 开发者，精准高效

联系我们

OpenRouter 配置 (200+ 模型)

llm:
  provider: "openrouter"
  api_key_env: "OPENROUTER_API_KEY"
  primary_model: "anthropic/claude-3.5-sonnet"
  fallback_models:
    - "google/gemini-pro-1.5"
    - "meta-llama/llama-3.1-70b-instruct"



export OPENROUTER_API_KEY="$YOUR_OPENROUTER_KEY"

ACP (Agent Client Protocol) — 无需 API 密钥

llm:
  provider: "acp"
  acp:
    agent: "claude"   # 或：codex, gemini, opencode, kimi
    cwd: "."

代理 CLI（例如 claude）会处理自身的身份验证。

OpenClaw 桥接器（可选高级功能）

openclaw_bridge:
  use_cron: true              # 计划研究运行
  use_message: true           # 进度通知
  use_memory: true            # 跨会话知识持久化
  use_sessions_spawn: true    # 并行子会话
  use_web_fetch: true         # 文献综述中的实时网络搜索
  use_browser: false          # 基于浏览器的论文收集

# 基本运行 — 完全自主，无需提示
researchclaw run --topic "您的研究想法" --auto-approve

# 使用显式配置文件运行
researchclaw run --config config.arc.yaml --topic "专家混合路由效率" --auto-approve

# 使用配置文件中定义的主题运行（省略 --topic 标志）
researchclaw run --config config.arc.yaml --auto-approve

# 交互模式 — 在关卡阶段暂停等待批准
researchclaw run --config config.arc.yaml --topic "您的主题"

# 检查流水线状态 / 恢复运行
researchclaw status --run-id rc-20260315-120000-abc123

# 列出过去的运行
researchclaw list

关卡阶段 (5, 9, 20) 在交互模式下会暂停等待人工批准。传递 --auto-approve 以跳过所有关卡。

from researchclaw.pipeline import Runner
from researchclaw.config import load_config

# 加载配置并运行
config = load_config("config.arc.yaml")
config.research.topic = "用于长上下文 LLM 的高效注意力机制"
config.auto_approve = True

runner = Runner(config)
result = runner.run()

# 访问输出
print(result.artifact_dir)          # artifacts/rc-YYYYMMDD-HHMMSS-<hash>/
print(result.deliverables_dir)      # .../deliverables/
print(result.paper_draft_path)      # .../deliverables/paper_draft.md
print(result.latex_path)            # .../deliverables/paper.tex
print(result.bibtex_path)           # .../deliverables/references.bib
print(result.verification_report)  # .../deliverables/verification_report.json



# 仅运行特定阶段
from researchclaw.pipeline import Runner, StageRange

runner = Runner(config)
result = runner.run(stages=StageRange(start="LITERATURE_COLLECT", end="KNOWLEDGE_EXTRACT"))



# 运行后访问知识库
from researchclaw.knowledge import KnowledgeBase

kb = KnowledgeBase.load(result.artifact_dir)
findings = kb.get("findings")
literature = kb.get("literature")
decisions = kb.get("decisions")

运行后，所有输出都位于 artifacts/rc-YYYYMMDD-HHMMSS-<hash>/：

artifacts/rc-20260315-120000-abc123/
├── deliverables/
│   ├── paper_draft.md          # 完整的学术论文（Markdown）
│   ├── paper.tex               # 符合会议要求的 LaTeX
│   ├── references.bib          # 真实的 BibTeX — 自动修剪为内联引用
│   ├── verification_report.json # 4 层引用完整性报告
│   └── reviews.md              # 多智能体同行评审
├── experiment_runs/
│   ├── run_001/
│   │   ├── code/               # 生成的实验代码
│   │   ├── results.json        # 结构化指标
│   │   └── sandbox_output.txt  # 执行日志
├── charts/
│   └── *.png                   # 自动生成的比较图表
├── evolution/
│   └── lessons.json            # 用于未来运行的自我学习经验
└── knowledge_base/
    ├── decisions.json
    ├── experiments.json
    ├── findings.json
    ├── literature.json
    ├── questions.json
    └── reviews.json

流水线阶段参考

阶段	阶段编号	名称	备注
A	1	TOPIC_INIT	解析并界定研究主题
A	2	PROBLEM_DECOMPOSE	分解为子问题
B	3	SEARCH_STRATEGY	构建搜索查询
B	4	LITERATURE_COLLECT	对 arXiv + Semantic Scholar 进行真实的 API 调用
B	5	LITERATURE_SCREEN	关卡 — 批准/拒绝文献
B	6	KNOWLEDGE_EXTRACT	提取结构化知识
C	7	SYNTHESIS	综合发现
C	8	HYPOTHESIS_GEN	多智能体辩论以形成假设
D	9	EXPERIMENT_DESIGN	关卡 — 批准/拒绝设计
D	10	CODE_GENERATION	生成实验代码
D	11	RESOURCE_PLANNING	GPU/MPS/CPU 自动检测
E	12	EXPERIMENT_RUN	沙盒执行
E	13	ITERATIVE_REFINE	失败时自我修复
F	14	RESULT_ANALYSIS	多智能体分析
F	15	RESEARCH_DECISION	继续 / 优化 / 转向
G	16	PAPER_OUTLINE	构建论文结构
G	17	PAPER_DRAFT	撰写完整论文
G	18	PEER_REVIEW	证据一致性检查
G	19	PAPER_REVISION	纳入评审反馈
H	20	QUALITY_GATE	关卡 — 最终批准
H	21	KNOWLEDGE_ARCHIVE	将经验保存到知识库
H	22	EXPORT_PUBLISH	输出 LaTeX + BibTeX
H	23	CITATION_VERIFY	4 层反幻觉检查

模式：关于某个主题的快速论文

export OPENAI_API_KEY="$OPENAI_API_KEY"
researchclaw run \
  --topic "用于蛋白质结构预测的自监督学习" \
  --auto-approve

模式：使用完整配置进行可复现运行

# config.arc.yaml
project:
  name: "protein-ssl-research"

research:
  topic: "用于蛋白质结构预测的自监督学习"

llm:
  provider: "openai"
  api_key_env: "OPENAI_API_KEY"
  primary_model: "gpt-4o"
  fallback_models: ["gpt-4o-mini"]

experiment:
  mode: "sandbox"
  sandbox:
    python_path: ".venv/bin/python"
  max_iterations: 3
  timeout_seconds: 300



researchclaw run --config config.arc.yaml --auto-approve

模式：通过 OpenRouter 使用 Claude 以获得最佳推理能力

export OPENROUTER_API_KEY="$OPENROUTER_API_KEY"

cat > config.arc.yaml << 'EOF'
project:
  name: "my-research"
llm:
  provider: "openrouter"
  api_key_env: "OPENROUTER_API_KEY"
  primary_model: "anthropic/claude-3.5-sonnet"
  fallback_models: ["google/gemini-pro-1.5"]
experiment:
  mode: "sandbox"
  sandbox:
    python_path: ".venv/bin/python"
EOF

researchclaw run --config config.arc.yaml \
  --topic "用于 Transformer 推理的高效 KV 缓存压缩" \
  --auto-approve

模式：在运行失败后恢复

# 列出运行以查找运行 ID
researchclaw list

# 从最后完成的阶段恢复
researchclaw run --resume rc-20260315-120000-abc123

模式：编程式批量研究

import asyncio
from researchclaw.pipeline import Runner
from researchclaw.config import load_config

topics = [
    "在有限硬件上的 LoRA 微调",
    "用于 LLM 推理的推测解码",
    "Flash attention 变体比较",
]

config = load_config("config.arc.yaml")
config.auto_approve = True

for topic in topics:
    config.research.topic = topic
    runner = Runner(config)
    result = runner.run()
    print(f"[{topic}] → {result.deliverables_dir}")

模式：OpenClaw 单行命令（如果使用 OpenClaw 代理）

将仓库 URL 分享给 OpenClaw，然后说：
"研究专家混合路由效率"

OpenClaw 会自动读取 RESEARCHCLAW_AGENTS.md，克隆、安装、配置并运行整个流水线。

# 导航到交付物目录
cd artifacts/rc-*/deliverables/

# 编译（需要 LaTeX 发行版）
pdflatex paper.tex
bibtex paper
pdflatex paper.tex
pdflatex paper.tex

# 或者直接将 paper.tex + references.bib 上传到 Overleaf

`researchclaw: command not found`

# 确保 venv 已激活且包已安装
source .venv/bin/activate
pip install -e .
which researchclaw

# 验证环境变量是否已设置
echo $OPENAI_API_KEY
# 应该打印您的密钥（非空）

# 为当前会话显式设置它
export OPENAI_API_KEY="sk-..."

流水线在第 13 阶段（ITERATIVE_REFINE）会自我修复。如果持续失败：

# 在配置中增加超时和迭代次数
experiment:
  max_iterations: 5
  timeout_seconds: 600
  sandbox:
    python_path: ".venv/bin/python"

第 23 阶段（CITATION_VERIFY）运行 4 层检查。如果引用被修剪：

这是预期行为 — 虚假引用会被自动移除
检查 verification_report.json 以了解哪些引用被拒绝及其原因

PIVOT 循环无限运行

第 15 阶段（RESEARCH_DECISION）可能会多次转向。要限制迭代次数：

research:
  max_pivots: 2
  max_refines: 3

# 检查缺失的包
pdflatex paper.tex 2>&1 | grep "File.*not found"

# 安装缺失的包（TeX Live）
tlmgr install <package-name>

实验期间内存不足

# 在配置中强制使用 CPU 模式
experiment:
  sandbox:
    device: "cpu"
    max_memory_gb: 4

转向/优化循环：第 15 阶段自主决定继续、优化（调整参数）或转向（新的假设方向）。所有工件都进行版本控制。
多智能体辩论：第 8、14、18 阶段使用结构化的多视角辩论 — 不是单一的 LLM 处理。
自我学习：每次运行都会提取经验，并带有 30 天的时间衰减。未来在类似主题上的运行会受益于过去的错误。
哨兵监视器：后台监视器检测结果中的 NaN/Inf，检查论文-证据一致性，评估引用相关性，并在整个运行过程中防范捏造。
4 层引用验证：arXiv 查找 → CrossRef 查找 → DataCite 查找 → LLM 相关性评分。一个引用必须通过所有层才能保留。

🇺🇸English

AutoResearchClaw — Autonomous Research Pipeline

Skill by ara.so — Daily 2026 Skills collection.

AutoResearchClaw is a fully autonomous 23-stage research pipeline that takes a natural language topic and produces a complete academic paper: real arXiv/Semantic Scholar citations, sandboxed experiments, statistical analysis, multi-agent peer review, and conference-ready LaTeX (NeurIPS/ICML/ICLR). No hallucinated references. No human babysitting.

Installation

# Clone and install
git clone https://github.com/aiming-lab/AutoResearchClaw.git
cd AutoResearchClaw
python3 -m venv .venv && source .venv/bin/activate
pip install -e .

# Verify CLI is available
researchclaw --help

Requirements: Python 3.11+

Configuration

cp config.researchclaw.example.yaml config.arc.yaml

Minimum config (`config.arc.yaml`)

project:
  name: "my-research"

research:
  topic: "Your research topic here"

llm:
  provider: "openai"
  base_url: "https://api.openai.com/v1"
  api_key_env: "OPENAI_API_KEY"
  primary_model: "gpt-4o"
  fallback_models: ["gpt-4o-mini"]

experiment:
  mode: "sandbox"
  sandbox:
    python_path: ".venv/bin/python"



export OPENAI_API_KEY="$YOUR_OPENAI_KEY"

OpenRouter config (200+ models)

llm:
  provider: "openrouter"
  api_key_env: "OPENROUTER_API_KEY"
  primary_model: "anthropic/claude-3.5-sonnet"
  fallback_models:
    - "google/gemini-pro-1.5"
    - "meta-llama/llama-3.1-70b-instruct"



export OPENROUTER_API_KEY="$YOUR_OPENROUTER_KEY"

ACP (Agent Client Protocol) — no API key needed

llm:
  provider: "acp"
  acp:
    agent: "claude"   # or: codex, gemini, opencode, kimi
    cwd: "."

The agent CLI (e.g. claude) handles its own authentication.

OpenClaw bridge (optional advanced capabilities)

openclaw_bridge:
  use_cron: true              # Scheduled research runs
  use_message: true           # Progress notifications
  use_memory: true            # Cross-session knowledge persistence
  use_sessions_spawn: true    # Parallel sub-sessions
  use_web_fetch: true         # Live web search in literature review
  use_browser: false          # Browser-based paper collection

Key CLI Commands

# Basic run — fully autonomous, no prompts
researchclaw run --topic "Your research idea" --auto-approve

# Run with explicit config file
researchclaw run --config config.arc.yaml --topic "Mixture-of-experts routing efficiency" --auto-approve

# Run with topic defined in config (omit --topic flag)
researchclaw run --config config.arc.yaml --auto-approve

# Interactive mode — pauses at gate stages for approval
researchclaw run --config config.arc.yaml --topic "Your topic"

# Check pipeline status / resume a run
researchclaw status --run-id rc-20260315-120000-abc123

# List past runs
researchclaw list

Gate stages (5, 9, 20) pause for human approval in interactive mode. Pass --auto-approve to skip all gates.

Python API

from researchclaw.pipeline import Runner
from researchclaw.config import load_config

# Load config and run
config = load_config("config.arc.yaml")
config.research.topic = "Efficient attention mechanisms for long-context LLMs"
config.auto_approve = True

runner = Runner(config)
result = runner.run()

# Access outputs
print(result.artifact_dir)          # artifacts/rc-YYYYMMDD-HHMMSS-<hash>/
print(result.deliverables_dir)      # .../deliverables/
print(result.paper_draft_path)      # .../deliverables/paper_draft.md
print(result.latex_path)            # .../deliverables/paper.tex
print(result.bibtex_path)           # .../deliverables/references.bib
print(result.verification_report)  # .../deliverables/verification_report.json



# Run specific stages only
from researchclaw.pipeline import Runner, StageRange

runner = Runner(config)
result = runner.run(stages=StageRange(start="LITERATURE_COLLECT", end="KNOWLEDGE_EXTRACT"))



# Access knowledge base after a run
from researchclaw.knowledge import KnowledgeBase

kb = KnowledgeBase.load(result.artifact_dir)
findings = kb.get("findings")
literature = kb.get("literature")
decisions = kb.get("decisions")

Output Structure

After a run, all outputs land in artifacts/rc-YYYYMMDD-HHMMSS-<hash>/:

artifacts/rc-20260315-120000-abc123/
├── deliverables/
│   ├── paper_draft.md          # Full academic paper (Markdown)
│   ├── paper.tex               # Conference-ready LaTeX
│   ├── references.bib          # Real BibTeX — auto-pruned to inline citations
│   ├── verification_report.json # 4-layer citation integrity report
│   └── reviews.md              # Multi-agent peer review
├── experiment_runs/
│   ├── run_001/
│   │   ├── code/               # Generated experiment code
│   │   ├── results.json        # Structured metrics
│   │   └── sandbox_output.txt  # Execution logs
├── charts/
│   └── *.png                   # Auto-generated comparison charts
├── evolution/
│   └── lessons.json            # Self-learning lessons for future runs
└── knowledge_base/
    ├── decisions.json
    ├── experiments.json
    ├── findings.json
    ├── literature.json
    ├── questions.json
    └── reviews.json

Pipeline Stages Reference

Phase	Stage #	Name	Notes
A	1	TOPIC_INIT	Parse and scope research topic
A	2	PROBLEM_DECOMPOSE	Break into sub-problems
B	3	SEARCH_STRATEGY	Build search queries
B	4	LITERATURE_COLLECT	Real API calls to arXiv + Semantic Scholar
B	5	LITERATURE_SCREEN	Gate — approve/reject literature
B	6	KNOWLEDGE_EXTRACT	Extract structured knowledge
C	7	SYNTHESIS	Synthesize findings

Common Patterns

Pattern: Quick paper on a topic

export OPENAI_API_KEY="$OPENAI_API_KEY"
researchclaw run \
  --topic "Self-supervised learning for protein structure prediction" \
  --auto-approve

Pattern: Reproducible run with full config

# config.arc.yaml
project:
  name: "protein-ssl-research"

research:
  topic: "Self-supervised learning for protein structure prediction"

llm:
  provider: "openai"
  api_key_env: "OPENAI_API_KEY"
  primary_model: "gpt-4o"
  fallback_models: ["gpt-4o-mini"]

experiment:
  mode: "sandbox"
  sandbox:
    python_path: ".venv/bin/python"
  max_iterations: 3
  timeout_seconds: 300



researchclaw run --config config.arc.yaml --auto-approve

Pattern: Use Claude via OpenRouter for best reasoning

export OPENROUTER_API_KEY="$OPENROUTER_API_KEY"

cat > config.arc.yaml << 'EOF'
project:
  name: "my-research"
llm:
  provider: "openrouter"
  api_key_env: "OPENROUTER_API_KEY"
  primary_model: "anthropic/claude-3.5-sonnet"
  fallback_models: ["google/gemini-pro-1.5"]
experiment:
  mode: "sandbox"
  sandbox:
    python_path: ".venv/bin/python"
EOF

researchclaw run --config config.arc.yaml \
  --topic "Efficient KV cache compression for transformer inference" \
  --auto-approve

Pattern: Resume after a failed run

# List runs to find the run ID
researchclaw list

# Resume from last completed stage
researchclaw run --resume rc-20260315-120000-abc123

Pattern: Programmatic batch research

import asyncio
from researchclaw.pipeline import Runner
from researchclaw.config import load_config

topics = [
    "LoRA fine-tuning on limited hardware",
    "Speculative decoding for LLM inference",
    "Flash attention variants comparison",
]

config = load_config("config.arc.yaml")
config.auto_approve = True

for topic in topics:
    config.research.topic = topic
    runner = Runner(config)
    result = runner.run()
    print(f"[{topic}] → {result.deliverables_dir}")

Pattern: OpenClaw one-liner (if using OpenClaw agent)

Share the repo URL with OpenClaw, then say:
"Research mixture-of-experts routing efficiency"

OpenClaw auto-reads RESEARCHCLAW_AGENTS.md, clones, installs, configures, and runs the full pipeline.

Compile the LaTeX Output

# Navigate to deliverables
cd artifacts/rc-*/deliverables/

# Compile (requires a LaTeX distribution)
pdflatex paper.tex
bibtex paper
pdflatex paper.tex
pdflatex paper.tex

# Or upload paper.tex + references.bib directly to Overleaf

Troubleshooting

`researchclaw: command not found`

# Make sure the venv is active and package is installed
source .venv/bin/activate
pip install -e .
which researchclaw

API key errors

# Verify env var is set
echo $OPENAI_API_KEY
# Should print your key (not empty)

# Set it explicitly for the session
export OPENAI_API_KEY="sk-..."

Experiment sandbox failures

The pipeline self-heals at Stage 13 (ITERATIVE_REFINE). If it keeps failing:

# Increase timeout and iterations in config
experiment:
  max_iterations: 5
  timeout_seconds: 600
  sandbox:
    python_path: ".venv/bin/python"

Citation hallucination warnings

Stage 23 (CITATION_VERIFY) runs a 4-layer check. If references are pruned:

This is expected behaviour — fake citations are removed automatically
Check verification_report.json for details on which citations were rejected and why

PIVOT loop running indefinitely

Stage 15 (RESEARCH_DECISION) may pivot multiple times. To cap iterations:

research:
  max_pivots: 2
  max_refines: 3

LaTeX compilation errors

# Check for missing packages
pdflatex paper.tex 2>&1 | grep "File.*not found"

# Install missing packages (TeX Live)
tlmgr install <package-name>

Out of memory during experiments

# Force CPU mode in config
experiment:
  sandbox:
    device: "cpu"
    max_memory_gb: 4

Key Concepts

PIVOT/REFINE Loop : Stage 15 autonomously decides PROCEED, REFINE (tweak params), or PIVOT (new hypothesis direction). All artifacts are versioned.
Multi-Agent Debate : Stages 8, 14, 18 use structured multi-perspective debate — not a single LLM pass.
Self-Learning : Each run extracts lessons with 30-day time decay. Future runs on similar topics benefit from past mistakes.
Sentinel Watchdog : Background monitor detects NaN/Inf in results, checks paper-evidence consistency, scores citation relevance, and guards against fabrication throughout the run.
4-Layer Citation Verification : arXiv lookup → CrossRef lookup → DataCite lookup → LLM relevance scoring. A citation must pass all layers to survive.

Weekly Installs

375

Repository

aradotso/trending-skills

GitHub Stars

First Seen

8 days ago

Security Audits

Gen Agent Trust HubFail SocketPass SnykWarn

Installed on

opencode371

gemini-cli371

github-copilot371

codex371

amp371

cline371

AI Elements：基于shadcn/ui的AI原生应用组件库，快速构建对话界面

54,900 周安装

AutoResearchClaw：全自动AI研究流水线，从主题到完整学术论文（含LaTeX格式）

🇨🇳中文介绍

AutoResearchClaw — 自主研究流水线

安装

配置

最小配置 (config.arc.yaml)

相关 Skills

OpenRouter 配置 (200+ 模型)

ACP (Agent Client Protocol) — 无需 API 密钥

OpenClaw 桥接器（可选高级功能）

关键 CLI 命令

Python API

输出结构

流水线阶段参考

常见模式

模式：关于某个主题的快速论文

模式：使用完整配置进行可复现运行

模式：通过 OpenRouter 使用 Claude 以获得最佳推理能力

模式：在运行失败后恢复

模式：编程式批量研究

模式：OpenClaw 单行命令（如果使用 OpenClaw 代理）

编译 LaTeX 输出

故障排除

researchclaw: command not found

API 密钥错误

实验沙盒失败

引用幻觉警告

PIVOT 循环无限运行

LaTeX 编译错误

实验期间内存不足

关键概念

🇺🇸English

AutoResearchClaw — Autonomous Research Pipeline

Installation

Configuration

Minimum config (config.arc.yaml)

OpenRouter config (200+ models)

ACP (Agent Client Protocol) — no API key needed

OpenClaw bridge (optional advanced capabilities)

Key CLI Commands

Python API

Output Structure

Pipeline Stages Reference

Common Patterns

Pattern: Quick paper on a topic

Pattern: Reproducible run with full config

Pattern: Use Claude via OpenRouter for best reasoning

Pattern: Resume after a failed run

Pattern: Programmatic batch research

Pattern: OpenClaw one-liner (if using OpenClaw agent)

Compile the LaTeX Output

Troubleshooting

researchclaw: command not found

API key errors

Experiment sandbox failures

Citation hallucination warnings

PIVOT loop running indefinitely

LaTeX compilation errors

Out of memory during experiments

Key Concepts

最新 Skills

最小配置 (`config.arc.yaml`)

`researchclaw: command not found`

Minimum config (`config.arc.yaml`)

`researchclaw: command not found`