README.md

metadata

license: cc-by-4.0
language:
  - en
pretty_name: OpenClaw as Language Infrastructure — Survey Corpus
tags:
  - nlp
  - llm-agents
  - systematic-review
  - openclaw
  - moltbook
  - language-infrastructure
size_categories:
  - n<1K
configs:
  - config_name: direct-corpus
    data_files:
      - split: train
        path: direct-corpus.csv
  - config_name: review-inventory
    data_files:
      - split: train
        path: review-inventory.csv
  - config_name: screening-audit
    data_files:
      - split: train
        path: screening-audit.csv

OpenClaw as Language Infrastructure

Public evidence release for “OpenClaw as Language Infrastructure: A Case-Centered Survey of a Public Agent Ecosystem in the Wild.”

Project Page · GitHub · Paper PDF

Snapshot

The corpus is frozen at 2026-03-10 and separates three analytical scopes:

Configuration	Records	Contents
direct-corpus	38	Ecosystem-specific studies with GATE, AERO, evidence-alignment, source-tier, and provenance labels
review-inventory	79	38 direct studies plus 41 contextual sources
screening-audit	85	Full identification, deduplication, eligibility, and inclusion lineage

The screening flow reconciles 85 identified → 83 after duplicate/version merging → 81 full texts assessed → 79 included. The 38 direct studies divide into Grounding 8, Action 9, Transfer 5, and Exchange 16.

Files

• direct-corpus.csv / .json — work-level survey corpus.
• review-inventory.csv / .json — included frozen review inventory.
• screening-audit.csv / .json — reason-coded screening ledger.
• schema.json, review-inventory.schema.json, and screening-audit.schema.json — machine-readable contracts.
• CC-BY-4.0.txt and THIRD_PARTY_NOTICES.md — licensing and source-boundary notices.

Evidence boundary

This release contains bibliographic facts and author-written annotations. It does not redistribute surveyed papers, platform data, or third-party full text. Each record links to a canonical public source. Do not infer autonomous authorship, verified identity, safety, or representativeness from inclusion in the corpus.

Authors

Chaoyue He, Xin Zhou, Di Wang, Hong Xu, Wei Liu, and Chunyan Miao.

Citation

@misc{he2026openclaw,
  title  = {OpenClaw as Language Infrastructure: A Case-Centered Survey of a Public Agent Ecosystem in the Wild},
  author = {He, Chaoyue and Zhou, Xin and Wang, Di and Xu, Hong and Liu, Wei and Miao, Chunyan},
  year   = {2026},
  url    = {https://chaoyue0307.github.io/openclaw-language-infrastructure/}
}

Funding

This research is supported by the RIE2025 Industry Alignment Fund--Industry Collaboration Projects (IAF-ICP) (Award I2301E0026), administered by A*STAR, and by Alibaba Group and NTU Singapore through the Alibaba--NTU Global e-Sustainability CorpLab (ANGEL).

License

Original corpus metadata and annotations are released under CC BY 4.0. Linked third-party works remain under their respective terms.