菜单
Alternative paper search backends — OpenAlex (~250M works, no auth, strong PDF extraction) and CORE (~49M full-text, BODY search). Use when hf-papers can't find a paper, when Semantic Scholar is rate-limited, when you need PDFs of older/long-tail papers, or when you need to search the actual body text of papers (not just titles/abstracts).
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Discover GitHub repositories, find example scripts within them, and read source files with line-range support and automatic Jupyter notebook conversion. Use BEFORE writing any ML training, fine-tuning, or inference code — your knowledge of library APIs is outdated and example scripts show the current patterns.
Inspect a HuggingFace dataset in one call — status, configs/splits, schema, sample rows, parquet file listing, and detailed messages-column analysis (roles, keys, tool-calls). REQUIRED before submitting any training job to verify column format matches the training method.
Explore and fetch HuggingFace library documentation (transformers, trl, datasets, peft, accelerate, trackio, vllm, optimum, courses, etc.). Use when you need current trainer arguments, config parameters, or API reference that your internal knowledge may have wrong.
Submit and manage HuggingFace compute jobs — run training / fine-tuning / inference / batch work on HF cloud infrastructure in Python or Docker mode. Supports one-shot and recurring (cron-style) jobs.
Discover ML research papers, read their contents by section, trace citations, find linked datasets/models — combining HuggingFace Papers, arXiv, and Semantic Scholar. Use whenever you need to research a method, find training recipes with published results, crawl a citation graph, or locate the datasets a paper used. This is your starting point for any literature-first workflow.
File and git-like operations on HuggingFace repositories (model / dataset / space). Covers list / read / upload / delete of files, branches, tags, pull requests, and repo management (create / update). Use to inspect repos, ship artifacts, and propose changes via PRs.
| name | paper-search-alt |
| description | Alternative paper search backends — OpenAlex (~250M works, no auth, strong PDF extraction) and CORE (~49M full-text, BODY search). Use when hf-papers can't find a paper, when Semantic Scholar is rate-limited, when you need PDFs of older/long-tail papers, or when you need to search the actual body text of papers (not just titles/abstracts). |
Two scripts, each wrapping a different open catalog:
scripts/openalex.py — OpenAlex-backed search + PDF download/extract.scripts/core_search.py — CORE-backed metadata + full-text body search.This skill is the fallback / broad-catalog companion to hf-papers. You should almost always start with hf-papers (it has HF's curated ML-paper metadata, arXiv section parsing, and Semantic Scholar citation graphs). Come here when:
snippet_search / recommend / filtered search).| Need | Best tool |
|---|---|
| ML-specific trending / curated / AI summaries | hf-papers --op trending / search / paper_details |
| Parsed paper sections (methodology, experiments) | hf-papers --op read_paper --section N |
| Citation graph, TL;DRs, influential citations | hf-papers --op citation_graph / recommend / snippet_search |
| Broad metadata search (incl. non-ML, older papers) | openalex.py --query ... |
| Downloading PDFs + extracting text | openalex.py --pdf --save-dir ... |
| Full-text body search (phrase inside paper) | core_search.py full-text --query ... |
| Fetching a specific paper by CORE ID | core_search.py get --id ... |
OPENALEX_MAILTO (optional) — your email, gets you into OpenAlex's "polite pool" with higher rate limits. Otherwise a generic mailto is used and public limits apply (still fairly generous).CORE_API_KEY (required for core_search.py) — free, request at https://core.ac.uk/services/api. Without it the script exits with a clear error.# 1. Try hf-papers first (rich, curated, ML-tuned)
./.venv/bin/python scripts/papers.py --op search --query "sparse mixture of experts quantization"
# 2. If the paper isn't there (or results are weak), go broad with OpenAlex
./.venv/bin/python scripts/openalex.py --query "sparse mixture of experts quantization" --limit 10
# 3. If an interesting result has a PDF but no arXiv HTML, grab full text:
./.venv/bin/python scripts/openalex.py --query "MoE expert pruning 2024" \
--pdf --save-dir ./texts
# 4. If you need to find a specific method/dataset mentioned inside a paper body:
./.venv/bin/python scripts/core_search.py full-text \
--query "adaptive scales per expert routing"
openalex.py referenceSingle-command script — one invocation per search.
openalex.py --query Q [--limit N] [--include-closed] [--pdf] [--save-dir DIR] [--json]
--query, -q (required): search query.--limit, -l: results (default 10, max 200).--include-closed: include non-OA papers (default: OA-only).--pdf, -p: download the best-OA PDF for each result and extract text with pypdf. Best-effort — scanned PDFs or paywalled OA stubs fall back to the abstract and a warning is printed to stderr.--save-dir DIR: write each result as DIR/<safe_id>.txt (with full text if --pdf, else abstract).--json: emit {"ok", "query", "count", "papers": [...], "saved": [...]}.ID extraction: if the paper's landing page URL contains an arXiv ID, that's used as id. Otherwise it's the OpenAlex suffix (e.g. W2141902134).
core_search.py referenceThree subcommands. All require CORE_API_KEY in the environment.
core_search.py search --query Q [--limit N] [--save-dir DIR] [--with-full-text] [--json]
core_search.py full-text --query Q [--limit N] [--save-dir DIR] [--json]
core_search.py get --id ID [--save-dir DIR] [--with-full-text] [--json]
search: metadata (title/abstract) search. Default for general lookups.full-text: BODY search — restricted to papers that have full text indexed (_exists_:fullText). Returns body excerpts in the output. Use when hunting a specific phrase.get: fetch a single paper by CORE ID. Useful to pull full text after finding an interesting ID via search or full-text.--with-full-text (save-dir only): write full body text into the saved .txt file instead of just the abstract. On search, only works if the result happens to include fullText; on get, always works if the paper has it.full-text is for.full-text returns nothing, try search (metadata-only) or fall back to OpenAlex.hf-papers --op citation_graph.--pdf follows best_oa_location.pdf_url. If that's paywalled or doesn't exist, no text is extracted. Many papers have OA copies on institutional repositories that OpenAlex indexes.hf-papers — the primary research entry point. Start here.research-subagent — if you spawn a research sub-agent, add scripts/openalex.py and scripts/core_search.py to its allowed tools when the crawl risks exhausting Semantic Scholar's rate limit or when the domain is non-ML.