面向量化社会科学、计算社会科学与 AI for Social Science 的可复现研究项目模板
A starter template for empirical research projects done with AI coding agents: shared memory in the repo, traceable results, built-in checks. Pre-release.
GitHub template for reproducible research repos — two-layer Python 3.10+/uv monorepo splitting generic infrastructure from per-project code, with pytest coverage gates (60% infra, 90% projects), a 17-stage build pipeline, markdown-to-PDF rendering via pandoc/LaTeX, and 24 canonical exemplars.
An AI-friendly research project template for reproducible, multi-dataset experiments and agent-assisted research workflows.
A starter template for Python-based research projects with a consistent folder layout, linting/type-checking config, editor settings, and a LaTex template for writing up results.
Template for composing multi-track active inference manuscripts — closed-form analytics, pymdp simulations, Lean 4 proofs, GNN/ontology notation — bound to shared sections by a typed sheaf engine with six machine-checked composition laws and a symbol gluing certificate.
Systematic literature review pipeline: dispatches ten retrieval engines (arXiv, OpenAlex, Semantic Scholar, Crossref, PubMed) with graceful degradation, deduplicates by DOI/arXiv/OpenAlex ID or title hash, then runs TF-IDF embeddings, NMF topic models and citation networks into an auto-written bibliometric manuscript.
Manuscript-refinement exemplar that models drafting as gold metallurgy: five stages (ore, smelting, assaying, cupellation, certification) monotonically raise a paper's measured purity from 37.5% to nine-nines via claim-support checks, citation audits, seeded SHA-256 token injection and 12 NetworkX/matplotlib figures.
Scaffold for publishing a dataset with a machine-readable data contract: JSON descriptors drive schema fingerprints, field-level constraint checks, SHA-256 file verification and quality gates, emitting data-dictionary and provenance-flow figures plus a FAIR release manifest with durable Zenodo identifiers.
Evaluation harness scaffold for self-improvement agents (SIA) — runs a Meta to Target to Feedback loop across generations, scores runs on the mini_classify task with split public/private data to catch overfitting, and replays recorded_generations fixtures for deterministic reruns. Python, uv, pytest; swap your agent into src/loop.py.
Testbed for the five coordination mechanisms in the AutoScientists paper on self-organizing agent teams: shared experiment log, dead-end registry (36 redundant re-probes to 0), effect-size ranking, noise-band confirmation, stagnation reorganization. Python, uv, no mocks.
Template that generates whole runnable child projects — src, tests, scripts, docs — from a combinatoric grammar over 5 domains and 8 kernel primitives, with SHA-256 byte-stable output, recompute verification of tree hashes, and meta-tests against green-by-construction test theater.
JAX/Equinox language-model research template adapted from TTT-E2E, with Hydra configuration, distributed training, and checkpointing.
Scaffold for registered reports, preregistrations, and replications — freezes hypotheses, outcomes, exclusion rules, and the analysis plan into a hash-stamped ledger, logs every deviation, and audits claims so confirmatory and exploratory results stay separate. Emits hypothesis-outcome maps, analysis DAGs, deviation timelines, permutation tests.
Research-infrastructure scaffold wiring three resource pools (fonds, rules, tools) into one self-validating Python pipeline: typed manifests with schema-versioned receipts, fail-closed jsonschema validation, soft vs strong rule checks, tool entrypoint discovery, and a generated manuscript with matplotlib figures.
Pitch decks as reproducible build artifacts: one YAML content source renders six deliverables — short/medium/long decks in PDF (ReportLab) and PPTX (python-pptx) — with byte-identical rebuilds, citation enforcement, a cliche linter, live token checks against repo test counts and DOIs, and per-slide QR deep links.
A human-led Research OS for rigorous, reproducible science with ChatGPT, Codex, and Git.
Academic project-page template fork reserved for Arcanea AI research pages, demos, and experiment writeups.
Personal paper-centric empirical research project template
Superseded first-generation research framework preserved as a public record. New projects should use Collaboration Station Research 2.0.
To associate your repository with the research-template topic, visit your repo's landing page and select "manage topics."