这是首个在三大顶级机器学习会议上公开、可复现的AI×生物学论文权威索引，召回率高达95.9%，且标签经过人工审核。
AI × Biology Papers Corpus
The definitive, machine-readable index of AI × Biology papers accepted at ICLR, ICML and NeurIPS.
Browse · Figures · Tags · Data · Load · Methods
Every AI × Biology paper accepted at ICLR / ICML / NeurIPS, from NeurIPS 2010 to ICLR 2026, in one place. 3,722 records, all with abstract, authors and PDF link.
Two-stage keyword filter with 95.9 % recall against a 315-paper hand-curated truth set. Ten topical tags (protein, molecule, single_cell, …) assigned per paper.
Ships as structured JSON + CSV plus 33 foldable markdown files you can browse on GitHub. Reproducible: the entire tree regenerates from one JSON.
git clone https://github.com/biotender/ai-bio-conference-papers.git
cd ai-bio-conference-papers
# Browse in your editor — every paper is a foldable card
open papers/ICLR/2026.md
# Or load the data
python -c "import json; papers = json.load(open('data/papers_master.json')); print(len(papers), 'papers')"
Stacked bars by venue, one bar per year. NeurIPS dominates the early record (2010–2017, ICLR/ICML absent from the corpus); ICLR takes over from 2020 onwards. The 2026 bar reaches 1,235 papers even without NeurIPS 2026, which had not yet run at snapshot time. The AlphaFold2 reference line marks the 2021 inflection point.
SVG (vector, editable):
figures/corpus_overview.svg
Per-year paper counts for the seven densest tags. protein climbs from ~2 papers/year (2018) to 294 in 2026 — a textbook post-AlphaFold2 explosion. single_cell lifts late (25 → 125 across 2020–2026). genomics roughly doubles between 2024 → 2025 (59 → 148). medical_imaging and clinical grow more steadily.
SVG (vector, editable):
figures/topic_trajectories.svg
Both figures are regenerated directly from
data/papers_master.json; the plotting code lives in the companionbiotender-ai-bio-corpusrepo.
3,722 AI × Biology papers from ICLR / ICML / NeurIPS (2010–2026) — browsable by venue & year
根据分类、Topic 和编程语言匹配的相似项目。
Harness Engineering is a methodology for improving coding agent outputs by carefully crafting the environment around them—providing curated context, tools, and executable constraints that encode an organization’s nonfunctional requirements and cumulative lessons.
A 28.9 million parameter language model runs on an $8 ESP32-S3 microcontroller entirely on-device, generating simple stories at about 9.5 tokens per second.
mimic captures traffic from any iOS or web app and automatically generates a Python client library that lets you call the app's API like a regular library.