Research infrastructure Preprint

Lacuna: A Research Map for Machine Learning

A living map that turns machine-learning papers into linked summaries, research directions, and testable proposals.

Martin Weiss, Miles Q. Li, Alejandro H. Artiles, Yacine Mkhinini, Chris Pal, Hugo Larochelle, Nasim Rahaman

Read on arXiv Explore Lacuna

The question

Why we’re working on this

Finding a useful paper often leaves you with more questions: what does it connect to, which claims hold up, and where could you contribute?

We built Lacuna to make those connections easier to follow. Researchers and agents can move from papers to ideas, research directions, and candidate questions, then trace each step back to its sources.

Approach

How it works

The pipeline reconciles scholarly records, creates readable paper summaries, extracts granular concept elements, and clusters those elements into research directions. Every generated layer preserves links back to the records and papers that support it.

Papers connect to concept elements, research directions, and candidate questions, with links back to their sources.
Follow a question through the map, then return to the papers that support it.
  1. 01

    Papers

    Reconcile scholarly records and produce readable, source-linked summaries.

  2. 02

    Concepts

    Extract granular methods, findings, and observations from each paper.

  3. 03

    Directions

    Cluster recurring concepts into coherent threads across the literature.

  4. 04

    Proposals

    Surface underexplored questions and hypotheses for researcher review.

What to keep in mind

  • Generated summaries, directions, and proposals remain research aids rather than substitutes for the primary literature.
  • Coverage and map quality depend on the scholarly sources, identity reconciliation, and models used to build each release.

Evidence

Results at a glance

0.538

LitSearch Recall@10

OpenScholar v3: 0.424

99

Expert-reference hits

GPT-Researcher: 72

7.82 / 10

RACE report quality

GPT-Researcher: 5.24

ReportBench-ML · 25 survey tasks
SystemCitation F1Reference hitsRACE / 10
Lacuna Deep Research0.052997.82
GPT-Researcher0.039725.24
STORM0.015212.90
LangChain Open Deep Research0.007137.42

ReportBench-ML covers 25 survey tasks. Reference hits measure overlap with expert bibliographies; RACE is a model-judged quality score. Systems use different model and retrieval setups. See the paper for the evaluation details.

Source: Lacuna v1, Tables 4–5 and Appendix D.3

Authors

People behind the work

Listed in the paper’s author order.

Martin Weiss

Tiptree staff

Tiptree Systems · Mila · Polytechnique Montréal

Builds tools that connect the literature, research questions, and the people working on them. Also studies peer-review verification and agent decision-making.

Contributed to

View on Lacuna

Miles Q. Li

Tiptree staff

Tiptree Systems · McGill University

Coauthor of Lacuna, connecting machine-learning papers through summaries, concepts, and research directions that people and agents can explore.

Contributed to

View on Lacuna

Alejandro H. Artiles

Tiptree intern

Tiptree Systems · Max Planck Institute for Human Development · Max Planck Institute for Intelligent Systems

First author of The Alien Space of Science. Now studies cognition–action tradeoffs in language agents through TextCraft.

Contributed to

View on Lacuna

Yacine Mkhinini

Tiptree staff

Tiptree Systems

Works on research infrastructure at Tiptree. Coauthor of Lacuna, our map of papers, concepts, and open research questions.

Contributed to

View on Lacuna

Christopher Pal

Scientific advisor

Polytechnique Montréal · Mila · Canada CIFAR AI Chair

Canada CIFAR AI Chair at Polytechnique Montréal. Scientific advisor and coauthor of Lacuna and The Alien Space of Science.

Contributed to

View on Lacuna

Hugo Larochelle

Scientific advisor

Scientific Director, Mila · Université de Montréal · McGill University

Scientific advisor and coauthor of Lacuna and Alien Science. Also collaborates on the behavioural economics of LLM agents.

Contributed to

View on Lacuna

Nasim Rahaman

Tiptree staff

Tiptree Systems

Brings a background in theoretical physics and machine learning to scientific discovery. Coauthor of Lacuna and The Alien Space of Science.

Contributed to

View on Lacuna

Citation

Cite this work

Copy this entry into your .bib file or import it into your reference manager.

Show BibTeX 
@misc{weiss2026lacuna,
  title = {Lacuna: A Research Map for Machine Learning},
  author = {Weiss, Martin and Li, Miles Q. and Artiles, Alejandro H. and Mkhinini, Yacine and Pal, Chris and Larochelle, Hugo and Rahaman, Nasim},
  year = {2026},
  publisher = {arXiv},
  doi = {10.48550/arXiv.2606.26246},
  url = {https://arxiv.org/abs/2606.26246}
}

Full paper

Read the full paper

Open the PDF or preview it here.

Open PDF

Continue exploring

Work with us

Have a question worth working on?

We work with researchers on scientific discovery, agent evaluation, and tools for navigating the literature. Tell us what you’re exploring.

Talk research with us