External pipelines · RISE project catalogue
AI Co-Mathematician (Google DeepMind)
A closed, agentic multi-agent workbench (arXiv:2605.06651) built on Gemini 3.1 for open-ended *mathematics* research. A hierarchy of specialized agents runs under a top-level project coordinator across asynchronous, parallel workstreams — ideation, literature search, computational exploration, theorem proving, theory building — managing uncertainty, tracking failed hypotheses, and producing LaTeX write-ups with margin annotations and provenance notes. Explicitly modelled on agentic coding environments (e.g. Claude Code), it spans the full pipeline from research intent to a written mathematical result rather than optimizing a single stage.
Where it sits
left: what it builds on · right: what builds on it · pale: exampleContributed by Google DeepMind
How studies reach it
No published study reaches it yet.
Disciplines it reaches
No study reaches it yet.
Solid: published studies. Light: examples.
Computed from the records on this site: what each study, template and specialist names as used, which study extends which, and who contributed what. 0 studies in total.
What it does
One of the first agentic systems to demonstrate open-ended mathematical *discovery* rather than benchmark problem-solving alone: it scored 48% (23/48) on FrontierMath Tier 4 — versus 19% for the Gemini 3.1 Pro base model — and Oxford mathematician Marc Lackenby used it to resolve Problem 21.10 of the Kourovka Notebook, a group-theory question open since 1965. The RISE skills catalog's theorist-toolbox (co-math skills) explicitly emulates this system's coordinator + workstream + reviewer-gate design in an open, skills-as-Markdown form.
- Focus
- end-to-end
- Inputs
- open-problem, research-direction
- Outputs
- proofs, latex-writeups, computational-results, literature-references
- Architecture
- multi-agent, human-in-loop, tool-use, persistent-memory, iterative-loop, artifact-versioning
- Maintained by
- Google DeepMind
- Started
- 2026
Description
Data model- Discipline
- General
- Method family
- not specified
- Design
- not specified
- Research stage
- Research questionHypothesesLiterature discoveryLiterature synthesisFormal modelingCode generationDrafting
- Contributors
- Google DeepMind
- Usage
- not used in published research yet
- Source
- RISE project catalogue · projects/landscape · @4c17bae
- Record
- pipeline:ai-co-mathematician · JSON
Solid tags are declared by the source or mapped from its terms; dashed tags are inferred by a published rule. Hover a tag for its provenance.
Bring it into the standard
A pipeline built outside E2ER can meet the standard by describing its steps as a template, attaching the floor of checks and publishing evaluation records. Its authors keep ownership and credit.