FAR pipeline mines 5,245 math papers into 4,717 open conjectures for AI discovery
wellecks · x · 2026-09-11
Sean Welleck presented his group's scalable mathematical discovery work at The Geometry of Machine Learning 2026, alongside the paper "The Problem Is the Problem" (arXiv:2608.16977). Key points:
- Bottleneck: expert problem selection and review, not model reasoning, limit AI-assisted math research.
- New paradigm: humans supply a research direction; the system searches literature corpora for candidate problems.
- FAR cascade: a Find–Attempt–Recommend pipeline, inspired by search and recommender systems, filters artifacts before they reach humans.
- Combinatorics pilot: from 5,245 papers it recovered 6,453 candidate conjectures, kept 4,717 well-posed open ones, and surfaced 598 potential resolutions.
More from Research
- NYU mathematician says OpenAI pushed him to drop co-author's name after private call — eyishazyer · 2026-09-11
- NYU mathematician says OpenAI pushed to scrub collaborator from Navier-Stokes proof announcement — eyishazyer · 2026-09-11
- Arch breakdown: dropping convs for Muon optimizer, 3x3 pixel unshuffle for vision — stochasticchasm · 2026-09-11
- The 'Trickle Test': a new eval measuring whether user models disclose information gradually like humans — gharik · 2026-09-11
- Humans& Releases Persimmon, First Large-Scale Model Simulating How People Talk — niloofar_mire · 2026-09-11
- Yacine: RL can run in real time — almost nobody is seriously trying — yacineMTB · 2026-09-11