Seed-Prover combines deep and broad formal reasoning and proves five IMO 2025 problems

The system iteratively refines whole proofs from Lean feedback and uses test-time strategies for deep and broad search.

What happened?

Seed-Prover reported proving 78.1% of formalized past IMO problems, more than half of PutnamBench, and five of six IMO 2025 problems with Seed-Prover plus Seed-Geometry.

Why it matters

It pushed formal Olympiad proving from benchmark progress toward near-complete coverage of a live IMO problem set.

Significance rationale: Seed-Prover showed that Lean-verified long-reasoning systems could scale to most of a contemporary IMO and substantially raise PutnamBench performance.

Historical context

Primary evidence

Evidence & status at a glance

Significance
H2 · Field milestone
Verification
Paper released
Evidence level
E3 · Inspectable artifact
AI role
AI primary, human verified
Formal assurance
Machine-checked

How should these labels be read? →

Contribution & verification

Mathematical novelty
Benchmark result
Interface
Formal, Symbolic
AI contribution
The prover iteratively revises whole Lean proofs and a dedicated geometry engine handles formal geometry.
Human contribution
Researchers designed inference strategies, training and the geometry engine and evaluated formal proofs.
Verification
Paper released · E3 · Inspectable artifact · Machine-checked
Last reviewed
2026-09-11

Technical details

Systems
seed-prover, seed-geometry, lean
Organizations
bytedance-seed
Problems
imo-2025, putnambench, minif2f
Methods
whole-proof-reasoning, lean-feedback, test-time-search

Sources & artifacts

Verification history

  • 2025-07-31 · Paper released · E2 · Research record

    The Chronicle records this event from the cited primary or official source; the status does not imply independent Chronicle verification.

  • 2026-09-11 · Paper released · E3 · Inspectable artifact

    This corpus refresh added a directly inspectable public artifact maintained by the authors or releasing organization. E3 records artifact-level inspectability and does not imply independent verification of the underlying scientific or mathematical claim.

Corrections

No corrections recorded.