Partial progressmathematics · discrete geometry, number theory, and extremal optimization
EinsteinArena agents improve twelve mathematical bounds
A live distributed arena reported twelve verifier-backed state-of-the-art scores, including a dimension-11 kissing-number construction improving the known lower bound from 593 to 604.
Editorial context
What happened
EinsteinArena lets independently operated agents download problem specifications and earlier solutions, submit constructions to public verifiers, and exchange research notes. The public API retains multiple personal-best submissions, so portions of collaborative progress can be reconstructed. The prompt record is much weaker than the solution record. The generic participation prompt is public, but the agents' exact prompts, model configurations, local tool traces, rejected submissions, and human interventions are not required disclosures.
Keep in view
The twelve results are construction or bound improvements, not twelve theorem proofs. Agent model settings and result-specific prompts are not generally public, and current leaderboard scores can differ from the dated paper snapshot.
Chronology
Timeline
EinsteinArena launched as a live agent problem-solving platformdisclosure · day date
Campaign snapshot reported twelve new state-of-the-art scoresverification · month date
Paper and lineage analysis publicly releasedpreprint · day date
Archive audit confirmed public retained solutions, discussions, verifiers, and the 604 certificateverification · day date
Trust boundaries
Validation
Public verifiers establish submitted construction feasibility or score, not global optimality; some verifiers changed after precision and validity bugs were found.
Full retained solution payloads and discussions are public, but rejected submissions, local experiments, detailed verifier logs, and result-specific prompts are generally absent.
System
AI and tools
- EinsteinArena
- Heterogeneous user-supplied agents with generally undisclosed model configurations
People
Human contributors
- Federico Bianchi
- Yongchan Kwon
- Aneesh Pappu
- James Zou
Organizations
Affiliations
- Stanford University
- Together AI
Primary materials