← Back to People
Achieved the highest score on ARC-AGI v1 (79.6%) and v2 (29.4%) using multi-agent collaboration with evolutionary test-time compute.
1
Mentions
1
Articles
1
Outlets
Associated Topics
Companies Covered
Coverage Patterns
How media typically covers Jeremy Berman
Article Types
- research1
Media Angles
- technical1
Narrative Framing
- breakthrough1
Associated AI Models
Grok-41Deepseek R11
Articles
Most recent firstMentioned In
Referenced in coverage
How I Got the Highest Score on ARC-AGI Again Swapping Python for English
Jeremy Berman SubstackresearchpositiveSep 17, 2025
Jeremy Berman achieved a new state-of-the-art score of 79.6% on ARC-AGI v1 and 29.4% on v2 using an evolutionary test-time compute architecture with English instructions instead of Python, achieving 25× greater efficiency than OpenAI's o3.
“Achieved the highest score on ARC-AGI v1 (79.6%) and v2 (29.4%) using multi-agent collaboration with evolutionary test-time compute.”
AI ResearchFoundation ModelsLLMsReinforcement Learning