Thanks @fchollet
. Indeed, the experiments used a large Pass@k which allowed us to focus on evaluating the search capability of AB-MCTS, rather than the official evaluation criteria based on k=2. We also used tasks in the public eval. Hopefully we’ll get down to Pass@2 someday! 🙂
RESEARCH
-

AB-MCTS Search Capability Evaluation with High Pass@k Metrics
By
–
-

Sakana AI ARC-AGI-2 Results: Clarification on Test-Time Search Method
By
–
Impressive results from Sakana AI on ARC-AGI-2 with a new method for test-time-search and ensembling!
— François Chollet (@fchollet) 1 juillet 2025
Please be mindful when reporting on this result: the method doesn't actually get 30% on ARC-AGI-2 public eval, since ARC-AGI scores should be computed by checking whether you… https://t.co/By3lUGdrrnImpressive results from Sakana AI on ARC-AGI-2 with a new method for test-time-search and ensembling! Please be mindful when reporting on this result: the method doesn't actually get 30% on ARC-AGI-2 public eval, since ARC-AGI scores should be computed by checking whether you
-

Microsoft advances medical superintelligence with latest AI breakthroughs
By
–
Top stories in AI today: – Microsoft’s ‘step towards medical superintelligence’
– Baidu open-sources ERNIE 4.5
– Automate YouTube content creation with AI
– Chai Discovery's AI designs working antibodies
– 4 new AI tools & 4 job opportunities Read more: https://
therundown.ai/p/microsoft-st
eps-towards-medical-superintelligence
… -

Multi-LLM AB-MCTS Combination Outperforms on ARC-AGI-2 Benchmark
By
–
The Multi-LLM AB-MCTS combination of o4-mini + Gemini-2.5-Pro + DeepSeek-R1-0528, current frontier AI models, achieves strong performance on the ARC-AGI-2 benchmark, outperforming individual models by a large margin. Implementation of AB-MCTS on GitHub: https://
github.com/SakanaAI/treeq
uest
… -
AI Transforming Chemistry: Drug Discovery Breakthroughs at Lindau
By
–
Happening soon! Join the livestream for Lindau Nobel Laureate Meeting on "Artificial Intelligence in Chemistry: How Is AI Changing the Game?" We will discuss the transformative impact of AI technologies in chemistry including groundbreaking applications in drug discovery,
-
Copyright Verdicts on AI Training and Fair Use Analysis
By
–
"essentially" doing a lot of heavy lifting. All the verdicts fell under Copyright law that's why they did the Fair Use analyses. Thus, it must have been expressive each time. Training often does capture expressions; verdict mentions this as "compressed copies". Keep up!
-
Training Data Copyright: Legal Analysis of AI Model Development
By
–
You speak in the future as if it didn't happen; deluding yourself? It did happen multiple times already and each time analysis was conducted under Copyright law. The uses before training were found to be expressive because all the work was used.
-
Copyright Fair Use Analysis in AI Model Training
By
–
You're confusing everything again, the reproductions before training and possible downstream uses. The downstream uses can potentially be non-expressive. If it was entirely non-expressive, it wouldn't fall under Copyright and there wouldn't have been Fair Use analysis needed.
-

Meta Launches Superintelligence Labs with Top AI Talent
By
–
Mark Zuckerberg officially introduced Meta Superintelligence Labs to employees The group has 11 hires from OpenAI, Google, Anthropic, Seasame and will be led by Alexandr Wang (Chief AI Officer) and Nat Friedman
-
Chai-2 AI Discovers Antibodies for 26 Diseases in Breakthrough
By
–
OpenAI-backed Chai Discovery dropped Chai-2, an AI that can design working antibodies
— Rowan Cheung (@rowancheung) 1 juillet 2025
It designed antibodies against 52 diseases and found successful treatments for half of them by testing just 20 candidates each
A big breakthrough in drug discovery!pic.twitter.com/2sov4rVHfVOpenAI-backed Chai Discovery dropped Chai-2, an AI that can design working antibodies It designed antibodies against 52 diseases and found successful treatments for half of them by testing just 20 candidates each A big breakthrough in drug discovery!