While human expert evaluation remains the gold standard for mathematical proofs, its cost and time intensity limit scalable research. To address this, we built #ProofAutoGrader, an automatic grader for IMO-ProofBench. The autograder leverages Gemini 2.5 Pro, providing it with a
RESEARCH
-

IMO-ProofBench: Evaluating AI Mathematical Reasoning Capabilities
By
–
IMO-ProofBench is our key focus designed to evaluate the ability of AI models in constructing rigorous and valid mathematical arguments. With 60 proof-based problems, the benchmark is divided into two subsets: a basic set covering pre-IMO to IMO-Medium difficulty levels, and an
-
Waterfall Development with LLMs: Inherent Challenges
By
–
Waterfall based development with LLMs have all the challenges with waterfall based dev?
-

IMO-Bench: New AI Model Evaluation Framework for Mathematics
By
–
IMO-Bench consists of three benchmarks that judge models on diverse capabilities: IMO-AnswerBench, a large-scale test on getting the right answer; IMO-ProofBench, a next-level evaluation for proof writing; and IMO-GradingBench, a new benchmarkto enable further progress in
-
AI in Academia vs AI in the World: Different Priorities
By
–
AI in academia: create architectures and algorithms for conference papers. AI in the world: create datasets and evaluations to engineer the models people use.
-
Managing AI Model Deprecation: Costs and Mitigation Strategies
By
–
Even when new AI models bring clear improvements in capabilities, deprecating the older generations comes with downsides. An update on how we’re thinking about these costs, and some of the early steps we’re taking to mitigate them:
-

AI Will Cure Alzheimer’s Disease Before 2050
By
–
Maxime, Toi oublies juste que l’Intelligence Artificielle guérira 100 % des Alzheimers bien avant 2050 Ne vis pas dans le passé !
-
Jamba Reasoning 3B outperforms Qwen 3 4B in speed test
By
–
Our Jamba Reasoning 3B model took on Qwen 3 4B 2507 by @Alibaba_Cloud in the same 60K token task. Jamba Reasoning 3B finished nearly 3× faster.
— AI21 Labs (@AI21Labs) 4 novembre 2025
🎥 Watch how it played out 👇
👉 Download Jamba Reasoning 3B here: https://t.co/W0ajfxd8ll pic.twitter.com/noeZNd6j1VOur Jamba Reasoning 3B model took on Qwen 3 4B 2507 by @Alibaba_Cloud in the same 60K token task. Jamba Reasoning 3B finished nearly 3× faster. Watch how it played out Download Jamba Reasoning 3B here: https://
huggingface.co/collections/ai
21labs/jamba-reasoning-3b
… -

16,800 Academic Papers Reference GPT-4 in 2025
By
–
There are 16,800 papers on Google Scholar that were published in 2025, which reference GPT-4, but not any future versions of GPT, e.g. no GPT-4o, 4.1, 4.5, or 5.
-
Partnerships Unite Technology Investment Policy for Sustainability
By
–
Partnerships can accelerate progress by uniting complementary strengths, allowing technology, investment, and policy to move together toward lasting sustainability.
