AI Briefing#mathematical-reasoning
The latest AI and developer news about #mathematical-reasoning, with the original source and a short summary.
The latest AI and developer news about #mathematical-reasoning, with the original source and a short summary.
The method uses DPO on self-generated traces to teach models when to stop reasoning, improving the accuracy-efficiency trade-off across math and code benchmarks.
ARBITRAGE predicts model superiority to reduce mathematical reasoning latency by up to 2x.
OpenAI's unreleased model Astra has produced new results on 10 open math problems.
Using IMO math problems, the reasoning performance of frontier models and open-weight models and the effectiveness of multi-agent usage were compared.
LG AI Research has unveiled 'EXAONE Deep,' an AI model with overwhelming reasoning performance in math, science, and coding.
Google DeepMind's AlphaProof Nexus autonomously solved decades-old math problems at low cost.
GPT-5.4 Pro solved Erdős's primitive set problem with a new proof.
OpenAI participated in the First Proof challenge, which verifies AI's mathematical reasoning ability, and confirmed the likelihood of correct answers for 5 or more out of 10 problems.
OpenAI has unveiled GPT-5.2 Pro and GPT-5.2 Thinking, which maximize science and mathematics research capabilities.
The visual reasoning model **QVQ-Max**, which analyzes visual information and reasons to present solutions, has been released.
LG AI Research has unveiled EXAONE Deep, a reasoning-specialized model with dramatically improved math, science, and coding capabilities.
Qwen has released a new PRM and benchmark called ProcessBench that identifies step-by-step errors in mathematical reasoning.
The Qwen team has released QVQ, an open-weight multimodal model with enhanced visual reasoning capabilities.