Alibaba Qwen Releases QVQ-Max: Visual Reasoning Model with Proofs

Alibaba Qwen officially releases QVQ-Max, its first version of a visual reasoning model. The model is capable of not only recognizing images and videos but also analyzing them, conducting reasoning, and offering solutions. In the MathVision benchmark, QVQ-Max demonstrates consistent accuracy improvement as the length of the thought process increases.

Alibaba/Qwen

DeepSeek hints at R2 model and introduces new inference scaling method SPCT

DeepSeek AI has published a paper on inference-time scaling for general reward models, introducing SPCT (Self-Principled Critique Tuning). The company hints at its upcoming R2 model. The research addresses reward sparsity and aims to improve scalability of reward models during inference.

DeepSeek OpenAI