It is amazing to see UI-Vision becoming a go-to benchmark for the community. This is exactly what we hoped for when we created the benchmark. So proud of the work we have done!

Internship ServiceNow Research to build the next generation of computer use agents that are safe and secure from malicious attacks. Focus on intervention strategies, defenses to make agents robust against unsafe behavior.. Apply here: bit.ly/3V3mmTg

Highly recommend this for those wanting to work on agents. Had a great experience working with Spandana Gella and team.

🚨Reasoning LLMs are e̵f̵f̵e̵c̵t̵i̵v̵e̵ ̵y̵e̵t̵ inefficient! Large language models (LLMs) now solve multi-step problems by emitting extended chains of thought. During the process, they often re-derive the same intermediate steps across problems, inflating token usage and

See you in San Diego 🚀 #NeurIPS2025

Our work (arxiv.org/abs/2509.13237) can be seen as one instantiation of the paradigm proposed by Andrej Karpathy here. The behavior handbook is a repository of problem solving strategies which we show can be reused to get better and more efficient reasoning in the future. 🧵 for more

NO verifiers. NO Tools. Qwen3-4B-Instruct can match DeepSeek-R1 and o3-mini (high) with ONLY test-time scaling. Presenting Recursive Self-Aggregation (RSA) — the strongest test-time scaling method I know of! Then we use aggregation-aware RL to push further!! 📈📈 🧵below!