PeopleofSuperintelligence

Researcher

Ruochen Xu

Ruochen Xu is an AI researcher whose work runs from cross-lingual language processing to language-model pretraining and embodied AI. He studied electronic engineering at Tsinghua University and completed a PhD at Carnegie Mellon's Language Technologies Institute under Yiming Yang in 2019. At Microsoft from 2019 to 2024, he coauthored G-Eval, a GPT-4 framework for evaluating generated text, in 2023, and Rho-1, a language model that received a Best Paper Runner-up award at NeurIPS 2024. He was a Principal Scientist at Om AI from 2024 to 2026 and coauthored the vision-language reinforcement-learning framework VLM-R1 in 2025. Since 2025, he is an embodied AI researcher at Galbot.

Now at
Galbot
Badges

Timeline

  1. 2011–2019

    Tsinghua, then a CMU PhD

    He studies electronic engineering at Tsinghua University from 2011 to 2015. In 2015, he starts a PhD at Carnegie Mellon's Language Technologies Institute with Yiming Yang, and in 2018 he coauthors "Unsupervised Cross-lingual Transfer of Word Embedding Spaces" with Yang, Naoki Otani and Yuexin Wu at EMNLP. His 2019 dissertation is titled "Learning Cross-language and Cross-style Mappings with Limited Supervision".

  2. 2019–2024

    Researcher at Microsoft

    After graduating in 2019, he joins Microsoft in Redmond as a researcher. From 2020 to 2024, he is a Senior Researcher at Microsoft AI.

  3. 2023

    G-Eval for text evaluation

    In March 2023, he is among the authors of G-Eval, which uses large language models with chain-of-thought prompting and a form-filling paradigm to assess the quality of generated text; with GPT-4 as the backbone it reaches a Spearman correlation of 0.514 with humans on summarization. The paper appears at EMNLP 2023.

  4. 2024

    Rho-1 at NeurIPS 2024

    He is among the authors of "Not All Tokens Are What You Need for Pretraining", which introduces Rho-1, a language model trained with Selective Language Modeling: a reference model scores the training tokens, and the model trains on those with higher scores. The paper is a NeurIPS 2024 oral and receives a Best Paper Runner-up award; after fine-tuning, Rho-1-7B reaches 51.8% on MATH, matching DeepSeekMath with 3% of its pretraining tokens.

  5. 2024–2026

    Principal Scientist at Om AI

    Following his years at Microsoft, he is a Principal Scientist at Om AI from 2024 to 2026.

  6. April 2025

    VLM-R1 for vision-language models

    In April 2025, he coauthors VLM-R1 with Haozhan Shen, Tiancheng Zhao and others, a framework that extends R1-style reinforcement learning to vision-language models. Their RL-trained model generalizes better than supervised fine-tuning.

  7. Since 2025

    Embodied AI at Galbot

    In 2025, he becomes an embodied AI researcher at Galbot, working on systems that connect perception, language, reasoning, and action. His research interests include robotics foundation models, multimodal reasoning, and tool-augmented agents.

Career

  1. Since 2025GalbotEmbodied AI researcher
  2. 2024–2026Om AIPrincipal Scientist
  3. 2019–2024MicrosoftResearcher, Senior Researcher

Education

  1. 2019Carnegie Mellon UniversityPhD
  2. 2015Tsinghua University

From the biography.

  1. Among employers with 30+ full-time cases, OpenAI leads primary technical role H-1B pay at $310,000, and Amazon, Google and Microsoft file 57% of all casesIncludes Microsoft. Across 440,669 certified US H-1B filings from 44 employers (FY2020–FY2026), OpenAI leads primary technical role base pay at $310,000 and xAI leads overall base pay at $290,000.
  2. Google had the most people shaping AI every year from 2014 to 2025Includes Microsoft. Our estimates of which organizations employed the most people shaping AI each year from 1955 to 2025, tracing the shift from university labs to companies.
  3. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Microsoft. We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  4. Google has the highest combined h-index among AI organizationsIncludes Microsoft. We estimate that Google holds the highest combined h-index at 8,016, while universities account for 73% of the deduplicated total across AI organizations.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Ruochen Xu yet. Send us a free-licensed photo.

  1. Ruochen Xu | alphaXiv alphaxiv.org. Checked 22 October 2026.
  2. Ruochen Xu's Website cs.cmu.edu. Checked 22 October 2026.
  3. Unsupervised Cross-lingual Transfer of Word Embedding Spaces - ACL Anthology aclanthology.org. Checked 22 October 2026.
  4. LTI Ph.D. Dissertations: 2021-2017 - Language Technologies Institute - School of Computer Science - Carnegie Mellon University lti-web-prod-vip.andrew.cmu.edu. Checked 22 October 2026.
  5. [2303.16634] G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment arxiv.org. Checked 22 October 2026.
  6. Ruochen Xu - Embodied AI Researcher xrc10.github.io. Checked 22 October 2026.
  7. NeurIPS 2024 Awards neurips.cc. Checked 22 October 2026.
  8. Not All Tokens Are What You Need for Pretraining proceedings.neurips.cc. Checked 22 October 2026.
  9. [2504.07615] VLM-R1: A Stable and Generalizable R1-style Large Vision-Language Model arxiv.org. Checked 22 October 2026.