PeopleofSuperintelligence

Researcher

Lala Li

Lala Li is a machine learning researcher who made a core contribution to Imagen, the text-to-image diffusion model from Google Research's Brain Team. In 2019, she coauthored a study of optimizers and batch sizes with Roger Grosse and others, and in 2020 and 2021 she worked with Ting Chen on embedding compression and contrastive losses. Pix2Seq, written with Chen, Saurabh Saxena, David Fleet and Geoffrey Hinton in 2021, cast object detection as a language modeling task. The Imagen paper received a NeurIPS 2022 Outstanding Paper award. Her later work used diffusion models for panoptic segmentation in 2023 and for 4D novel view synthesis in 2025, when she coauthored the ICLR paper 4DiM at Google DeepMind.

Timeline

  1. 2019

    Optimizers and critical batch size

    In 2019, she coauthors Which Algorithmic Choices Matter at Which Batch Sizes? with Guodong Zhang, Roger Grosse and other coauthors for NeurIPS 2019. The paper shows that preconditioned optimizers such as Adam and K-FAC reach much larger critical batch sizes than stochastic gradient descent with momentum.

  2. 2020–2021

    Embeddings and contrastive learning

    In 2020, she coauthors Differentiable Product Quantization with Ting Chen and Yizhou Sun, a framework that compresses embedding layers and appears at ICML 2020. In 2021, she coauthors Intriguing Properties of Contrastive Losses with Ting Chen and Calvin Luo for NeurIPS 2021.

  3. 2021–2022

    Pix2Seq object detection

    In 2021, she coauthors Pix2Seq with Ting Chen, Saurabh Saxena, David Fleet and Geoffrey Hinton. The ICLR 2022 paper casts object detection as a language modeling task, with bounding boxes and class labels written as sequences of discrete tokens.

  4. May 2022

    Co-authoring Imagen

    In May 2022, she makes a core contribution to Imagen, a text-to-image diffusion model from Google Research's Brain Team. The model encodes text with a pretrained language model such as T5 and reaches a state-of-the-art FID score of 7.27 on COCO without training on COCO.

  5. 2022

    NeurIPS 2022 Outstanding Paper

    She is a coauthor of the Imagen paper, Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding, which receives a NeurIPS 2022 Outstanding Paper award.

  6. 2022–2023

    Sequence and diffusion for vision

    In 2022, she coauthors A Unified Sequence Interface for Vision Tasks at NeurIPS, which trains one model on object detection, instance segmentation, keypoint detection and image captioning. In 2023, she coauthors A Generalist Framework for Panoptic Segmentation of Images and Videos at ICCV, which uses a diffusion model to model panoptic masks.

  7. May 2023

    The FIT transformer

    In May 2023, she and Ting Chen post FIT: Far-reaching Interleaved Transformers, which splits tokens into groups and alternates local and global transformer layers. The architecture can serve as an encoder, a diffusion decoder or an autoregressive decoder.

  8. 2025

    4DiM at Google DeepMind

    In 2025, she coauthors Controlling Space and Time with Diffusion Models at Google DeepMind, with Daniel Watson, Saurabh Saxena, Andrea Tagliasacchi and David Fleet. The ICLR 2025 paper presents 4DiM, a cascaded diffusion model for 4D novel view synthesis from one or more images.

Career

  1. 2025Google DeepMind
  2. 2022Google Brain

From the biography.

  1. Google had the most people shaping AI every year from 2014 to 2025Includes Google. 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.
  2. Google has the highest combined h-index among AI organizationsIncludes Google. 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.
  3. 20% of people at companies who shape AI also hold a university role, led by Google and Stanford UniversityIncludes Google. We estimate that 539 of the 6,766 people with a known current organization (8%) hold both a university role and a company role, including 20% of people at companies and 13% of people at universities.
  4. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Google. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.

Colleagues

People named in this biography, and people with dated positions at the same organization in the same years.

Sources and credits

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

  1. Which Algorithmic Choices Matter at Which Batch Sizes? Insights From a Noisy Quadratic Model arxiv.org. Checked 21 October 2026.
  2. Lala Li - Google Research research.google. Checked 21 October 2026.
  3. Differentiable Product Quantization for End-to-End Embedding Compression arxiv.org. Checked 21 October 2026.
  4. Intriguing Properties of Contrastive Losses arxiv.org. Checked 21 October 2026.
  5. Pix2seq: A Language Modeling Framework for Object Detection arxiv.org. Checked 21 October 2026.
  6. Imagen: Text-to-Image Diffusion Models imagen.research.google. Checked 21 October 2026.
  7. Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding arxiv.org. Checked 21 October 2026.
  8. NeurIPS 2022 Awards neurips.cc. Checked 21 October 2026.
  9. A Unified Sequence Interface for Vision Tasks arxiv.org. Checked 21 October 2026.
  10. A Generalist Framework for Panoptic Segmentation of Images and Videos arxiv.org. Checked 21 October 2026.
  11. FIT: Far-reaching Interleaved Transformers arxiv.org. Checked 21 October 2026.
  12. Controlling Space and Time with Diffusion Models | ICLR Proceedings proceedings.iclr.cc. Checked 21 October 2026.
  13. Controlling Space and Time with Diffusion Models arxiv.org. Checked 21 October 2026.