Fei Xia (@xf1280) / X

Fei Xia

959 posts

Fei Xia

@xf1280

Ex Research Scientist, TLM at @GoogleDeepMind, ✨♊, Gemini & Robotics, PhD from @StanfordAILab @StanfordSVL, previously @Tsinghua_Uni. #AGI through Embodiment

Earth

Joined June 2015

Fei Xia
@xf1280
Mar 25, 2025
Gemini 2.5 Pro is such a strong coding model. I was able to use it to convert an image into a 3d-printable object, and bring it to live. More in 🧵👇
252K
Fei Xia
@xf1280
Feb 22, 2023
Text-to-image generative models, meet robotics! We present ROSIE: Scaling RObot Learning with Semantically Imagined Experience, where we augment real robotics data with semantically imagined scenarios for downstream manipulation learning. Website: diffusion-rosie.github.io 🧵👇
00:00
198K
Fei Xia
@xf1280
Jun 16, 2023
🤖Excited to share our project where we propose to use rewards represented in code as a flexible interface between LLMs and an optimization-based motion controller. website: language-to-reward.github.io Want to learn more about how we make a robot dog do moonwalk MJ style?🕺🕺
00:00
357K
Fei Xia
@xf1280
Nov 7, 2024
Excited to share a project where we leverage Gemini’s long context window and multimodal understanding in a creative way, led by @JasonMa2020. Earlier this year we @GoogleDeepMind published Gemini 1.5 series, achieving state of the art on multiple image understanding, video
97K
Fei Xia
@xf1280
Feb 12, 2024
We propose PIVOT: Iterative Visual Prompting, a new visual prompting method to elicit knowledge about action / motion / spatial understanding in VLMs. For example, we can ask "How should I make a smiley face with the fruits" and our method will generate arrows to guide you.
00:00
60K
Fei Xia
@xf1280
Dec 4, 2022
GeoGuesser with #chatGPT
Fei Xia
@xf1280
Sep 25, 2025
Excited that we are launching Gemini Robotics 1.5, a pair of models that can work together to solve long horizon tasks. Gemini Robotics-ER 1.5, the high-level reasoning brain, is available for public preview in AIStudio and Gemini API for robotics developers around the world:
Google DeepMind
@GoogleDeepMind
Sep 25, 2025
We’re making robots more capable than ever in the physical world. 🤖 Gemini Robotics 1.5 is a levelled up agentic system that can reason better, plan ahead, use digital tools such as @Google Search, interact with humans and much more. Here’s how it works 🧵
00:00
36K
Fei Xia
@xf1280
Oct 19, 2025
The mission is calling! #Google DeepMind Robotics is looking for talented researchers to join our team and help build the new frontier of robotics intelligence through world modeling.
job-boards.greenhouse.io
DeepMind
91K
Fei Xia
@xf1280
Jan 13, 2023
In 2021 I trained a model-free RL agent that can go into a room through a closed door. It took 50k episodes for the agent to figure out the right strategy. However, we can get such prior out of the box from Foundation Models. Our team explored this direction in 2022 🧵👇
00:00
Karol Hausman
@hausman_k
Jan 12, 2023
Here are a few examples from our work in robotics that leverage bitter lesson 2.0 This is something that I believe we'll see a lot more of (including our own work in 2023)🧵
117K
Fei Xia
@xf1280
May 6, 2025
Gemini 2.5 Pro is here! It is the new workhorse for vibe coding! It codes up a garden bed planner that generates layouts based on plant height, orientation to the sun, optimal spacing, and companion planting. Really helpful for square foot gardening!🌱
00:00
00:32
Demis Hassabis
@demishassabis
May 6, 2025
Very excited to share the best coding model we’ve ever built! Today we’re launching Gemini 2.5 Pro Preview 'I/O edition' with massively improved coding capabilities. Ranks no.1 on LMArena in Coding and no.1 on the WebDev Arena Leaderboard. It’s especially good at building
57K
Fei Xia
@xf1280
Apr 2, 2025
Lately I have been pretty fascinated by the concept of a neural network training another neural network. I probe Gemini with this quest, and found it can generate the code and data to train NN for simple digit recognition akin to the MNIST task. gemini.google.com/share/b4dfb725… more 👇🧵
57K
Fei Xia
@xf1280
Oct 26, 2025
Feeling honored and grateful to be named one of @techreview Innovators Under 35 Asia Pacific this year. It’s inspiring to see 6 out of 35 innovators in this cohort exploring #EmbodiedAI and #Robotics from different directions. I’ve learned so much from their work and
innovatorsunder35.com
Fei XIA | Innovators Under 35
Our annual list of the young dreamers, tinkerers, and innovators that are creating technology’s future.
24K
Fei Xia
@xf1280
Feb 20, 2024
When robots do not work out of the box, we teach them, and they can improve over time. What is interesting is that their "teachability" also improves over time, meaning it takes fewer attempts to shape their behavior and achieve task success. Website: robot-teaching.github.io
00:00
00:40
Jacky Liang
@jackyliang42
Feb 20, 2024
We can teach LLMs to write better robot code through natural language feedback. But can LLMs remember what they were taught and improve their teachability over time? Introducing our latest work, Learning to Learn Faster from Human Feedback with Language Model Predictive Control
29K
Fei Xia
@xf1280
Mar 7, 2023
Super excited this work is out! PaLM-E is a general purpose vision-language model that can do long-horizon robot planning, visual question answering, and complex visual reasoning. Check out the deepdive thread by @DannyDriess 🌴🤖 The future of LLMs is multimodal and embodied!
47K