Kaustubh Sridhar
@kaustubhsridhar.bsky.social
Research Scientist at Google Deepmind.
Prev: @UPenn @Amazon @IITBombay
http://kaustubhsridhar.github.io/
Prev: @UPenn @Amazon @IITBombay
http://kaustubhsridhar.github.io/
Very excited to share our work on SIMA, a general embodied agent for 3D worlds.
Even more excited to share our work on self-improvement where our gemini-based SIMA 2 agent self-improves to human performance (no humans needed) using a gemini utility model and a gemini task setter! More info ⏬
Even more excited to share our work on self-improvement where our gemini-based SIMA 2 agent self-improves to human performance (no humans needed) using a gemini utility model and a gemini task setter! More info ⏬
Can't tell you how great it is to finally be able to release and talk about this work, SIMA 2, the next step toward embodied intelligence in rich, interactive 3D worlds!
deepmind.google/sima
deepmind.google/sima
SIMA 2: A Gemini-Powered AI Agent for 3D Virtual Worlds
Introducing SIMA 2, the next milestone in our research creating general and helpful AI agents. By integrating the advanced capabilities of our Gemini models, SIMA is evolving from an instruction-foll…
deepmind.google
November 13, 2025 at 4:17 PM
Very excited to share our work on SIMA, a general embodied agent for 3D worlds.
Even more excited to share our work on self-improvement where our gemini-based SIMA 2 agent self-improves to human performance (no humans needed) using a gemini utility model and a gemini task setter! More info ⏬
Even more excited to share our work on self-improvement where our gemini-based SIMA 2 agent self-improves to human performance (no humans needed) using a gemini utility model and a gemini task setter! More info ⏬
REGENT will be presented as an Oral at ICLR 2025 in Singapore 🇸🇬, given to the top 1.8% of 11672 submissions! More details at our website: bit.ly/regent-research
February 22, 2025 at 6:55 PM
REGENT will be presented as an Oral at ICLR 2025 in Singapore 🇸🇬, given to the top 1.8% of 11672 submissions! More details at our website: bit.ly/regent-research
Reposted by Kaustubh Sridhar
Is scaling current agent architectures the most effective way to build generalist agents that can rapidly adapt?
Introducing 👑REGENT👑, a generalist agent that can generalize to unseen robotics tasks and games via retrieval-augmentation and in-context learning.
Introducing 👑REGENT👑, a generalist agent that can generalize to unseen robotics tasks and games via retrieval-augmentation and in-context learning.
December 14, 2024 at 9:50 PM
Is scaling current agent architectures the most effective way to build generalist agents that can rapidly adapt?
Introducing 👑REGENT👑, a generalist agent that can generalize to unseen robotics tasks and games via retrieval-augmentation and in-context learning.
Introducing 👑REGENT👑, a generalist agent that can generalize to unseen robotics tasks and games via retrieval-augmentation and in-context learning.
Is scaling current agent architectures the most effective way to build generalist agents that can rapidly adapt?
Introducing 👑REGENT👑, a generalist agent that can generalize to unseen robotics tasks and games via retrieval-augmentation and in-context learning.
Introducing 👑REGENT👑, a generalist agent that can generalize to unseen robotics tasks and games via retrieval-augmentation and in-context learning.
December 14, 2024 at 9:50 PM
Is scaling current agent architectures the most effective way to build generalist agents that can rapidly adapt?
Introducing 👑REGENT👑, a generalist agent that can generalize to unseen robotics tasks and games via retrieval-augmentation and in-context learning.
Introducing 👑REGENT👑, a generalist agent that can generalize to unseen robotics tasks and games via retrieval-augmentation and in-context learning.
Reposted by Kaustubh Sridhar
Cool demo of Gemini 2.0 Flash's new streaming API, by @simonwillison.net.
www.youtube.com/watch?v=mpgW...
www.youtube.com/watch?v=mpgW...
Gemini 2.0 Flash multi-modal streaming demo
YouTube video by Simon Willison
www.youtube.com
December 12, 2024 at 9:29 PM
Cool demo of Gemini 2.0 Flash's new streaming API, by @simonwillison.net.
www.youtube.com/watch?v=mpgW...
www.youtube.com/watch?v=mpgW...
Vancouver is so beautiful!
December 10, 2024 at 7:28 PM
Vancouver is so beautiful!
In the hitchhikers guide to the galaxy, when they built a huge computer (Deep Thought) to answer the ultimate question (of Life, the Universe and Everything), and it took 7.5 million years, it seems like they clearly did both train-time and test-time scaling.
December 5, 2024 at 5:24 PM
In the hitchhikers guide to the galaxy, when they built a huge computer (Deep Thought) to answer the ultimate question (of Life, the Universe and Everything), and it took 7.5 million years, it seems like they clearly did both train-time and test-time scaling.
Reposted by Kaustubh Sridhar
Can no longer tell if LLMs are sounding like humans or some humans have always sounded like LLMs
December 4, 2024 at 1:45 AM
Can no longer tell if LLMs are sounding like humans or some humans have always sounded like LLMs
Reposted by Kaustubh Sridhar
I'd like to introduce what I've been working at @hellorobot.bsky.social: Stretch AI, a set of open-source tools for language-guided autonomy, exploration, navigation, and learning from demonstration.
Check it out: github.com/hello-robot/...
Thread ->
Check it out: github.com/hello-robot/...
Thread ->
December 3, 2024 at 4:51 PM
I'd like to introduce what I've been working at @hellorobot.bsky.social: Stretch AI, a set of open-source tools for language-guided autonomy, exploration, navigation, and learning from demonstration.
Check it out: github.com/hello-robot/...
Thread ->
Check it out: github.com/hello-robot/...
Thread ->
I'm still waiting for the "react/respond to the author rebuttal" from a couple of reviewers :_(
Dear reviewers:
As you react/respond to the author rebuttal can you please articulate the answers to these questions in 1-2 sentences each?
1. Why not a lower score
2. Why not a higher score
This significantly helps bring everyone (authors/reviewers/AC/SAC) on the same page.
As you react/respond to the author rebuttal can you please articulate the answers to these questions in 1-2 sentences each?
1. Why not a lower score
2. Why not a higher score
This significantly helps bring everyone (authors/reviewers/AC/SAC) on the same page.
November 30, 2024 at 5:15 PM
I'm still waiting for the "react/respond to the author rebuttal" from a couple of reviewers :_(
@deep-mind.bsky.social Can we make the Gemini chat history searchable?
November 28, 2024 at 11:21 PM
@deep-mind.bsky.social Can we make the Gemini chat history searchable?
Reposted by Kaustubh Sridhar
We got pretty amazing in context generalization results without lots of data (but in simulation) in our recent paper: kaustubhsridhar.github.io/regent-resea...
Not sure that I agree that an OOM more data is all we need.
Not sure that I agree that an OOM more data is all we need.
REGENT: A Retrieval-Augmented Generalist Agent That Can Act In-Context In New Environments.
REGENT: A Retrieval-Augmented Generalist Agent That Can Act In-Context In New Environments.
kaustubhsridhar.github.io
November 27, 2024 at 8:08 PM
We got pretty amazing in context generalization results without lots of data (but in simulation) in our recent paper: kaustubhsridhar.github.io/regent-resea...
Not sure that I agree that an OOM more data is all we need.
Not sure that I agree that an OOM more data is all we need.
Reposted by Kaustubh Sridhar
Do checkout REGENT: kaustubhsridhar.github.io/regent-resea...
RAG training enables generalization to new robotics and game environments purely via in context learning.
RAG training enables generalization to new robotics and game environments purely via in context learning.
REGENT: A Retrieval-Augmented Generalist Agent That Can Act In-Context In New Environments.
REGENT: A Retrieval-Augmented Generalist Agent That Can Act In-Context In New Environments.
kaustubhsridhar.github.io
November 17, 2024 at 7:44 PM
Do checkout REGENT: kaustubhsridhar.github.io/regent-resea...
RAG training enables generalization to new robotics and game environments purely via in context learning.
RAG training enables generalization to new robotics and game environments purely via in context learning.
Great starter pack from @cpaxton.bsky.social go.bsky.app/DfAoaJ1
November 18, 2024 at 4:28 AM
Great starter pack from @cpaxton.bsky.social go.bsky.app/DfAoaJ1
Reposted by Kaustubh Sridhar