Danijar Hafner ✓
@danijarh · SF · joined 10 Aug 2013
Building general agents for the physical world. Prev @GoogleDeepMind @UCBerkeley @UofT
34 009Followers
1 223Following
2 734Posts total
684.1KViews on collected posts
Derniers posts
Recherche de nouveaux posts…
Proud to share that I've been recognized as one of the 35 Innovators under 35 by MIT Technology Review!
Developing world models has been an incredible journey over the past 8 years, both training the world models and leveraging them for imagination training and motor control 🌎 h
15.4K views · 271 likes
· 12 reposts · 33 replies
Open on X →
Really enjoyed this conversation on the BuzzRobot podcast! 🤖⚡
We covered a lot of ground on the future of the field:
- the path from frontier models to AGI
- discovering instead of distilling reasoning
- continual learning
- scalable objectives
- how to solve robotics
9K views · 51 likes
· 5 reposts · 6 replies
Open on X →
Key points from my recent conversation with @danijarh
- Architecture doesn’t matter to achieve AGI. What is missing: algorithmic advances and objective function improvements. More compute still matters.
- AI shouldn’t be limited by human reasoning. It will be able to go beyond h
12.1K views · 29 likes
· 1 reposts · 2 replies
Open on X →
0:46
✨ Excited to share this AMA with @hackclub, a high school community hosting @elonmusk @realGeorgeHotz @3blue1brown and many others.
We talk about world models, robotics, and careers in AI. Check it out for an accessible intro to cutting edge research! 🚀
https://t.co/WNmcd1zls0
9.5K views · 65 likes
· 6 reposts · 2 replies
Open on X →
@danijarh @TalkRLPodcast 😄 Exactly what I was waiting for, thank you for that! Always inspiring to listen to your thoughts. I'm curious what makes you all the time, including reconstruction objective in your world model training?
229 views · 2 likes
· 0 reposts · 1 replies
Open on X →
@danijarh @TalkRLPodcast That podcast sounds super interesting, Danijar! Looking forward to hearing about Dreamer 4 and the future of robotics, sounds like a good time!
134 views · 0 likes
· 0 reposts · 0 replies
Open on X →
Excited for this podcast episode with TalkRL to be out! 🎙️
We talk about the story behind Dreamer 4, the details of scalable world models, and the future of robotics (and beyond) 🤖🌏🚀
Thanks for the fun conversation, @TalkRLPodcast
28.1K views · 136 likes
· 12 reposts · 4 replies
Open on X →
E73: Danijar Hafner on Dreamer v4
@danijarh (ex-@GoogleDeepMind RS) on offline world models for safe robotics, Shortcut Forcing for fast diffusion video models, outperforming OpenAI’s VPT with 100× less data, his “APD” theory unifying exploration and empowerment, and more! https:
17.2K views · 24 likes
· 5 reposts · 4 replies
Open on X →
@danijarh @wilson1yan I love that Minecraft is still the benchmark
4.2K views · 37 likes
· 0 reposts · 1 replies
Open on X →
Big thanks to my co-lead @wilson1yan and to @countzerozzz!
Check out the website for videos and the paper for details & many ablations
Website: https://t.co/fyWmC9141Q
Paper: https://t.co/7Y5T0ZJ7g4
Happy to answer any questions! ✨
9.5K views · 159 likes
· 6 reposts · 4 replies
Open on X →
We have come a long way since Dreamer 3, which is based on a more lightweight but less scalable RNN with variational objective
While the lightweight approach still makes sense for easier tasks, Dreamer 4 allows scaling to much more diverse datasets and environments 🚀 https://t.c
9.6K views · 89 likes
· 3 reposts · 1 replies
Open on X →
✅ We find that imagination training not only makes policies more robust but also more efficient, so they achieve milestones towards the diamond faster
✅ Moreover, using the WM representations for behavioral cloning outperforms using the general representations of Gemma 3 https:/
9K views · 89 likes
· 1 reposts · 1 replies
Open on X →
📈 On the offline diamond challenge, Dreamer 4 outperforms OpenAI's VPT offline agent despite using 100x less data
It also outperforms modern behavioral cloning recipes, even when they are based on powerful pretrained models such as Gemma 3 https://t.co/fdvv5tfcvG
9.7K views · 101 likes
· 1 reposts · 1 replies
Open on X →
▶️ Shortcut forcing builds on diffusion forcing and shortcut models, training a sequence model with both the noise level and requested step size as inputs
This enables much faster frame-by-frame generations than diffusion forcing, without needing a distillation phase ⏱️ https://
21.2K views · 130 likes
· 3 reposts · 1 replies
Open on X →
For accurate and fast generations, we use an efficient transformer architecture and a novel shortcut forcing objective ⚡
We first pretrain the WM, finetune agent tokens into the same transformer to predict policy & reward, and then improve the policy by imagination training http
11.5K views · 125 likes
· 2 reposts · 2 replies
Open on X →
The Dreamer 4 world model predicts complex object interactions while achieving real-time interactive inference on a single GPU
It outperforms previous world models by a large margin when put to the test by human interaction 🧑💻 https://t.co/KzVTIK10LF
12.6K views · 155 likes
· 1 reposts · 2 replies
Open on X →
0:27
🧠 Dreamer 4 learns a scalable world model from offline data and trains a multi-task agent inside it, without ever having to touch the environment. During evaluation, it can be guided through a sequence of tasks.
These are visualizations of the imagined training sequences https:/
15.6K views · 172 likes
· 3 reposts · 1 replies
Open on X →
0:07
Excited to introduce Dreamer 4, an agent that learns to solve complex control tasks entirely inside of its scalable world model! 🌎🤖
Dreamer 4 pushes the frontier of world model accuracy, speed, and learning complex tasks from offline datasets.
co-led with @wilson1yan https://t.
461.3K views · 2.6K likes
· 352 reposts · 85 replies
Open on X →
0:09
💎 Enabled by imagination training, Dreamer 4 is the first agent to mine diamonds in Minecraft entirely from offline data!
This setting is crucial for fields like robotics, where online interaction is not practical. The task requires 20k+ mouse/keyboard actions from raw pixels ht
27.9K views · 323 likes
· 12 reposts · 8 replies
Open on X →
2:55
Face aux comptes de taille comparable
1 posts des 90 derniers jours, à côté de la tranche de 10K–100K abonnés. diffusé bien au-delà de son propre public, et ce public réagit.
Vues médianes15 411ce compte924médiane pour 10K–100K
Portée, %45.31%ce compte3.62%médiane pour 10K–100K
Engagement, %2.07%ce compte1.52%médiane pour 10K–100K
| Indicateur | Ce compte | Médiane pour 10K–100K | Rapport |
|---|---|---|---|
| Vues médianes par post | 15 411 | 924 | 16.7× |
| Portée (vues ÷ abonnés) | 45.31% | 3.62% | 12.5× |
| Taux d'engagement | 2.07% | 1.52% | 1.36× |
Autres comptes de cette tranche → Comparer avec un autre compte → Comment ces repères sont établis →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
9.7K30 Sep
9K
9.6K
9.5K
4.2K
17.2K10 Nov
28.1K
237
13411 Nov
22913 Nov
9.5K9 Dec
12.1K15 Jan
9K
15.4K10 Sep
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
1.06%30 Sep
1.01%
0.97%
1.80%
0.90%
0.20%10 Nov
0.54%
0.84%
0.00%11 Nov
1.31%13 Nov
0.77%9 Dec
0.27%15 Jan
0.69%
2.07%10 Sep
Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.
What the audience does
Likes68.1%4 564 in total
Reposts6.3%425 in total
Replies2.4%159 in total
Quotes1.6%104 in total
Bookmarks21.6%1 451 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
Comptes similaires
teamfk
@freekickerz
68K followers
· 9.4K posts
Federal Reserve Bank of San Francisco
@sffed
68K followers
· 15.4K posts
Circa Survive
@circasurvive
68K followers
· 8K posts
TaytawanFC ✓
@taytawanfc
68K followers
· 97.4K posts
SYDUS ONIICHAN💢💢😭😭
@SydusArtsLinks
68K followers
· 600 posts
Dawn M. Bennett
@DawnMBennettVA
68K followers
· 39.1K posts
Rahul Sidhu ✓
@rahul
68K followers
· 2.8K posts
Stop Arab Hate
@StopArabHate
68K followers
· 2.4K posts