
applied scientist @alexa99,⚡️ @reboot_hq | prev @Umich #AlexaPrize
BOS
Joined March 2008
- Tweets3.1K
- Following1.8K
- Followers330
- Likes25.2K
Really cool. nvidia based open model coalition
Nvidia offers DGX Cloud to AI foundation model labs for new Nemotron Coalition - DCD datacenterdynamics.com/en/ne…
4
146
I was able to try @yutori_ai as a beta user. collecting latest financial trends to gathering neurips after events from an online agent. I suggest folks to try!
Waymo scientists found scaling laws in autonomous driving! waymo.com/blog/2025/06/scali…
4
391
Interesting take on q-learning - on-poliy RL algorithms. thank you for simplifying the research into a narrative @seohong_park
1
3
439
홍정훈 Chung Hoon Hong retweeted
vincentwoo.com/3d/sutro_towe…
Drone 으로 찍어서 3D Gaussian Splat 으로 렌더링.
홍정훈 Chung Hoon Hong retweeted
A few months ago, I started a series of blog posts to explain what it is like to work as a researcher in the industry in the hope of documenting some useful information for those facing this potential transition (especially junior folks). Today, I've published the last part
🧵
홍정훈 Chung Hoon Hong retweeted
Last week our paper on SynthID-text, a method from @GoogleDeepMind for watermarking LLM-generated text, was published in @Nature. 🧵
nature.com/articles/s41586-0…
홍정훈 Chung Hoon Hong retweeted
AI companies are constructing private highways on public land—commercializing closed products built on public data, public resources, open research, and the open web.
We can’t rely on companies to build everything we need from AI, and we can’t afford the risk that they won’t.
Today, Mozilla is releasing a new vision for Public AI
We’re making a big push to promote public goods, public orientation, and public use throughout every step of AI development and deployment 🧵
blog.mozilla.org/en/mozilla/…
홍정훈 Chung Hoon Hong retweeted
New short course on Pretraining LLMs! Developed with @UpstageAI and taught by their CEO @hunkims and CSO @echojuliett.
While prompting or fine-tuning existing models works well for many general language tasks, pretraining is valuable for specialized domains or languages with limited representation in current models.
This course walks you through the LLM pretraining pipeline:
1. Data preparation: Learn to source, clean, and prepare training data using HuggingFace.
2. Model architecture: Configure transformer networks, including modifying existing models.
3. Training: Set up and run training using open-source libraries.
4. Evaluation: Benchmark performance using popular evaluation strategies.
As an example use case, you'll also compare the output of a base model with its fine-tuned and further pretrained variants, to see the impact of pretraining on a model's ability to write Python.
The course also explores an innovative technique called depth up-scaling, which Upstage used to train their Solar model family, reducing pretraining compute costs by up to 70%. This technique works by first duplicating layers of a smaller pretrained model to form a larger model, and then further pretraining the result.
Sign up here! deeplearning.ai/short-course…
홍정훈 Chung Hoon Hong retweeted
Just published the second part of my blog post to reflect on my journey as an early-career industry researcher, where I talk about how I would approach choosing a team today and what I'd factor into my consideration.
Post: qipeng.me/blog/what-do-indus…
홍정훈 Chung Hoon Hong retweeted
New essay in @reboot_hq in which I recount my own history with the lost art of iPhone jail breaking
홍정훈 Chung Hoon Hong retweeted
🎉Introducing Upstage/SOLAR-10.7B🚀 It's compact, yet powerful, delivering the best performance among models under 30B 🙌. The secret? We use vertical depth upscaling instead of horizontal upscaling methods like MOE. We found depth upscaling to be super efficient 🤓. Check it out at huggingface.co/upstage/SOLAR…! Also check out the leaderboard at huggingface.co/spaces/Huggin… (Post written by Solar 10.7B)
홍정훈 Chung Hoon Hong retweeted
Amazon is bringing @agilityrobotics Digit humanoids into its warehouses:
theguardian.com/technology/2…
This week is the launch of 3rd issue of the @kernel_magazine !!
The future doesn’t just arrive; it has to be built. What does that mean for writing about technology?
The third issue of @kernel_magazine is out today, and you can get your copy now at shop.kernelmag.io ⚡
Excited to share our #INTERSPEECH2023 paper on Record deduplication for entity distribution modeling in ASR transcripts. Interested in hearing more about how we make voice assistants to keep up with trending search queries.
1
5
1,198
Here is link to our paper amazon.science/publications/…
1
354
Also, if you are intereseted in other papers from Amazon amazon.science/blog/a-quick-…
247
Hi I am at #INTERSPEECH2023 this week. Excited to hear about your work. Interested in ASR, parameter efficient training, etc(LLMS 😆) calendly.com/chunghoon-resea…
1
328


















