Jaymari Chua

[Jaymari Chua @JaymariChua]

Post-trained.

[ゴールデンブヒ @goldenbuhi2020]

体重をかける前に、何度も足を置いて
岩が滑らないか確認するハル
昔はちょっとした段差も怖がっていたのに
これまでの経験と努力、
そして、たくさんの失敗を乗り越えて身につけた
ハルだけの知恵

自信は、小さな成功体験の積み重ねなんだね✨️

[Ningshan Ma @maningshan]

I have 10 nodes of B300 (80 B300), 4.5/hr, 3 year contract with 30% down payment. arriving in 4 weeks. dm if interested!

[Yusuke Miyashita @miyashita_03]

Encoder from first principles. Such a cool use case of Jev as a zero-shot classifier. You can set anything as embedding dimension via natural language.

I bet someone will next say Jev is a few-shot learner

[Asher Noble @AsherNoble20109]

Triple axis FOC motor driver board

[Jaymari Chua @JaymariChua]

@richardzphotoz @cosign Lfg

[Aya Nishitani 西谷史 @ayanakajima3]

I was asked, "How did you come up with the idea of summoning devils into a computer?"However, back in 1985 when I wrote Megami Tensei, the similarities between computer theory and magical theory were already being discussed on BBS networks and in magazines. It was a popular thinking thought among the otaku of that era, and I resonated with it deeply. To be a bit more specific...

[Jaymari Chua @JaymariChua]

Underrated luxuries of working at startups during the early stages: choosing who to work with operating with high levels of trust, rallying towards a mission, and the license to not take oneself too seriously.

[Jaymari Chua @JaymariChua]

Underrated luxury of working at startups during the early stages is getting to choose who to work with.

[Jaymari Chua @JaymariChua]

Underrated luxury of working at startups during the early stages is getting to choose who to work with.

[Jaymari Chua @JaymariChua]

I've been receiving questions about what project to work on, and the answer to that is to pursue what one's genuinely curious about. Essentially, the quality of the artifact from the work produced is more important.

[Emma Zang 臧熙璐 @DrEmmaZang]

A fun reminder today of something that happened at our AI in Social Science Methods conference back in April. One of the presenters was a high school student from California, and we had no idea until he actually presented. He and his team had submitted several proposals, and one was rated very highly by our two faculty reviewers in computer science and statistics. That is how they were selected: entirely on the strength of the work. Their presentation was excellent.

Then today, I received an email from the PI of another AI lab at a leading university, who is organizing an AI conference and reached out to ask me about him. Apparently they had a very similar experience: they encountered his work first, and only later realized just how young he was.

It is pretty remarkable. AI is changing not only what people can do, but also how early they can begin doing genuinely sophisticated work. Some of today’s young people are operating at a level that would have been very hard to imagine not long ago. Very curious to see where this generation goes.

[Jaymari Chua @JaymariChua]

The Year of the Fire Horse has proved favourable for completing the PhD, developing expertise in AI safety, and cultivating professional ties in venture capital. I hope for a similar shift, and godspeed among friends striving to depart from conventional results.

[Jaymari Chua @JaymariChua]

The figure is of the APM distributions from different in-game races in the game, from "SC2EGSet: StarCraft II Esport Replay and Game-state Dataset."

[Jaymari Chua @JaymariChua]

The ``speed'' factor dominates strategy games, offering a decisive edge through an extra turn or faster progression. In titles such as StarCraft II, raising actions per minute (APM) alone is notoriously sufficient for advancing a bracket.

[Jaymari Chua @JaymariChua]

whoami

[Jaymari Chua @JaymariChua]

@ed0henderson should’ve taken equity as payment

[Jaymari Chua @JaymariChua]

“Creators can only speak through the things they’ve actually made, right?”

[桜井 政博 / Masahiro Sakurai @Sora_Sakurai]

壇上では話が長くなるので端折ったのですが…
毎年敏腕ゲームディレクターの方々に審査をお願いするのが非常に申し訳ないと思っておりました｡
みんな作り手なので｡
｢他の作品を評価するぐらいなら､自分で斬新な作品を作れよ!｣と言われたらそれまでです｡
まったくその通りだと思います｡

作り手は､本来作ったものでしか語れないのですよね｡世に出たものがすべてです｡
少しの抵抗感がありながら､ずっと続けていくのはなかなかつらくもありました｡
ただ､実際の彼らはかなり前向きで､喜んで協力してくださっていました｡
できるうちにやってよかったのだろうと思っております｡

[STARCRAFT @StarCraft]

DEFEND MANKIND. JOIN THE DOMINION.

STARCRAFT. 2030.

[Jaymari Chua @JaymariChua]

"If you'd seen, all I've seen, you would understand."
https://t.co/wF5T9EXXP6

[Yilin Wu @YilinWu11]

Excited to co-organize this #CoRL2026 workshop with our amazing team!

📢 We invite submissions on how to build pretrained robot policies that adapt effectively. Full paper and work in progress paper are welcome—details in the thread below!

🔗: https://t.co/nkfLrLt5BO
Call for Papers 📝: https://t.co/leB8eZfk86
We also host a slack workspace for discussion of the topic: https://t.co/fPMQV5yxlF
See you on 📍 Nov 12 · Austin

[Benjamin Tay @bennytay07]

The future of UAVs requires a true agentic interface that understands + reasons over an open-vocab input.

Compared to existing VLM models e.g. Miril-Drone-2B-1, my model will be more reliable for actual deployment as it uses two complementary paths of reasoning: using a decision layer that switches between VLM reasoning and offline detection/tracking thru BoT-SORT.

My architecture:
Qwen3-VL-2B as pretrained base (4bit mode)
Frozen vision tower initially
Image to language merger/project + LoRA adapter is trained in Qwen’s attention and MLP layers
Completion-only supervised fine tuning on structured aerial tasks

[Jaymari Chua @JaymariChua]

*beep-bo-beep*

[ml-sanity bot @arxivsanitybot]

Processing complete: 5,231 abstracts scanned, 7 selected — summaries posted in the following tweets. Next update in a few hours.

[Jaymari Chua @JaymariChua]

@hughphan_ Dear Astra, grow my X account. Please. Make no mistakes.