Junlin Han

MTS @ @amilabs | Multimodal pre-training & post-training. Ex Meta FAIR. PhD Oxford. Opinions are my own.

[Junlin Han @han_junlin]

We live in a multimodal world. We see, talk, act, and dream.

Yet most LLMs still start with language pretraining. Why not train them natively with multimodal I/O from scratch?

Because it’s SUPER HARD, adding modalities triggers training instability, design complexity, and often modal competition

So what’s the path forward?

Introducing: Towards Physics of Multimodal Pretraining (https://t.co/xgfbdDPFq4)

We unpack the underlying mechanics of multimodal pretraining across 4 aspects: Knowledge Flow, Modality Synergy, Early Unification, and Recipe.

[Junlin Han @han_junlin]

Nice model with native multimodality!!! Congrats to the team cooking it!

[Thinking Machines @thinkymachines]

Introducing Inkling 2 Small. Scoring 31 on the AA Intelligence Index, it's a big jump over its predecessor.

Inkling 2 Small is available for fine-tuning and inference on Tinker today, with full weight release planned soon.

https://t.co/VR02tw73fJ

[Junlin Han @han_junlin]

@poppingG5 @amilabs thanks a lot Chenhui!!

[Junlin Han @han_junlin]

I joined @amilabs ~a month ago as an MTS, working on world model pre-training and capability emergence.

There has never been a better time to build AI for the real world with a scaling mindset. Thrilled to join forces with friends old and new to build the next frontier🚀!

[Junlin Han @han_junlin]

@YuxuanEasonFan @amilabs Thank you Yuxuan!!

[Junlin Han @han_junlin]

@zinho_leung @amilabs Thank you Zhanhao!!

[Junlin Han @han_junlin]

@xingjian_leng @sainingxie @amilabs Thanks so much Xingjian!

[Junlin Han @han_junlin]

@LeotheHorseman @amilabs Thanks a lot 🐴!

[Junlin Han @han_junlin]

@TianzheC @amilabs Work!

[Junlin Han @han_junlin]

@hanlin_hl @amilabs Thanks a lot Han!

[Junlin Han @han_junlin]

@matheold @amilabs Massive ambition, relentless execution.

[Junlin Han @han_junlin]

@vhjf36495872 @sainingxie @amilabs haha thank you Jinfa!

[Junlin Han @han_junlin]

@Kuvvius @amilabs Thanks a lot Jiawei!!

[Junlin Han @han_junlin]

@zzzoooeee321 @amilabs Thank you very much Zoe!!

[Junlin Han @han_junlin]

@ElectronDev @amilabs Thanks, Peter! Yeah, the digital world is mostly solved, so it's a good time to explore physical world (and connect both)!

[Junlin Han @han_junlin]

@purshow04 @amilabs Vision is a vision and there will be more than vision haha. Thanks yuwei!

[Junlin Han @han_junlin]

@Haian_Jin @amilabs Thank you so much Haian!!

[Junlin Han @han_junlin]

@SimulatedAnneal @amilabs Thank you very much Xingjian!!

[Junlin Han @han_junlin]

@Junhua_Yao13 @amilabs Thanks Junhua!!

[Junlin Han @han_junlin]

@kaiwynd @wenhaocha1 @amilabs Thank you Kaifeng!!

[Junlin Han @han_junlin]

@DYDYYDYYYD @amilabs Thanks Jianhao!!

[Junlin Han @han_junlin]

@ChuanxiaZ @amilabs thanks chuanxia!! Yeaaaa that would be super fun 😆.