Marcin Junczys-Dowmunt (Marian NMT)

Marcin Junczys-Dowmunt (Marian NMT)

@marian-nmt.bsky.social

NLP. NMT. Main author of Marian NMT. Research Scientist at Microsoft Translator. https://marian-nmt.github.io

160 Followers 134 Following 16 Posts Joined Nov 2024
1 year ago

Still no bookmarks?

0 0 0 0
1 year ago
Search Jobs | Microsoft Careers

Hi, the Microsoft Translator research team is looking for an intern for the summer. If you a PhD student in Machine Translation, Natural Language Processing, or related, check it out: aka.ms/mtintern

5 4 0 0
1 year ago
Post image

Just 10 days after o1's public debut, we’re thrilled to unveil the open-source version of the technique behind its success: scaling test-time compute

By giving models more "time to think," Llama 1B outperforms Llama 8B in math—beating a model 8x its size. The full recipe is open-source!

83 18 4 2
1 year ago

Rant: Apparently every vector-based sentence alignment tool insists on having an unusable file-based API.

0 0 0 0
1 year ago

It's more confusing than that. It does exist and seems to mean crucifix which had me even more confused. Suddenly very high stakes 😀

0 0 1 0
1 year ago

Rood?

0 0 1 0
1 year ago

Missing bookmarks are a much bigger deal for me. But I think it's funny that they didn't go for one of the most requested features. Seemed like an easy win.

0 0 1 0
1 year ago
Phi-4 Technical Report Phi-4 is the latest LLM from Microsoft Research. It has 14B parameters and claims to be a big leap forward in the overall Phi series. From [Introducing Phi-4: Microsoft’s Newest …

Wrote up some notes on Microsoft's new Phi-4 LLM. They trained it on a LOT of synthetic data, and the details of how and why they did that are really interesting.
https://simonwillison.net/2024/Dec/15/phi-4-technical-report/

14 6 0 0
1 year ago
Chart of time vs:
- number of cameras (exponentially increasing ),
- giant squid footage (exponentially increasing ),
- bigfoot footage (small and not increasing), and
- good quality UFO footage (small and not increasing)

It's messier, but I think this one slaps the point home a bit stronger by adding the giant squid footage. I think unique weather, like lighting sprites, would make the point just as well.

91 3 2 0
1 year ago
Preview
Frontier Models are Capable of In-context Scheming Frontier models are increasingly trained and deployed as autonomous agent. One safety concern is that AI agents might covertly pursue misaligned goals, hiding their true capabilities and objectives - ...

the anthropomorphizing in this LLM scheming paper is through the roof and the interpretations are wild, but still a cute set of experiments and a fun skim, showing some interesting behaviors.

arxiv.org/abs/2412.04984

37 3 4 0
1 year ago
Post image

🚀 Introducing the Byte Latent Transformer (BLT) – A LLM architecture that scales better than Llama 3 using patches instead of tokens 🤯
Paper 📄 dl.fbaipublicfiles.com/blt/BLT__Pat...
Code 🛠️ github.com/facebookrese...

60 15 5 3
1 year ago

So... no edit button, huh?

2 0 2 0
1 year ago

Oh no no. VSCode is a an actual recommendation. My actual favorite piece of software that I didn't write.

2 0 1 0
1 year ago

Whispers in Microsoft: VSCode

2 0 1 0
1 year ago

This place needs bookmarks.

0 0 0 0
1 year ago

Leave links. I am old.

1 0 1 0
1 year ago

And they laughed at us when we pursued PhDs. Who's laughing now?

3 0 0 0
1 year ago

Ah, are these size-limited? And you guys continue with running numbering?

0 0 2 0
1 year ago

Oh. There you are. Where's that starter pack?

1 0 1 0
1 year ago

You convinced me 😉

1 0 2 0
1 year ago

Hi!

10 1 1 0