Ondra Platek's Avatar

Ondra Platek

@oplatek.bsky.social

BottleCapAI. LLM. Engineer. Researcher. Speaker. Father.

162 Followers  |  225 Following  |  28 Posts  |  Joined: 19.11.2024
Posts Following

Posts by Ondra Platek (@oplatek.bsky.social)

Endure: Mind body and the curiously elastic limits of human performance by Alex Hutchinson.

Good theory before to learn anything from any Dummies series book

29.01.2026 21:40 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Interesting did not know that arxiv requires peer-reviewed papers
arxiv.org/abs/2601.17036

29.01.2026 17:56 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 0    πŸ“Œ 1
Preview
Pulse: Community News App - App Store Download Pulse: Community News by BottleCap AI on the AppΒ Store. See screenshots, ratings and reviews, user tips and more games like Pulse: Community News.

apps.apple.com/cz/app/pulse...

29.01.2026 16:47 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
Jaroslav Beck a TomΓ‘Ε‘ Mikolov zΓ­skΓ‘vajΓ­ 150 milionΕ― pro svΕ―j AI startup. Investuje do nΔ›j i Ivo Lukačovič ČeskΓ½ startup BottleCap AI chce zefektivnit trΓ©novΓ‘nΓ­ modelΕ― umΔ›lΓ© inteligence. MΓ‘ na to novou investici a uvΓ‘dΓ­ prvnΓ­ aplikaci pro iOS.

Czech news adopted it fast cc.cz/jaroslav-bec...

29.01.2026 16:41 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
BottleCapAI Announcement, Seed Round, Pulse: Community News
YouTube video by BottleCapAI BottleCapAI Announcement, Seed Round, Pulse: Community News

Our Pulse News App -- our first app is live on iOS today.

We also announce star investors cap.
See the announcement video to see our plans.
If you are LLM guru at research or product building, join us to make the plans a reality
youtu.be/cA8-XJeoZAQ?...

29.01.2026 16:41 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 3    πŸ“Œ 0

Thank you to all coauthors to push the article to acceptance!

29.01.2026 15:42 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

In general using LLM annotations for self-improvement is great future direction

29.01.2026 15:41 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

Annotations using LLM are pretty under researched topics.

I would be interested how it performs on coding tasks and if it helps during CoT for coding tasks

29.01.2026 15:41 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Post image

Well done @zdenekkasner.bsky.social et al!

LLMs as Span Annotators: A Comparative Study of LLMs and Humans is accepted to multilingual-multicultural-evaluation.github.io πŸŽ‰

See paper arxiv.org/abs/2504.08697

29.01.2026 15:35 β€” πŸ‘ 8    πŸ” 2    πŸ’¬ 2    πŸ“Œ 1

What you plan to read on vacation?

29.01.2026 15:28 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Post image Post image

Dr.LLM: Adaptive Layer Routing for Efficient Inference

Avoiding computation is pretty obvious way how to be more efficient.
The question is how to maintain quality. MoE and router is an efficient approach but suffer from exposure bias fixed thresholds.

arxiv.org/abs/2510.12773

29.01.2026 15:25 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 3    πŸ“Œ 0

so what it is?

29.01.2026 14:00 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

What do you want to do with the deeper layers

29.01.2026 12:23 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Do you think future is MoE or dense models?

29.01.2026 12:21 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Post image

Qwen3 Embedder is amazing
but it is tricky to use.
Left padding confused lot of early adopters.

However, clearly it is the future. Reuse of pre-trained LLM for everything.
Including vector search where BERT-like models ruled.

1/N

arxiv.org/abs/2506.05176
huggingface.co/Qwen/Qwen3-E...

29.01.2026 12:18 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 5    πŸ“Œ 0

So would you like to prune them?

29.01.2026 12:13 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
The Flexibility Trap: Why Arbitrary Order Limits Reasoning Potential in Diffusion Language Models Diffusion Large Language Models (dLLMs) break the rigid left-to-right constraint of traditional LLMs, enabling token generation in arbitrary orders. Intuitively, this flexibility implies a solution sp...

The no free lunch idea is here again.

Saw it in TTS models where Guided Attention worked very well. Idea is simple: For speech conversion prefer guide the attention to focus on narrow context. It learns faster.

Looking forward to deep dive on this one - expect similarities
arxiv.org/abs/2601.15165

29.01.2026 12:11 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

It can be used in RAG, so for fetching relevant documents in clawbot it makes very much sense

29.01.2026 12:09 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

The last token pooling is compressing the embeddings too much.

However it is engineering marvel.
Including tricks like the spherical last checkpoint averaging is awesome.

29.01.2026 12:06 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
πŸ‘€ WE ARE LOOKING FOR YOU! (yes, you!) 🚨NoCap test deadline 11/11/2025 πŸ† $3K / $2K / $1K prizes Our new team members shared tips&tricks which helped them to go through our test! TRY IT NOW! πŸ‘‰β€¦ | ... πŸ‘€ WE ARE LOOKING FOR YOU! (yes, you!) 🚨NoCap test deadline 11/11/2025 πŸ† $3K / $2K / $1K prizes Our new team members shared tips&tricks which helped them to go through our test! TRY IT NOW! πŸ‘‰ https:/...

The video is on LinkedIn. Check it out it turn out fun www.linkedin.com/posts/bottle...

07.11.2025 22:04 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
LinkedIn This link will take you to a page that’s not on LinkedIn

πŸ‘€ WE ARE LOOKING FOR YOU! (yes, you!)
🚨NoCap test deadline 11/11/2025
πŸ† $3K / $2K / $1K prizes

Our new team members shared tips&tricks which helped them to go through our test!
TRY IT NOW! πŸ‘‰ github.com/BottleCapAI/...

07.11.2025 21:54 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Czech President Delivers Powerful Speech at UN, Slams Russia, Iran, Israel Aggression | AC1G
YouTube video by DRM News Czech President Delivers Powerful Speech at UN, Slams Russia, Iran, Israel Aggression | AC1G

The right words at the right time with the right style for the venue. Proud of our president πŸ‡¨πŸ‡Ώ
youtu.be/d3wT84egi-g?...

26.09.2025 11:12 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Me: Scaling algorithms is just much more fun, than just burning electricity on more and more GPUs.

22.09.2025 13:41 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
GitHub - BottleCapAI/NoCap-Test: Open Test for BottleCapAI Open Test for BottleCapAI. Contribute to BottleCapAI/NoCap-Test development by creating an account on GitHub.

Official BottleCapAI: we believe AI shouldn’t cost tens of millions to train.
We are now opening challenge for those who want to help with that!

πŸ† $3K / $2K / $1K prizes
⏰ Deadline: 11/11/2025
πŸ–₯️ 1 GPU. Your ideas.

πŸ‘‰ Join the NoCap Test:

github.com/BottleCapAI/...

22.09.2025 13:40 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Terminology Translation Task

πŸ“£Take part in 3rd Terminology shared task @WMT!πŸ“£
This year:
πŸ‘‰5 language pairs: EN->{ES, RU, DE, ZH},
πŸ‘‰2 tracks - sentence-level and doc-level translation,
πŸ‘‰authentic data from 2 domains: finance and IT!

www2.statmt.org/wmt25/termin...

Don't miss an opportunity - we only do it once in two years😏

06.06.2025 15:54 β€” πŸ‘ 3    πŸ” 2    πŸ’¬ 0    πŸ“Œ 2
Preview
Algorithmic Simplicity This is an educational channel for all things algorithmic, including but not limited to computer science, machine learning, physics, and mathematics.

I just found awesome channel on YouTube.
It has 77k subscribers with just 7 videos.
Because the videos are just awesome!

I wish I could explain stuff that simply!

Does anybody know how long take to prepare such video?

www.youtube.com/@algorithmic...

04.05.2025 09:00 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Thanks to all my awesome collaborators!

πŸ‘‰οΈ VilΓ©m @zouharvi.bsky.social
πŸ‘‰οΈ PatrΓ­cia @patuchen.bsky.social
πŸ‘‰οΈ Ivan @ivankartac.bsky.social
πŸ‘‰οΈ KristΓ½na OnderkovΓ‘
πŸ‘‰οΈ OndΕ™ej P. @oplatek.bsky.social
πŸ‘‰οΈ Dimitra @dimitrag.bsky.social
πŸ‘‰οΈ Saad @saad.me.uk
πŸ‘‰οΈ OndΕ™ej D. @tuetschek.bsky.social
πŸ‘‰οΈ Simone Balloccu

15.04.2025 11:10 β€” πŸ‘ 5    πŸ” 3    πŸ’¬ 1    πŸ“Œ 0
Preview
Large Language Models as Span Annotators Website for the paper Large Language Models as Span Annotators

How do LLMs compare to human crowdworkers in annotating text spans? πŸ§‘πŸ€–

And how can span annotation help us with evaluating texts?

Find out in our new paper: llm-span-annotators.github.io

Arxiv: arxiv.org/abs/2504.08697

15.04.2025 11:10 β€” πŸ‘ 20    πŸ” 7    πŸ’¬ 1    πŸ“Œ 2
Preview
Čech povede evropský výzkum AI - 3. února 05:59 - Studio 6 | ČeskÑ televize Jan Hajič

We've been making the media rounds!
πŸ‘‰πŸ“Ί @hajicjan.bsky.social talked about the new OpenEuroLLM project on Czech TV's Studio 6 www.ceskatelevize.cz/porady/10969...
πŸ‘‰πŸ“» @tuetschek.bsky.social discussed #LLMs on Czech Radio radiozurnal.rozhlas.cz/proc-umela-i...).

07.02.2025 21:44 β€” πŸ‘ 7    πŸ” 3    πŸ’¬ 0    πŸ“Œ 1
ŠÑrka ZikÑnovÑ | ÚFAL

Congrats to ✨ŠÑrka ZikΓ‘novÑ✨on her promotion to Associate Professor! πŸŽ‰ Her research explores syntax, information structure, and discourse relations in Czech and beyond. She contributed to key linguistic corpora and now focuses on psycholinguistic studies of discourse. ufal.mff.cuni.cz/sarka-zikanova

25.02.2025 10:44 β€” πŸ‘ 6    πŸ” 1    πŸ’¬ 0    πŸ“Œ 0