Markus Frohmann

2

100

Markus Frohmann@FrohmannM·7 Mar

@usmasfr Cool use case! You can start with our default setup, using 30 epochs, here: github.com/segment-any-te… Otherwise, I'd focus more on clean segmented training data than tuning epochs; >100 segmented training sentences would be good. You can also tune LoRA rank a bit if needed.

English

0

1

23

usama@usmasfr·4 Mar

@FrohmannM Fine-tuning sat via LoRA to detect scene boundaries in fiction manuscripts. Any tips on optimal training data size and epoch tuning for this kind of domain adaptation? Appreciate the work 🙌

English

0

42

Markus Frohmann@FrohmannM·4 Mar

wtpsplit now supports length-constrained segmentation ✂️ min/max chunk length (chars) while preserving semantic chunks - should be great for RAG! Example (≤30 chars): [Landing 5pm → Beimen.] [Let's meet at: Ximen Exit 6.] [Then: Ningxia Night Market...] [Late-night snack!']

Introducing 🪓Segment any Text! 🪓 A new state-of-the-art sentence segmentation tool! Compared to existing tools (and strong LLMs!), our models are far more: 1. efficient ⚡ 2. performant 🔝 3. robust 🚀 4. adaptable 🎯 5. multilingual 🗺

English

2

0

146

Markus Frohmann@FrohmannM·4 Mar

More info + how to use: #new-v22-length-constrained-segmentation" target="_blank" rel="nofollow noopener">github.com/segment-any-te…

English

30

Markus Frohmann@FrohmannM·3 Kas

I'm at #EMNLP2025 in Suzhou this year! Looking forward to connecting with the community after a year's break and spending some time abroad。。。再見！

English

0

16

2.3K

Markus Frohmann@FrohmannM·28 Tem

🤖🗣️ Double Entendre will be presented today, 18:00-19:30 at Hall 4/5 by @m_schedl! Check it out if you're at #ACL2025!

Excited to share two new papers on AI-generated music detection from my research internship at @Deezer, published in @ismir_conf #ISMIR2025 and @aclmeeting #ACL2025 Findings! 🎶🤖 The problem: most AI music detectors are impractical or unreliable in real-world settings.

English

5

0

2

248

Markus Frohmann@FrohmannM·25 Tem

@Deezer @aclmeeting I had a great time working on this with @deezer in Paris! Big thanks to my mentors @evpure, @Gabolsgabs, and @m_schedl! 💻 Code: github.com/deezer/robust-… 📄 ISMIR Paper (foundation): arxiv.org/abs/2506.18488 📄 ACL Paper (Multi-View Double Entendre): aclanthology.org/2025.findings-…

English

1

93

Markus Frohmann@FrohmannM·25 Tem

@Deezer @aclmeeting I view this work as an important extension of current single-modality detectors while maintaining flexibility and modularity. It's not production-ready, but it highlights key paradigms for detection: Using all available information from just the audio and a focus on robustness.

English

0

90

Markus Frohmann@FrohmannM·25 Tem

Excited to share two new papers on AI-generated music detection from my research internship at @Deezer, published in @ismir_conf #ISMIR2025 and @aclmeeting #ACL2025 Findings! 🎶🤖 The problem: most AI music detectors are impractical or unreliable in real-world settings.

English

5

0

3

457

Markus Frohmann@FrohmannM·8 May

Wtpsplit, our text segmentation tool, just reached ⭐️1000 stars⭐️ on GitHub! Excited to see it is proving useful! Check it out here: github.com/segment-any-te… 🎉

Introducing 🪓Segment any Text! 🪓 A new state-of-the-art sentence segmentation tool! Compared to existing tools (and strong LLMs!), our models are far more: 1. efficient ⚡ 2. performant 🔝 3. robust 🚀 4. adaptable 🎯 5. multilingual 🗺

English

1

7

511

Markus Frohmann retweetledi

Benjamin Minixhofer@bminixhofer·2 Nis

We created Approximate Likelihood Matching, a principled (and very effective) method for *cross-tokenizer distillation*! With ALM, you can create ensembles of models from different families, convert existing subword-level models to byte-level and a bunch more🧵

English

2

26

88

6.5K

Markus Frohmann@FrohmannM·11 Kas

Curious about our SoTA text segmentation tool? 🪓 It's gonna help you across all kinds of NLP tasks! Learn more at our poster session: Tuesday, 4pm, Jasmine room at #EMNLP2024! 🗓️ See you there! I'll be attending the whole conference - happy to connect with everyone! 👋

Introducing 🪓Segment any Text! 🪓 A new state-of-the-art sentence segmentation tool! Compared to existing tools (and strong LLMs!), our models are far more: 1. efficient ⚡ 2. performant 🔝 3. robust 🚀 4. adaptable 🎯 5. multilingual 🗺

English

5

238

Markus Frohmann@FrohmannM·29 Eki

Excited to share that I joined @researchdeezer as a research intern to work with @evpure and @Gabolsgabs on detecting AI-generated lyrics !🎶 The first few weeks have been amazing, and I am excited about what is to come—life in Paris certainly has unparalleled charm!

English

6

204

Markus Frohmann@FrohmannM·2 Eki

This was an awesome summer! I can only recommend ETH's summer research fellowship program 🏔️ Also happy about the project's progress - integrating videos into existing architectures is quite exciting, stay tuned! Super grateful to Ryan Cotterell and @glnmario for supervising me.

Excited to share that I joined @ETH Zürich as a summer research fellow, supervised by Prof. @ryandcotterell, working on ✨Multimodal LLMs! ✨ The first few weeks have been a blast, and I'm looking forward to the weeks ahead! 📽️

English

0

6

368

Markus Frohmann retweetledi

Cohere Labs@Cohere_Labs·30 Eyl

Congratulations to C4AI Research Grant recipient @FrohmannM and all authors of "Segment Any Text: A Universal Approach for Robust, Efficient and Adaptable Sentence Segmentation" for their EMNLP acceptance!🥳

This got accepted to #EMNLP 24 Main! ✨ Same goes for our paper on unsupervised debiasing 🤓 ✈️🇺🇲

English