MT Group at FBK

🎉🎓Congratulations to our PhD student @dennisfucci.bsky.social on a very successful thesis defense! 👏

Many thanks to the evaluation committee members @deboranozza.bsky.social, Mirco Ravanelli, and Leonardo Badino for their insightful feedback and appreciation of his work!

#nlproc

@bsavoldi.bsky.social from @fbk-mt.bsky.social will present Translation in the Hands of Many: Centering Lay Users in Machine Translation Interactions at the poster session on Wed Nov. 5th, 11:00-12:30 in Hall C

Our #PickOfTheWeek by @beomseok-lee.bsky.social: "Can Speech LLMs Think while Listening?" by Yi-Jen Shih, @rdesh26.bsky.social, Chunyang Wu, Wei Zhou, SK Bong, Yashesh Gaur, Jay Mahadeokar, Ozlem Kalinli, Mike Seltzer (2025).

#Speech #SpeechLLM #LLM #SpeechTech #AI

Our next presentation is by @sarapapi.bsky.social: "How real is your real-time simultaneous speech-to-text translation system?"

Look for the answer in her TACL paper: direct.mit.edu/tacl/article...

#lt2025fbk

Our Marco Gaido presenting FAMA, the first family of large-scale open-science speech foundation models for English and Italian.

Joint work with the @speechtekfbk.bsky.social group.

Data, code, models publicly available, check all info in the paper:
clic2025.unica.it/wp-content/u...

#lt2025fbk

@bsavoldi.bsky.social presenting our new multilingual benchmark for evaluating LLMs on gender-neutral translation.

Catch our paper at #EMNLP2025
ℹ️ arxiv.org/pdf/2501.09409

#lt2025fbk

Now it's the turn of our @dennisfucci.bsky.social presenting the #ACL2025NLP paper on explaining gender bias in speech translation

📖 aclanthology.org/2025.acl-sho...
#lt2025fbk

The Language Technology at FBK workshop has just started with a truly insightful talk by @deboranozza.bsky.social: "A Roadmap for the Everyday Use of LLMs: Emerging Risks and Research Directions" #LT2025FBK

Our pick of the week by @linaconti.bsky.social: "Lost in Transcription, Found in Distribution Shift: Demystifying
Hallucination in Speech Foundation Models" by Hanin Atwany, @abdulwaheed.bsky.social, Rita Singh, Monojit Choudhury, and Bhiksha Raj (ACL Findings 2025)

aclanthology.org/2025.finding...

LT Highlights @ FBK 2025

🚀 Join us for the LT@FBK day 2025! Discover cutting-edge research and highlights in speech and language technologies from Fondazione Bruno Kessler (FBK)

📅 October 28, 2025
📍FBK, Trento
ℹ️ lt-highlights.fbk.eu

Our pick of the week by @bsavoldi.bsky.social: "Acoustic-based Gender Differentiation in Speech-aware Language Models" by Junhyuk Choi, Jihwan Seol, Nayeon Kim, Chanhee Cho, EunBin Cho, Bugeun Kim.

arxiv.org/abs/2509.21125

#Gender #SpeechLLM #Speech

Language Technology Research Highlights 2025 The Language Technology Research Highlights 2025 (LT@FBK2025) event aims to bring together scientists, students, practitioners, and enthusiasts who are interested in language technologies and want to ...

🚀 Our annual, full-day Language Technologies showcase is back! Dive into the latest research highlights from FBK groups. Want in? We'd love to see you, but don't forget to register!

www.fbk.eu/en/event/346...

Marco Gaido introducing SimulStream, an #OpenSource Tool for Simultaneous #Speech #Translation 🗣️🖥️📝 at the DI Center Demo Day at FBK!

The tool is going to be released soon. Stay tuned! 👀

Marco Gaido and Roldano Cattoni presenting our SimulStream Demo at the DI Center Demo Day at FBK!

The open-source tool, which is going to be released soon, natively supports any speech-to-text #HuggingFace models! 🤖

#SpeechTech #Translation

The Unheard Alternative: Contrastive Explanations for Speech-to-Text Models Contrastive explanations, which indicate why an AI system produced one output (the target) instead of another (the foil), are widely regarded in explainable AI as more informative and interpretable th...

🎉 Excited to share that my paper "The Unheard Alternative" was accepted to @blackboxnlp.bsky.social 2025!
We introduce contrastive explanations for speech-to-text, identifying which audio features ST models use to assign a grammatical gender to the speaker.
📄 Preprint: arxiv.org/abs/2509.265...

Our very own @sarapapi.bsky.social presenting FAMA at #clicit2025:

📗Paper: clic2025.unica.it/wp-content/u...
🔗 Models: hf.co/collections/...
📊 Data: hf.co/datasets/FBK...
💻 Code: github.com/hlt-mt/FBK-f...

Joint work with @speechtekfbk.bsky.social

🚀 Excited to present FAMA, the first large-scale #OpenScience #Speech foundation model for 🇮🇹 Italian & 🇬🇧 English, at #clicit2025 (17:30–18:45 oral session)!

🔗 Models: hf.co/collections/...
📊 Data: hf.co/datasets/FBK...
💻 Code: github.com/hlt-mt/FBK-f...
📄 Preprint: arxiv.org/pdf/2505.22759

We are on our way to Casteddu for #clicit2025 with a guest from @fbk-mt.bsky.social @ailc-nlp.bsky.social

Our pick of the week by @sarapapi.bsky.social: "Retrieval-Augmented Generation for AI-Generated Content: A Survey" by Penghao Zhao, Hailin Zhang, Qinhan Yu, Zhengren Wang, Yunteng Geng, Fangcheng Fu, Ling Yang, Wentao Zhang, Jie Jiang, Bin Cui.

arxiv.org/pdf/2402.19473

#RAG #survey

Context-Driven Dynamic Pruning for Large Speech Foundation Models Speech foundation models achieve strong generalization across languages and acoustic conditions, but require significant computational resources for inference. In the context of speech foundation mode...

Our pick of the week by Marco Gaido: "Context-Driven Dynamic #Pruning for Large #Speech #Foundation Models" by Masao Someki, Shikhar Bharadwaj, Atharva Anand Joshi, Chyi-Jiunn Lin, Jinchuan Tian, Jee-weon Jung, @shinjiw.bsky.social, et al. #INTERSPEECH2025.

arxiv.org/abs/2505.18860

Our pick of the week by @zhihangxie.bsky.social: "SimulMEGA: MoE Routers are Advanced Policy Makers for Simultaneous Speech Translation" by Chenyang Le, Bing Han, Jinshun Li, Songyong Chen, and Yanmin Qian (2025)

#Speech #Simultaneous #Translation #MOE #SpeechTech

Our pick of the week by @beomseok-lee.bsky.social: "Speech Discrete Tokens or Continuous Features? A Comparative Analysis for Spoken Language Understanding in SpeechLLMs" by Dingdong Wang, Junan Li, Mingyu Cui, Dongchao Yang, Xueyuan Chen, and Helen Meng (EMNLP 2025)

Our pick of the week by @linaconti.bsky.social: "I Have No Mouth, and I Must Rhyme: Uncovering Internal Phonetic Representations in LLaMA 3.2" @jackmerullo.bsky.social, Arjun Khurana, Oliver McLaughlin (ICML 2025 Workshop on Assessing World Models)

arxiv.org/abs/2508.02527

#XAI #LLM

Finally, we contributed to "NUTSHELL: A Dataset for Abstract Generation from Scientific Talks" presented by @maikezufle.bsky.social from @ai4lt.bsky.social

👉 aclanthology.org/2025.iwslt-1...
(6/6)

Marco Gaido took a deep dive into "The Warmup Dilemma" @iwslt.bsky.social

👉 aclanthology.org/2025.iwslt-1...
(5/6)

We organized 5 tasks at #IWSLT:
📌 Offline
📌 Simultaneous
📌 Subtitling
📌Model compression
📌 Instruction following

👉 iwslt.org/2025/#shared...
(4/6)

@dennisfucci.bsky.social shared insightful findings on gender bias through the lens of interpretability: "Different speech translation models encode and translate speaker gender differently"

👉 aclanthology.org/2025.acl-sho...
(3/6)

@sarapapi.bsky.social presented her TACL paper: “How real is your real-time simultaneous speech-to-text translation system?”

👉 aclanthology.org/2025.tacl-1.14/
(2/6)

Heading home after an exciting and intense @aclmeeting.bsky.social in Vienna! We had a great time presenting our work and connecting with the community.

Thanks to everyone who came by!

#acl2025 #nlproc
(1/6)