iiinova — where your digital self keeps evolving

Niyati Bafna

@niyatibafna.bsky.social

103 followers 170 following 65 posts

PhD student @jhuclsp. Previously @AIatMeta, @InriaParisNLP, @EM_LCT| #NLProc

Niyati Bafna @niyatibafna.bsky.social · 23/09/2026

We all know about the curse of multilinguality. We know that empirical performance degrades as you add languages to a model. But *in theory*, does it have to? Let’s talk about the theoretical curse of multilinguality for embedding space structure.

1121

Niyati Bafna @niyatibafna.bsky.social · 04/09/2026

Colourful idioms, false friends, garden paths, cultural experience, signboards, menus, dialectal slang, archaic language, minority scripts, self-referential logic, social conventions, audio dialogue, daily life - find it in Last Translation Benchmark. Join us for the next one :-)

040

Reposted by Niyati Bafna

Patricia Schmidtova @patuchen.bsky.social · 17/06/2026

Are you tired of getting meh results from LLMs in your native language and resorting to English instead? We are too!

331

Reposted by Niyati Bafna

Kaiser Sun @kaiserwholearns.bsky.social · 12/03/2026

Multimodal LLMs can read text in images, but why do they often perform worse than when the same text is given as tokens? Our work studies the modality gap of models perceiving text as pixels and shows how to close it. 📄 arxiv.org/abs/2603.09095 🧵👇 #NLProc #LLM #ComputerVision

134

Reposted by Niyati Bafna

changemily.bsky.social @changemily.bsky.social · 24/11/2025

Frustrated with how most of the world’s low-resource languages have NO evaluation resources? 📢 Check out ChiKhaPo, a massively multilingual lexical comprehension and generation benchmark covering 2700+ languages. www.arxiv.org/abs/2510.16928

122

Niyati Bafna @niyatibafna.bsky.social · 29/07/2025

Accepted at ACL main! Come chat about dialectal MT at our poster today at 4 pm. Also, check out this largely bug-free package for generating your own synthetic dialectal data: pypi.org/project/dial...

000

Reposted by Niyati Bafna

Vilém Zouhar @zouhar.bsky.social · 15/07/2025

You have a budget to human-evaluate 100 inputs to your models, but your dataset is 10,000 inputs. Do not just pick 100 randomly!🙅 We can do better. "How to Select Datapoints for Efficient Human Evaluation of NLG Models?" shows how.🕵️ (random is still a devilishly good baseline)

2333

Niyati Bafna @niyatibafna.bsky.social · 04/07/2025

🔈When LLMs solve tasks with a mid-to-low resource input or target language, their output quality is poor. We know that. But can we put our finger on what breaks inside the LLM? We introduce the 💥 translation barrier hypothesis 💥 for failed multilingual generation with LLMs. arxiv.org/abs/2506.22724

2267

Niyati Bafna @niyatibafna.bsky.social · 07/06/2025

We know that speech LID systems flunk on accented speech. But why? And what can we do about it? 🤔 Our work arxiv.org/abs/2506.00628 (Interspeech '25) finds that *accent-language confusion* is an important culprit, ties it to the length of feature that the model relies on, and proposes a fix.

163

Niyati Bafna @niyatibafna.bsky.social · 11/04/2025

Presented DialUp (MT, dialect continua, robustness, etc.; arxiv.org/abs/2501.16581) to some new people this week! Thanks Hale and @schmidtsciences.bsky.social for inviting me up to New York 🥯 Saw some magnolias too :)

020

Niyati Bafna @niyatibafna.bsky.social · 27/02/2025

Dialects lie on continua of (structured) linguistic variation, right? And we can’t collect data for every point on the continuum...🤔 📢 Check out DialUp, a technique to make your MT model robust to the dialect continua of its training languages, including unseen dialects. arxiv.org/abs/2501.16581

1135