Aditya Yadavalli (@AdityaYadavall2) header
Aditya Yadavalli (@AdityaYadavall2) profile picture

Aditya Yadavalli

@AdityaYadavall2

PhD Student @UCSanDiego | Formerly Speech/NLP Engineer @karya_inc, CS + Linguistics @iiit_hyderabad

adityayadavalli1.github.io
Joined May 2021
  • Tweets77
  • Following351
  • Followers599
3Media
Photo by Aditya Yadavalli (@AdityaYadavall2)Photo by Aditya Yadavalli (@AdityaYadavall2)
What can LLMs tell us about the nature of spoken communication? 🗣️🤖 From an info-theory perspective, speech has two main channels: words (text) and melody of speech (prosody). We propose a new method using spoken LLMs to quantify how these channels conveys meaning (1/n)
Title of the paper, "What Do Prosody and Text Convey? Characterizing How Meaningful Information is Distributed Across Multiple Channels", authors, and the venn diagram depicting the informational relationship between the linguistic channels (prosody P, text T, Audio A) and meaningful feature Y.

ALT Title of the paper, "What Do Prosody and Text Convey? Characterizing How Meaningful Information is Distributed Across Multiple Channels", authors, and the venn diagram depicting the informational relationship between the linguistic channels (prosody P, text T, Audio A) and meaningful feature Y.

1
12
34
3,273
Aditya Yadavalli retweeted
Jonny Cook (@JonnyCoook) profile picture
Jonny Cook@JonnyCoookOct 28
Internship opportunity! I'll be hosting a Student Researcher next year @GoogleDeepMind. Humans got to the moon with roughly the same brains as our ancient ancestors; culture did the work. Can a population of frontier agents compound their abilities by accumulating artifacts?
16
30
5
537
37,513
Aditya Yadavalli retweeted
EleutherAI (@AiEleuther) profile picture
EleutherAI@AiEleutherOct 24
nitter.cf/realshivamsingh/status…
Shivam Singh (@realshivamsingh) profile pictureShivam Singh@realshivamsinghSep 14
1/ Every language is unique, but the big AI companies release a single model to cover dozens of languages at once. BuzzASR 🐝 takes a different approach: We release a swarm of 102 monolingual speech recognition models, beating Whisper on 77 languages and achieving SOTA on 27.
Image from Shivam Singh (@realshivamsingh)
1
1
3
337
Aditya Yadavalli retweeted
EleutherAI (@AiEleuther) profile picture
EleutherAI@AiEleutherOct 24
Come say hi to us at EMNLP! @Aflah02101 @linguist_cat
Image from EleutherAI (@AiEleuther)
1
4
21
2,448
Aditya Yadavalli retweeted
Kanishka Misra 🌊 (@kanishkamisra) profile picture
Kanishka Misra 🌊@kanishkamisraOct 19
Come join the 4 of us and answer exciting research questions! I’m looking for PhD students who want to: 1) understand the role of input in shaping generalization behavior — how do different sources of positive and negative evidence conspire together to give rise to learning?…
Lindia Tjuatja (@lltjuatja) profile pictureLindia Tjuatja@lltjuatjaOct 19
The UT comp ling faculty (@jessyjli, @kanishkamisra, @kmahowald & myself) are looking for PhD students to join us! sites.utexas.edu/compling/ 🤘 We are an interdisciplinary group with interests across linguistics, cogsci and NLP. Apply by 12/1: liberalarts.utexas.edu/lingu…
1
12
1
58
5,047
My personal takeaway: the modeling techniques are there and enough data (surprisingly!) is already there. Someone just had to put in the effort to build language-specialized ASR models. What a paper to start your research journey, @realshivamsingh!
Shivam Singh (@realshivamsingh) profile pictureShivam Singh@realshivamsinghSep 14
1/ Every language is unique, but the big AI companies release a single model to cover dozens of languages at once. BuzzASR 🐝 takes a different approach: We release a swarm of 102 monolingual speech recognition models, beating Whisper on 77 languages and achieving SOTA on 27.
Image from Shivam Singh (@realshivamsingh)
1
1
7
529
Aditya Yadavalli retweeted
EleutherAI (@AiEleuther) profile picture
EleutherAI@AiEleutherSep 15
If you’re looking for SOTA ASR capabilities in a language other than Chinese or English, we got you covered!
Shivam Singh (@realshivamsingh) profile pictureShivam Singh@realshivamsinghSep 14
1/ Every language is unique, but the big AI companies release a single model to cover dozens of languages at once. BuzzASR 🐝 takes a different approach: We release a swarm of 102 monolingual speech recognition models, beating Whisper on 77 languages and achieving SOTA on 27.
Image from Shivam Singh (@realshivamsingh)
1
4
8
2,465
Aditya Yadavalli retweeted
Alex Warstadt (@a_stadt) profile picture
Alex Warstadt@a_stadtSep 14
🐝New preprint🐝 If you want open source speech technology that's actually optimized for your target language, BuzzASR might be your best option! All models on Huggingface 🤗 Congrats to @realshivamsingh on an amazing first paper!! Appearing in EMNLP Findings.
Shivam Singh (@realshivamsingh) profile pictureShivam Singh@realshivamsinghSep 14
1/ Every language is unique, but the big AI companies release a single model to cover dozens of languages at once. BuzzASR 🐝 takes a different approach: We release a swarm of 102 monolingual speech recognition models, beating Whisper on 77 languages and achieving SOTA on 27.
Image from Shivam Singh (@realshivamsingh)
1
2
21
1,762
We have a preprint out our EMNLP Findings paper plus over 100 ASR models, each adapted for a single language. for Well done @realshivamsingh for pulling this off!
Shivam Singh (@realshivamsingh) profile pictureShivam Singh@realshivamsinghSep 14
1/ Every language is unique, but the big AI companies release a single model to cover dozens of languages at once. BuzzASR 🐝 takes a different approach: We release a swarm of 102 monolingual speech recognition models, beating Whisper on 77 languages and achieving SOTA on 27.
Image from Shivam Singh (@realshivamsingh)
1
5
34
1,876
Aditya Yadavalli retweeted
Shivam Singh (@realshivamsingh) profile picture
Shivam Singh@realshivamsinghSep 14
1/ Every language is unique, but the big AI companies release a single model to cover dozens of languages at once. BuzzASR 🐝 takes a different approach: We release a swarm of 102 monolingual speech recognition models, beating Whisper on 77 languages and achieving SOTA on 27.
Image from Shivam Singh (@realshivamsingh)
3
8
6
26
10,236
Aditya Yadavalli retweeted
Nathan Godey @COLM 🌉 (@nthngdy) profile picture
Nathan Godey @COLM 🌉@nthngdyAug 19
🧵New paper: Matryoshka LM Suites Instead of training all models of a suite independently, we nest them within a single architecture and train the suite in a single run => much less compute, same performance Also allows free distillation, better speculative decoding & more 👇
Image from Nathan Godey @COLM 🌉 (@nthngdy)
44
147
23
1,382
74,741
89 languages sota speech models. There's plenty of speech data it appears, so the simplest fine-tuning plus tokenization on public data just improves everything substantially. And it's not even with any tricks or all the data... #conll #acl
Image from Leshem (Legend) Choshen 🤖🤗 @ACL @ICML (@LChoshen)
2
4
10
1,398
Aditya Yadavalli retweeted
Saujas Vaduguru (@saujasv) profile picture
Saujas Vaduguru@saujasvJun 30
I'll be at #ACL2026 presenting our work on convention formation! I'm thinking about collaborating and communicating with agents these days, and am excited to chat with others working on these problems! Find us at Poster Session E on Monday, July 6 to check out the work!
Saujas Vaduguru (@saujasv) profile pictureSaujas Vaduguru@saujasv10 Nov 2025
People adapt their language to communicate more efficiently over time. How can we make models do this? In our recent work, we trained models in self-play, and found that using the right incentives can make models adapt to communicate efficiently even without human demonstrations.
Image from Saujas Vaduguru (@saujasv)
1
3
20
1,413
Aditya Yadavalli retweeted
XiulinYang (@xiulin_yang) profile picture
XiulinYang@xiulin_yangApr 21
🎉 Happy to share that our paper on function words & language learning (w/ Heidi Getz & @weGotlieb) is accepted to #ACL2026! A little late to the party, but still worth celebrating 🥳 We ask: what statistical properties help a learner abstract grammatical knowledge from linear input? Turns out function words, though often overlooked, play an important role. Check out our updated preprint: arxiv.org/pdf/2601.21191 🧵 1/4
Image from XiulinYang (@xiulin_yang)
2
14
1
75
5,848
While we focused on some specific case studies in English, our methodology can be applied elsewhere. Prosodic functions differ cross-linguistically 🌍, and our approach can be applied to other (para)linguistic channels too, like facial expressions  🎭. (5/n)
1
1
283
For more details, check out our preprint: arxiv.org/pdf/2512.16832. I had a blast collaborating on this with @tpimentelms, @tamaregev, @weGotlieb, and @a_stadt! (n/n)
1
279
Aditya Yadavalli retweeted
Ishaan Watts (@IshaanWatts18) profile picture
Ishaan Watts
@IshaanWatts1825 Jun 2024
🚀 Excited to share my latest work at @MSFTResearch, "PARIKSHA: A Large-Scale Investigation of Human-LLM Evaluator Agreement on Multilingual and Multi-Cultural Data"! 🌍. (1/n)
1
6
40
6,423
Load more