--- title: "Kushal Lakhotia — author profile and papers" description: "Kushal Lakhotia: 3 papers on Elamur Research (2021), h-index 15 — affiliations, topics, co-authors and who cites them in our library." canonical: "https://research.elamur.ai/free/en/authors/kushal-lakhotia/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/kushal-lakhotia/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / Kushal Lakhotia # Kushal Lakhotia Kushal Lakhotia works on self-supervised learning for [speech and spoken language](https://md.research.elamur.ai/free/en/topics/speech/). The papers here, all from 2021, ask how models can pick up useful structure from raw audio that has no transcripts. They cover masked prediction over audio frames, scaling pretraining across many languages, and modeling spoken language without any text. The work connects to [language models and pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) and to multilingual speech processing, and it builds on Transformer encoders of the wav2vec 2.0 kind. The most cited of these works is [HuBERT](https://md.research.elamur.ai/free/en/papers/hsu-2021-hubert-self-supervised-speech-representation/). It clusters acoustic frames offline, starting from k-means on MFCC features, and treats the resulting labels as targets that the encoder must predict for masked spans of audio. The clusters are then refined using features the network itself has learned. [XLS-R](https://md.research.elamur.ai/free/en/papers/babu-2021-xls-r-self-supervised-cross/) pushes cross-lingual pretraining to models of up to 2B parameters, trained on 436K hours of publicly available speech in 128 languages, with sampling that gives low-resource languages extra weight. A further 2021 paper, [On Generative Spoken Language Modeling from Raw Audio](https://md.research.elamur.ai/free/en/papers/lakhotia-2021-generative-spoken-language-modeling-raw/), turns from learning representations to generating speech. Its main contribution is a suite of evaluation measures covering both encoding and generation, at the level of sound and at the level of language. Generated audio is scored by transcribing it with a pretrained recognizer, and the measures are compared with human ratings and with lighter zero-shot tests drawn from earlier benchmark challenges on unlabeled speech. Bylines on this work list [Meta](https://md.research.elamur.ai/free/en/companies/meta/) in Menlo Park for 2021 and 2022, and two papers from 2022 also name the [École des hautes études en sciences sociales](https://md.research.elamur.ai/free/en/labs/ecole-des-hautes-etudes-en-sciences-sociales/) in Paris, which fits the spoken language modeling work. Recurring co-authors include [Wei-Ning Hsu](https://md.research.elamur.ai/free/en/authors/wei-ning-hsu/), [Abdelrahman Mohamed](https://md.research.elamur.ai/free/en/authors/abdelrahman-mohamed/), Alexei Baevski and Benjamin Bolte, with further collaboration involving [Emmanuel Dupoux](https://md.research.elamur.ai/free/en/authors/emmanuel-dupoux/), Alexis Conneau and Juan Pino. Kushal Lakhotia has 3 papers in our library, in 2021, most of them on [Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/), [Speech & spoken language](https://md.research.elamur.ai/free/en/topics/speech/), [Multilingual & cross-lingual NLP](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/); often with [Abdelrahman Mohamed](https://md.research.elamur.ai/free/en/authors/abdelrahman-mohamed/), [Alexei Baevski](https://md.research.elamur.ai/free/en/authors/alexei-baevski/), [Benjamin Bolte](https://md.research.elamur.ai/free/en/authors/benjamin-bolte/). 3 papers in our library · 2021 [OpenAlex](https://openalex.org/A5050838145) · [**h-index 15**](https://openalex.org/authors/A5050838145) · [31 works](https://openalex.org/works?filter=author.id:A5050838145) · [4 910 citations](https://openalex.org/works?filter=author.id:A5050838145&sort=cited_by_count:desc) OpenAlex, 2026-10-14 [3 works in our library](https://md.research.elamur.ai/free/en/authors/kushal-lakhotia/#papers) · [906 citations in our library](https://md.research.elamur.ai/free/en/authors/kushal-lakhotia/cited-by/) · 1 more work in our library · [⭐ Paper Star rank #588](https://md.research.elamur.ai/free/en/paper-stars/page/6/#star-588) The profile is the library's match of the name; two people with one name are not always told apart. Elsewhere: [Google Scholar](https://scholar.google.com/citations?view_op=search_authors&mauthors=Kushal+Lakhotia) · [Semantic Scholar](https://www.semanticscholar.org/search?q=Kushal+Lakhotia) · [dblp](https://dblp.org/search?q=Kushal+Lakhotia) · [ACL Anthology](https://aclanthology.org/search/?q=Kushal+Lakhotia) ## Affiliations - [Meta (United States)](https://md.research.elamur.ai/free/en/companies/meta/) · Menlo Park, United States —2021–2022 on 16 papers - [École des hautes études en sciences sociales](https://md.research.elamur.ai/free/en/labs/ecole-des-hautes-etudes-en-sciences-sociales/) · Paris, France —2022 on 2 papers From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine. 2 more institutions on single bylines are not listed. ## Topics [Language models & pretraining 1](https://md.research.elamur.ai/free/en/topics/language-models/) · [Speech & spoken language 1](https://md.research.elamur.ai/free/en/topics/speech/) · [Multilingual & cross-lingual NLP 1](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/) ## Works by year ### 2020s [**2021** 3 papers](https://md.research.elamur.ai/free/en/authors/kushal-lakhotia/year/2021/) ## Papers The most cited first (OpenAlex and our library). 1. **[HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://md.research.elamur.ai/free/en/papers/hsu-2021-hubert-self-supervised-speech-representation/)** *(2021 · 3 441 citations)* Unsupervised speech representation learning is harder than in text or vision for three reasons. Each utterance contains many sounds, there is no ready-made inventory of discrete units, and unit boundaries are unknown. HuBERT (Hidden-Unit BERT) handles this by first clustering acoustic frames offline, starting with… 2. **[XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale](https://md.research.elamur.ai/free/en/papers/babu-2021-xls-r-self-supervised-cross/)** *(2021 · 799 citations)* Self-supervised speech models had been scaled mainly for English or in supervised setups. Cross-lingual ones stayed small: XLSR-53 had about 300M parameters and roughly 50K hours of audio. XLS-R tests how far this approach can go. The authors pretrain wav2vec 2.0 Transformers with 0.3B, 1B and 2B parameters on 436K… 3. **[On Generative Spoken Language Modeling from Raw Audio](https://md.research.elamur.ai/free/en/papers/lakhotia-2021-generative-spoken-language-modeling-raw/)** *(2021 · 165 citations)* The paper introduces Generative Spoken Language Modeling (GSLM): learning both the sound system and the linguistic structure of a language directly from raw audio, without transcripts or labels. ## Also in our library 1 more work of theirs is in our library without a page here yet; it will appear as we publish them. ## Co-authors here 12 of 18 co-authors with a page, the most papers together first. - [Abdelrahman Mohamed](https://md.research.elamur.ai/free/en/authors/abdelrahman-mohamed/) ×2 - [Alexei Baevski](https://md.research.elamur.ai/free/en/authors/alexei-baevski/) ×2 - [Benjamin Bolte](https://md.research.elamur.ai/free/en/authors/benjamin-bolte/) ×2 - [Wei-Ning Hsu](https://md.research.elamur.ai/free/en/authors/wei-ning-hsu/) ×2 - [Alexis Conneau](https://md.research.elamur.ai/free/en/authors/alexis-conneau/) - [Andros Tjandra](https://md.research.elamur.ai/free/en/authors/andros-tjandra/) - [Arun Babu](https://md.research.elamur.ai/free/en/authors/arun-babu/) - [Changhan Wang](https://md.research.elamur.ai/free/en/authors/changhan-wang/) - [Emmanuel Dupoux](https://md.research.elamur.ai/free/en/authors/emmanuel-dupoux/) - [Eugene Kharitonov](https://md.research.elamur.ai/free/en/authors/eugene-kharitonov/) - [Jade Copet](https://md.research.elamur.ai/free/en/authors/jade-copet/) - [Juan Pino](https://md.research.elamur.ai/free/en/authors/juan-pino/) ## Cited by Papers of our library cite their works 906 times (of the 140 884 papers whose bibliographies we read); [58 citing papers with a page here →](https://md.research.elamur.ai/free/en/authors/kushal-lakhotia/cited-by/).