Ashwanth retweeted
ACL2024 (@aclmeeting) paper posts. If you are attending ACL in person, then please do drop by.
Paper #1:
🚀 Excited to share our research paper: "An Empirical Study of In-context Learning in LLMs for Machine Translation", which we'll be presenting at ACL 2024!
📄Paper: arxiv.org/abs/2401.12097
🛠️Github: github.com/PranjalChitale/in…
🗓️Slides: docs.google.com/presentation…
▶️Video: underline.io/events/466/sess…
Our work delves into the ICL capabilities of LLMs for MT, including focusing on how different aspects of demonstration impact performance.
joint work w/ @jaygala24 @pranjalchitale
Ashwanth retweeted
I will be attending @aclmeeting #ACL2024 in person in Thailand.
If you are interested in Multilinguality (evaluation/meta-evaluation/LLMs/Machine Translation), Efficient Methods in NLP, do reach out for a chat.
Looking forward to meeting the community!
Ashwanth retweeted
gofund.me/87383192
Please 🙏 Support Dr. Harika Jetti’s Family in their Time of Loss 🥹
@Swathireddytdp @Shiva4TDP @naralokesh @MilagroMovies @M9News_ @turagasudhakar @TdpVenkata @JaiTDP @PemmasaniOnX
#Harikajetti #USA #TDP #TeluguDesam @ysjagan @YSRCParty #NaraLokesh
Ashwanth retweeted
📣 📣 📣 New instruction-tuned LLM! 📣 📣 📣
Today, we announce an initial release of "Airavata", an instruction-tuned LLM for Hindi.
Blog: ai4bharat.github.io/airavata…
Model: huggingface.co/ai4bharat/Air…
Datasets: huggingface.co/datasets/ai4b…
(1/N)
Ashwanth retweeted
🚨🚨🚨 Romanization Based Multilingual LLM Adaptation Technique
Presenting "RomanSetu", a technique to exploit Romanization to efficiently unlock or bridge the multilingual capabilities of LLMs like LLaMA. Setu means bridge in Hindi.
arxiv.org/abs/2401.14280
Ashwanth retweeted
It was almost 10 months since I posted by first tweet:
teapaw.org/VarunGumma23/status/16…
And now, we finally were able to do it (a late post though). Do check out our ported IndicTrans2 and IndicTrans2-Distilled models on HuggingFace.
@jaygala24 @pranjalchitale @prajdabre1
Ashwanth retweeted
Finally, you can read the TMLR accepted version. You don't need to be FAANG to create SOTA models if you focus on a niche and keep grinding!
@ai4bharat @jaygala24 @pranjalchitale @anoopk @MiteshKhapra @pratykumar @oneraghavan @sumanthd17 @VarunGumma23 @NameIsAshwanth @vivekrag
IndicTrans2: Towards High-Quality and Accessible Machine Translation Models for all 22 Scheduled ...
Jay Gala, Pranjal A Chitale, A K Raghavan et al..
Action editor: W Ronny Huang.
openreview.net/forum?id=vfT4…
#corpus #multilingual #corpora
Ashwanth retweeted
I am extremely pleased to announce that IndicTrans2 will be published in TMLR (@TmlrOrg). This is a tremendous achievement for my coauthors and me that took nearly 1.5 years of hard work. The camera ready version will be out soon but for now we are over the moon!
#NLProc #ACL
🚨 📢 Preprint Alert
After more than a year of hard work, we are pleased to introduce IndicTrans2, the first machine translation system supporting all 22 scheduled Indic languages.
📎: arxiv.org/abs/2305.16307
💻: github.com/AI4Bharat/IndicTr…
▶️: models.ai4bharat.org/#/nmt/v…
Thread👇[1/n]
Ashwanth retweeted
Led my first work at @MSFTResearch where we build upon amazing work done on MEGA and introduce MEGAVERSE, where we benchmark new tasks and languages on state-of-the-art commercial and open-source models. (1/n)
We learn a regression model, CTQ Scorer (Contextual Translation Quality), which selects examples based on multiple features to maximize machine translation quality in few-shot prompting.
1
1
261
Ashwanth retweeted
🚨 📢 Preprint Alert
After more than a year of hard work, we are pleased to introduce IndicTrans2, the first machine translation system supporting all 22 scheduled Indic languages.
📎: arxiv.org/abs/2305.16307
💻: github.com/AI4Bharat/IndicTr…
▶️: models.ai4bharat.org/#/nmt/v…
Thread👇[1/n]
Ashwanth retweeted
Please retweet!
Job opening at NICT, Japan.
Post: Technical Researcher
Topic: Indic Natural Language Generation
Minimum qualification: Masters in CS (plans to pursue a PhD and some publications is a plus)
nict.go.jp/en/employment/ind…
www2.nict.go.jp/employment/t…
#NLP
#NLProc












