18 authors submitted more than 20 papers to #ACL2024NLP!!! How can this be possible? Unless of course they are running paper mills.

#ACL2024NLP in a nutshell by Eduard Hovy (in #@lm_knowledge). Funny but true.

First in-person conference after 2019. Missed many conferences first due to covid and then due to lack of money. Anyway, glad I could present in-person at ACL 2026 and now heading home with many new ideas. Good bye Bangkok.

What's wrong with Google scholar? Just noticed that my citation count has gone down by about 50!

Journal-IEEE Transactions on Audio, Speech and Language Processing Paper submitted - July 2023 Finally accepted and published- March 2025. Journal publication system is broken! ieeexplore.ieee.org/document/10933…

Most-cited papers in the 21st century. ResNet tops the list and Transfomers, ImageNet and Random Forest made it to the top 10. nature.com/articles/d4158…

I am meta reviewing for the current ARR cycle and a worrying number of papers have simply used LLM as judge to evaluate complex domain-specific tasks without any human evaluation. Luckily, most reviewers complained about this. This trend shouldn't continue.

We release SinLlama, an LLM for Sinhala. We continually pre-trained Llama-3-8B with 10 million Sinhala sentences. It always feels good to give back to your mother tongue.

CFP: Journal of Autonomous Agents and Multiagent Systems (JAAMAS) Special Issue: When Foundation Models Meet Multi-Agent Systems Guest editors: Stephen Cranefield, Bastin Tony Roy Savarimuthu, Shuyue He, Surangika Ranathunga More info: link.springer.com/journal/10458/…

Excited to share our #EMNLP2025 paper: "Improving the quality of Web-mined Parallel Corpora of Low-Resource Languages using Debiasing Heuristics" We analyse disparities across multiPLMs for parallel data curation & propose a heuristic-based solution. arxiv.org/abs/2502.19074

We met Prof Pushpak only once in 2019. But after 3 years, in 2022, he was thinking about our safety. Rest in peace, beautiful soul 🙏

It was pretty much a fun project to take part on with Nisansa de Silva. Went back down the memory lane to find unique examples from the Sri Lankan culture. We attest to the 100% quality of the Sinhala data.

What a cool idea. Now even the journals can offer a similar tool for the authors to check the quality of their papers before doing the actual submission. This way, it would significantly reduce the load on human reviewers who have to review low-quality papers.