After getting published in ICLR as an Independent Researcher, I have received nearly 100 messages from others who are looking to do the same. So I wrote a blog post on why I decided to do it and my advice to others. medium.com/@andreas_madse…

arxiv.org/abs/2001.05140 I'm pretty confused. with this level of details, they have presumably run some experiments already to be certain about this approach, and somehow couldn't wait a couple of weeks to put results in the manuscript.

I see many papers that begin with a sentence equivalent to "Topic X is popular". Popularity is not a sound scientific reason for studying a topic, so such opening sentences strike me as lame. How about "This paper shows how to solve issue Y with method M for X"? 1/2

I think our "Neural Arithmetic Units" ICLR paper, provides a nice list of sanity checks when developing a new unit. Looking at: initialization, gradients, loss space, and redundant parameters, are generally important. I hope to see more of this :) - openreview.net/forum?id=H1gNO…

NLP research has focused a lot on SOTA chasing recently We got better metrics but with a worrisome increase in energy consumption to use/train models One reason: the lack of incentives to keep models efficient & find optimal perf/efficiency tradeoffs Let's try to change this👇

How can you successfully train transformers on small datasets like PTB and WikiText-2? Are LSTMs better on small datasets? I ran 339 experiments worth 568 GPU hours and came up with some answers. I do not have time to write a blog post, so here a twitter thread instead. 1/n

UPDATE: We have spent the past month “fine-tuning” our approach for Closed Book QA (CBQA, no access to external knowledge) w/ T5 and now our appendix is overflowing with interesting results and new SoTAs on open domain WebQuestions and TriviaQA! arxiv.org/abs/2002.08910 (1/7)

📢 Exciting update! If you're seeking additional resources and a deeper understanding of the AToMiC track, we've got you covered. Development queries and relevance labels are now available: trec-atomic.github.io/annoucements/d… 📚🔍 [1/4]

That's a wrap! The Waterloo (Waterloo's Cheriton School of Computer Science) team had fun attending the ACL 2023 Conference in Toronto, Canada! #ACL2023NLP 🇨🇦 We would like to congratulate ralphtang.eth Linqing Liu Gin Jiang Jimmy Lin et al. for winning the Best Paper Award at ACL 2023!!🏆 Next stop is SIGIR 2023.

We also have Jheng-Hong Yang presenting AToMiC our cross modal resource paper at #SIGIR2023 in collaboration with Jimmy Lin Miriam Redi Krishna Srinivasan Rafael S Rezende and Stéphane Clinchant

📢 Calling all enthusiasts of image-text cross-modal retrieval for multimedia content creation! 📷📝 We've extended the submission deadline for TREC-AToMiC to August 7th, 9:00 am (EST) ⏳ Need support? Reach out #TREC2023 #AToMiC #MultimediaRetrieval

Excited to delve into the world of AToMiC today! 🚀 Grateful to TwelveLabs (twelvelabs.io) for the invite! 🙌 #AToMiC #Innovation >> twelvelabs-20253029.hs-sites.com/multimodal-wee…

RAG is all the RAGe these days, but we (still) don't quite know how to evaluate it properly... This year, we are taking a stab at it in the context of TREC, building on 30+ years of experience in evaluating IR systems. trec-rag.github.io

Our paper on understanding variability in text-to-image models was accepted at #EMNLP2024 main track! Lots of thanks to my collaborators Xinyu Crystina Zhang Yao Lu Wenyan Li Ulie Xu and mentors Jimmy Lin Pontus Ferhan Ture. Check out w1kp.com