Akari Asai
Ph.D. student @ Paul G. Allen School of Computer Science & Engineering, University of Washington
Visiting Student Researcher @ Meta AI
I am currently in my 5th year of pursuing a Ph.D. in NLP at Paul G. Allen School of Computer Science & Engineering, University of Washington. I am fortunate to be advised by Prof. Hannaneh Hajishirzi. I am also spending some time at Meta AI Research as a visiting student researcher, under the supervision of Dr. Wen-tau Yih. Prior to joining UW, I obtained a B.E. in Electrical Engineering and Computer Science from The University of Tokyo, Japan.
My primary research interests are centered around natural language processing and machine learning. My research is driven by a desire to create NLP systems that are both practical and impactful, with the ultimate goal of improving access to information for people from all walks of life. Recently, my research has focused on several key areas, including:
-
Retrieval-augmented LMs – I’ve pioneered and advanced the area of retrieval-augmented LMs to build more reliable, adaptable and attributable intelligent systems: Self-RAG (ICLR 2024), Adaptive Retrieval-augmented LM (ACL 2023), Evidentiality-guided RAG (NAACL 2022), FAVA (Retrieval-augmented fine-grained hallucination Detection). I’ve co-taught the first tutorial of retrieval-augmented LMs at ACL 2023. See our latest position paper, Reliable, Adaptive and Attributable LMs with Retrieval (Asai et al., 2024) on why retrieval-augmented LMs could be the next generation of LMs and we should contribute to their advancements together.
-
General-purpose knowledge retrieval systems – Reliable retrieval systems are the key to build powerful and successful retrieval-augmented LMs. I’ve developed advanced search and representation systems that better capture complex real-world user queries, robust to diverse domains, and are more efficient: Path Retriever (ICLR 2020), the first Instructable Retriever (Findings of ACL 2023), LUKE (EMNLP 2020) and Binary Passage Retriever (ACL 2021). Our methods have been integrated into multiple real-world systems, such as COVID-19 Research Search.
-
Serving intelligent systems for underrepresented populations – Today’s NLP systems are primarily developed and tested on a handful of high-resource languages, leaving many world languages left behind. I believe developing NLP systems accessible for everyone is important. I’ve been contributing to the area of multilingual NLP, in particular, advancing our understanding of NLP in low-resource setups and developing systems that can help linguistic minorities to access the information they need: XOR QA (NAACL 2021), CORA (NeurIPS 2021), BUFFET (NAACL 2024) and Large-scale meta survey on multilingual resources (Findings of EMNLP 2022). I was the lead organizer of Workshop on Multilingual Information Access at NAACL 2022 and hosted the first cross-lingual retrieval and open-domain QA shared task. This line of my work was featured in UW CSE News.
I am also passionate about teaching, mentoring and helping students to learn research, especially students from underrepresented groups. I have been the Head TA for CSE473: Intro to AI (undergrad) and CSE599J: Data-centric ML (grad) at UW. To reduce the barrier to start research or Ph.D. in this area, I’m hosting weekly office hours open to everyone (please sign up from Calendly!), and am a mentor for UW CSE Ph.D. Pre-Application Mentorship Service (PAMS).
news
Apr 19, 2024 | Gave an invited lecture on state of the art of retrieval-augmented generation (slides) at CMU Advanced NLP class! |
---|---|
Mar 15, 2024 | Our BUFFET, a large-scale few-shot cross-lingual transfer benchmark across 54 langauges and 15 tasks, has been accepted to NAACL 2024 main conference! |
Mar 05, 2024 | Our position paper on retrieval-augmented LMs is out! |
Mar 01, 2024 | Gave an invited lecture on retrieval-augmented LMs (slides; video) at UW CSE CSE 447/517: Natural Language Processing. I gave the same lecture at Yale and Washington University in St. Louis. |
Feb 23, 2024 | Gave an invited talk about Reliable, Adaptable and Attributable LMs with Retrieval at University College London Web Intelligence Group (WI). |
selected publications
See my full publications at the publication page!