Zhiyu Chen

Applied Scientist at Amazon focusing on IR and NLP problems. My opinions are my own.

[Kawin Ethayarajh @ethayarajh]

📢 Models like #ChatGPT are trained on tons of human feedback. But collecting this costs $$$!

That's why we're releasing the Stanford Human Preferences Dataset (🚢SHP), a collection of 385K *naturally occurring* *collective* human preferences over text.
https://t.co/cRY1F8TjFz

[Gabriel Bénédict @le__gab]

Just started a compilation of generative models for search / information retrieval.

https://t.co/LNiyij82yl

Mostly divided in:
- Grounded Answer Generation (retrieval-augmented, attribution, ...)
- Generative Document Retrieval (DSI, ...)

What is still missing?

[Jundong Li @LiJundong]

We just released a comprehensive survey paper on causal inference in recommender systems https://t.co/t7aX0OA3RN. It covers widely-used strategies for bias mitigation, explanation, and generalization. Please check it out!

[Dan Luo @DanLuo19]

I am thrilled to share that our work paper "Model-based Unbiased Learning to Rank" is accepted as a full paper at #WSDM23 with @lixin26814124 @QingyaoAi @colozoy Dawei Yin and @BrianDavison. Updated arxiv version and code come soon.

[Shuo Zhang @imsure318]

Thrilled that share our work "StruBERT: Structure-aware BERT for Table Search and Matching" w/Mohamed Trabelsi, @colozoy @BrianDavison, and Jeff Heflin has been accepted at @TheWebConf #WWW2022 as a full paper!

[Fangyu Liu @hardy_qr]

Sharing our recent tacl paper... 😎 okay, our recent arxiv preprint called token-aware contrastive learning (TaCL),
by Yixuan Su, me, @mengzaiqiao, Lei Shu, @EhsanShareghi and @nigelhcollier.

[Besnik Fetahu @FetahuBesnik]

Thread: We are organizing a #SemEval22 shared task #multiconer on Multilingual Complex #NER.

The training data is already available for download. Make sure you check it out and participate to the task. https://t.co/qDco5vOrtW #NLProc 1/3

[Huan Sun @hhsun1]

Thanks for tweeting! Yep, please check out our SIGKDD'21 tutorial if you are interested in table understanding. https://t.co/vksInsCdp0

[Google AI @GoogleAI]

Check out a case study with Know Your Data — a dataset exploration tool introduced earlier this year at Google I/O — that highlights how biases can be traced to both dataset collection and annotation practices. https://t.co/CZ9sXYG7NN

[Chenyan Xiong @XiongChenyan]

Just read through this new survey. Very well organized and helped me gain a quick overview of this topic in one hour. Thanks!

[Jimmy Lin @lintool]

For those working on MS MARCO v2 for the TREC 2021 Deep Learning Track, we're happy to share an augmented passage corpus that pulls in metadata from the doc corpus: provides a nearly 20 point bump on recall@100 on the dev queries for first-stage retrieval! https://t.co/2lV913VD2G

[Sergey Levine @svlevine]

Empirical studies observed that generalization in RL is hard. Why? In a new paper, we provide a partial answer: generalization in RL induces partial observability, even for fully observed MDPs! This makes standard RL methods suboptimal.
https://t.co/u5HaeJPQ29

A thread:

[Tech At Bloomberg @TechAtBloomberg]

During #SIGIR2021's Poster Session C this afternoon (12 PM EDT), @LehighCSE's Zhiyu Chen will present "WTR: A Test Collection for Web Table Retrieval," a paper co-written with our #AI Research Scientist @imsure318 and @BrianDavison
https://t.co/Gt1AF5AM1v

[Zhiyu Chen @colozoy]

very good advice😂 “I'll often ask authors of a paper whether it really works”

[Eugene Vinitsky 🦋 @EugeneVinitsky]

Telling which results in RL are worth building on is hard. This blog post discusses three heuristics I use. Curious to learn about what other folks do! https://t.co/fWZ8jtIfxr

[Jiaao Chen @jiaao_chen]

Data augmentation has been one of the most common approaches for mitigating the need for labeled data&improving data efficiency. We provide an empirical*survey of data augm for limited data learning in NLP: https://t.co/y4jTDOSgTg

w/ Derek Tam @colinraffel @mohitban47 @Diyi_Yang

[Claudia Hauff 🇪🇺 🇺🇦 🇩🇪 🇳🇱 @CharlotteHase]

Three ACL 2021 papers read in a row, three times they point to GitHub and three times there is just a placeholder README. A tad frustrating.

[Anjie Fang @anjiefang]

Our team is still looking summer intern candidates. The candidate is expected to have a solid background in NLP. They will be working on NLP related research projects. You can reach out with your CV (send to anjiefang at gmail com). #AmazonScience #InternAtAmazon

[Shuo Zhang @imsure318]

Our resource paper accepted at #SIGIR2021 “WTR: A Test Collection for Web Table Retrieval”, w/@colozoy and @BrianDavison, has extended the table retrieval test collection, WikiTables (w/@krisztianbalog ), with more diverse Web tables and labels on contextual fields.