about ▽
Hello! I am a researcher in NLP/AI, currently working on synthetic data and retrieval, although my interests are still extremely broad in the LLM space. Currently, I work at Microsoft. Previously, I completed my PhD in 2022 at JHU's Center for Language and Speech Processing working with Benjamin Van Durme.
I also solve and write puzzles and participate in puzzlehunts. These days I tend to write for Teammate Hunt and Galactic Puzzle Hunt. Some other hobbies include cycling, rock climbing, and board/video games.
contact: patrickxia at microsoft dot com | discord: @patrickx
misc: pitrack | @nlpaxia
Last updated: Aug 5 2024
news ▽
Aug Our paper on Task Decomposition will be presented at IJCAI.
Mar I will be co-hosting Millicent Li's internship this summer, with Tongfei Chen.
papers ▽
Interpreting User Requests in the Context of Natural Language Standing Instructions
Nikita Moghe, Patrick Xia, Jacob Andreas, Jason Eisner, Benjamin Van Durme, Harsh Jhamtani NAACL - Findings (2024)= arxiv =
SCREWS: A Modular Framework for Reasoning with Revisions
Kumar Shridhar, Harsh Jhamtani, Hao Fang, Benjamin Van Durme, Jason Eisner, Patrick Xia= arxiv code =
Few-Shot Adaptation for Parsing Contextual Utterances with LLMs
Kevin Lin, Patrick Xia, Hao FangIJCNLP - Findings (2023)
= arxiv code =
Natural Language Decomposition and Interpretation of Complex Utterances
Harsh Jhamtani, Hao Fang, Patrick Xia, Eran Levy, Jacob Andreas, Ben Van DurmeIJCAI (2024)
= arxiv =
Multilingual Coreference Resolution in Multiparty Dialogue
Boyuan Zheng, Patrick Xia, Mahsa Yarmohammadi, Benjamin Van DurmeIn TACL (2023)
= arXiv =
Automatic Document Selection for Efficient Encoder Pretraining
Yukun Feng, Patrick Xia, Benjamin Van Durme, João SedocIn EMNLP (2022)
= pdf =
Adapting Coreference Resolution Models through Active Learning
Michelle Yuan, Patrick Xia, Benjamin Van Durme, Jordan Boyd-GraberACL (2022)
= pdf =
Pruning Pretrained Encoders with a Multitask Objective
Patrick Xia, Richard ShinIn ENLSP at NeurIPS (2021)
= pdf =
On Generalization in Coreference Resolution
Shubham Toshniwal*, Patrick Xia*, Sam Wiseman, Kevin Gimpel, Karen LivescuIn CRAC (2021) [best short paper]
= pdf code =
Moving on from OntoNotes: Coreference Resolution Model Transfer
Patrick Xia, Benjamin Van DurmeIn EMNLP (2021)
= pdf code slides (pdf, ppt) poster =
Which *BERT? A Survey Organizing Contextualized Encoders
Patrick Xia, Shijie Wu, Benjamin Van Durme In EMNLP (2020)= pdf slides (pdf (long), pdf (short), ppt) video =
Incremental Neural Coreference Resolution in Constant Memory
Patrick Xia, João Sedoc, Benjamin Van Durme In EMNLP (2020)= pdf slides (pdf, ppt) code video =
Multi-Sentence Argument Linking
Seth Ebner*, Patrick Xia*, Ryan Culkin, Kyle Rawlins, Benjamin Van Durme In ACL (2020)= pdf data code slides (ppt, pdf) =
Can You Tell Me How to Get Past Sesame Street? Sentence-Level Pretraining Beyond Language Modeling
Alex Wang, Jan Hula, Patrick Xia, Raghavendra Pappagari, R. Thomas McCoy, Roma Patel, Najoung Kim, Ian Tenney, Yinghui Huang, Katherin Yu, Shuning Jin, Berlin Chen, Benjamin Van Durme, Edouard Grave, Ellie Pavlick, Sam Bowman In ACL (2019)= (outdated) OpenReview (outdated) arXiv pdf code =
Improved Lexically Constrained Decoding for Translation and Monolingual Rewriting
J. Edward Hu, Huda Khayrallah, Ryan Culkin, Patrick Xia, Tongfei Chen, Matt Post and Benjamin Van DurmeIn NAACL (2019)
= pdf demo website paraphrased MNLI =
What do you learn from context? Probing for sentence structure in contextualized word representations
Ian Tenney, Patrick Xia, Berlin Chen, Alex Wang, Adam Poliak, R Thomas McCoy, Najoung Kim, Benjamin Van Durme, Sam Bowman, Dipanjan Das, Ellie PavlickIn ICLR (2019)
= OpenReview pdf related slides poster code =
Annotating Character Relationships in Literary Texts
Philip Massey, Patrick Xia, David Bamman, Noah A. SmitharXiv (2015)
= pdf =
puzzles ▽
I've been writing puzzles since ~2013, from PuzzlehuntCMU, Galactic Puzzle Hunt, Teammate Hunt, and more recently I was an editor-in-chief of the 2023 MIT Mystery Hunt. This list contains a few select puzzles along with my interpretation of their difficulty. Some puzzles are "paired," meaning they are meant to be solved simultaneously -- the pairing was not given during the event, which made these puzzles even harder.
I was also part of creating Conjuri's Quest, a puzzle roguelite that was part of the 2023 MIT Mystery Hunt event. The game contains a full round of (1-4 ★) puzzles and the game itself is a (meta)puzzle. However, it is also playable and winnable as a game. I also occasionally write puzzles that are not for a specific event, like Puzzlebox.
Puzzle |
Difficulty |
Puzzle Type |
Solution |
Beige Flags | ★☆☆☆☆☆ | Words | Solution |
Hyperspace | ★★☆☆☆☆ | Words | Solution |
We Made a Quiz Bowl Packet but Somewhere Things Went Horribly Wrong | ★★★★★★ | Research | Solution |
My Fun Bow | ★★★★★★ | Swahili | Solution |
Book Spade | ★★☆☆☆☆ | Words | Solution |
Pin The Tail / Meta-Eval Times | ★★★★★☆ | Interactive game / Blackbox | Solution & Solution |
DC Meet and Greet / Ring Toss | ★★★★★★ | Words, Reference | Solution & Solution |
Solving Pumpkin Patch with NO Given Pumpkins?! (paired with Menagerial Problems) | ★★★★☆☆ | Logic / Math | Solution (& Solution) |
Symbols | ★★☆☆☆☆ | Words | Solution |
Thrifty/Thrifty | ★★★★☆☆ | Black box | Solution |
Drip Quote | ★★★★☆☆ | Words | Solution |
Connect the Dots | ★★★★★★ | Research | Solution |
This Anagram Does Not Exist | ★★★★★☆ | Words | Solution |
Exceptions to the Rule | ★☆☆☆☆☆ | Words | Solution |
50/50 | ★★★★☆☆ | Words | Solution |
Now Get Ready For... | ★★☆☆☆☆ | Words | Solution |
Fan Video | ★★★☆☆☆ | Reference | Solution |
Chromesthesia | ★★☆☆☆☆ | Research | Solution |