Freda Shi
@fredahshi
Followers
2K
Following
2K
Media
22
Statuses
415
https://t.co/ZkHLb0G4Cw | Assistant Professor @UWCheritonCS, Faculty Member & Canada CIFAR AI Chair @VectorInst | Past: PhD @TTIC_Connect, BS @PKU1898
Waterloo, ON
Joined December 2016
I'm probably (not too) late in joining the migration, but I'm here:
0
0
2
Personal update: I'll be starting in July 2024 as an Assistant Professor @UWCheritonCS and a Faculty Member @VectorInst! Looking forward to working with all the amazing folks!.Prospective students: if you are interested in NLP and/or comp. linguistics, please consider applying!.
33
19
318
šØLong thread warning: excited to share that I defended my PhD thesis earlier in May!.Here's my thesis, Learning Language Structures through Grounding: 1/
30
11
307
Large language models show reasoning abilities in English with chain-of-thought prompting - how are their multilingual reasoning abilities? . New preprintš: Language models are multilingual chain-of-thought reasoners. (1/n)
5
51
267
Honored to receive the 2021 Google PhD fellowship in natural language processing. Thanks @GoogleAI for the support! Kudos to my advisors and mentors: thanks for teaching me everything over the past years, and for showing me concrete examples of best researchers---yourselves!.
Continuing our tradition of supporting outstanding graduate students in their pursuit of research in computer science and related fields, we congratulate our 13th annual PhD Fellowship Program recipients! See the list of 2021 Fellowship recipients below:
9
4
178
Iām extremely honored to be appointed as a CIFAR AI chair and grateful to everyone who offered generous help along the path. Even more exciting news that makes today super special to me: I just officially received my doctorate hood!.
Freda Shi (@fredahshi @UWaterloo @UWCheritonCS) works on computational linguistics and natural language processing, striving towards a deeper understanding of human language comprehension in order to make more efficient and effective AI systems.
19
10
162
šØ(Not Really) Old Paper AlertšØ: sharing our 2-year-old NeurIPS paper that Iām still quite excited about. We learn grounded, neuro-symbolic CCGs from multi-modal data and demonstrate nearly perfect compositional generalization to unseen sentences and scenes. (1/)
2
17
107
Late post but letās do this! Happy to share our #EMNLP2022 work on translating natural language to executable code with execution-aware minimum Bayes risk decoding.šPaper: šCode: š¦Data (codex output):(1/n)
3
19
103
Just got a paper w/ scores 4, 4, 4 rejected by #acl2020nlp, but the comments from the meta-reviewer and all reviewers are super, super constructive. Would like to say thank you to them all!.
1
1
89
I'm heading to Bangkok for ACL 2024ā Looking forward to meeting old and new friends. I'd also be glad to chat about PhD opportunities in my group at the University of Waterloo, and job search experience in Canada. Feel free to send me an email to secure a slot! 1/.
7
8
68
Iād always be proud of receiving my PhD from TTIC, a magic place which gives you the most unique (in a positive sense, of course!) experience among all PhD programs. Do apply to @TTIC_Connect !.
A reminder that there is NO APPLICATION FEE to apply for TTICās PhD program
0
2
43
Happening right now!
My coauthors and I will present 2 papers at ACL: .1. Structured Tree Alignment for Evaluation of (Speech) Constituency Parsing .Aug 14 10:30 Poster Session 6.Calculating the similarity of two "speech C-parse trees" with continuous time spans as nodes. 2/.
0
2
42
I'm taking a few international interns through the Mitacs Globalink program! .Interested in working with me next summer? Learn more at and apply by Sep 18, 2024!.
1
7
42
Though time it quite limited, I'm happy to spend most of my weekend reviewing for #iclr2023 - my assigned papers are all interesting, carefully written and relevant (to me), as most ICLR papers I've reviewed before - kudos to the ICLR matching system (and my ACs)!.
1
0
39
#ACL2023 attendees: Welcome to Canada! šØš¦. I'll be at the conference from Monday to Wednesday. First time attending a conference without presenting a paper, and Iām sure Iāll enjoy all the cool presentations. Old & new friends: please donāt hesitate to come & say hi!.
1
0
38
I'm very excited about this collaborative effort: it's known that vision-language models perform poorly recognizing spatial relations---in this work, we, for the first time, systematically analyze VLM behaviors from the view of ambiguity from reference frames!.
Do Vision-Language Models represent space, and how? . Spatial terms like "left" or "right" may not be enough to match images with spatial descriptions, as we often overlook the different frames of reference (FoR) used by speakers and listeners. See Figure 1 for examples!
1
5
37
Looking forward to visiting tomorrow!.
š¢Delighted to host @fredahshi's #AI Seminar on "Learning Syntactic Structures from Visually Grounded Text and Speech"!. TOMORROW, OCT. 24 @ 4pm ET:.
1
1
36
Yes, we are looking for PhD students at Waterloo! Come join us ā apply by Dec 1!.
I am hiring NLP/ML PhD students at UWaterloo, home to 5 NLP professors! Apply by Dec 1. Strong consideration will be given to those who can tackle the below challenge: Can we use LM's hidden states to reason multiple problems simultaneously?. āāRetweets/shares appreciatedš„°
1
0
36
Finally, I'll be presenting this work at EMNLP 2022 in person! Cannot wait to meet old and new friends - come and say hi!.
Late post but letās do this! Happy to share our #EMNLP2022 work on translating natural language to executable code with execution-aware minimum Bayes risk decoding.šPaper: šCode: š¦Data (codex output):(1/n)
0
4
32
This has been one of the most exciting posters Iāve visited at EMNLP2022. Neat results showing syntax and semantics are learnably separated in spectrums!.
For #EMNLP2022, we (w/ @robvanderg, @barbara_plank) look through differentiale, rainbow-colored glasses to find linguistic timescale profiles for 7 #NLProc tasks across 6 languages š. š š½ļø š¬ 10th Dec 9:00 at Poster Session 7 & 8
0
0
21
My coauthors and I will present 2 papers at ACL: .1. Structured Tree Alignment for Evaluation of (Speech) Constituency Parsing .Aug 14 10:30 Poster Session 6.Calculating the similarity of two "speech C-parse trees" with continuous time spans as nodes. 2/.
3
2
20
Are there any resource/study showing which words (in any language) are more likely to be mispronounced (by either native speakers or L2 learners)? Any pointer is appreciated!.
2
0
17
Starting this year, I'm also participating as an advisor in the collaborative program between @CIFAR_News and @ELLISforEurope. Consider applying to the ELLIS PhD program if you are interested in joint supervision from me and other ELLIS advisors!.
The #ELLISPhD application portal is now open! Apply to top #AI labs & supervisors in Europe with a single application, and choose from different areas & tracks. The call for applications: Deadline: 15 November 2024. #PhD #PhDProgram #MachineLearning #ML
0
1
18
I very much enjoyed this paper, and of course, the poster! Large-sized data and LLMs present a fantastic opportunity for studying cultural differences.
"ą¤ą¤-ą¤ą¤² NLP Research ą¤ą„ ą¤øą¤¾ą¤„ ą¤¬ą¤Øą„ ą¤°ą¤¹ą¤Øą¤¾ ą¤ą¤¤ą¤Øą¤¾ ą¤¹ą„ ą¤ą¤øą¤¾ą¤Ø ą¤¹ą„ ą¤ą¤æą¤¤ą¤Øą¤¾ ą¤ą¤æ ą¤®ą¤¾ą¤Øą¤øą„ą¤Ø ą¤®ą„ą¤ ą¤ą„ą¤ą¤Øą„ ą¤øą„ ą¤¬ą¤ą„ ą¤°ą¤¹ą¤Øą¤¾!" . Did you understand? How about LMs? Our #ACL2023 Findings paper explores multilingual models' cultural understanding through figurative language in 7 langs š(1/9).
2
0
17
And she feels so lucky to be a student at @TTIC_Connect ;).
Third-year PhD student Freda Shi bridges the gap between linguistics and computer science in her natural language processing research. Follow the link to learn more: #computerscience #womeninstem
1
0
14
If you're at ICML, chat with @xinyun_chen_ on this paper in the poster session 3 11am tomorrow!.
Come by the 11am posters on Wednesday to learn how irrelevant context effects LLMs:.
0
1
13
We are hiring šØš¦.
The Cheriton School of Computer Science @UWCheritonCS at the University of Waterloo is the best computer science program in Canada šØš¦. We are hiring multiple tenure-track faculty positions, with a focus on data systems @dsg_uwaterloo. Deadline November 30.
0
1
15
Surprisingly, PaLM-540B shows decent multilingual reasoning ability, solving >40% problems in any of the 10 investigated languages, including the underrepresented ones (such as Bengali and Swahili) that only cover <0.01% tokens of the pretraining data. (3/n)
1
2
14
Back to 2017, when thinking about visually grounded syntax induction (, I dreamed for 1 second if we could parse image in similar ways---apparently it's too difficult for me then (and now), so, super excited to see this! Congrats on the nice work!.
Introducing #CVPR2022 GroupViT: Semantic Segmentation Emerges from Text Supervision šØāš©āš§. Without any pixel label ever, Our Grouping ViT can group pixels bottom-up to open vocabulary semantic segments. The only training data is 30M noisy image-text pairs.
0
0
14
In the coming year, I'll finish my PhD @TTIC_Connect, and visit @roger_p_levy. Huge thanks to my advisors @kevingimpel and Karen, my mentors @LukeZettlemoyer, @sidawxyz and @denny_zhou , and everyone who helped me along the way!.
1
0
13
Again, welcome to check out our paper and dataset for more details! .Paperš:Dataš¾: (8/n).
1
0
12
Same here. Even worse: I feel I'm probably not qualified to review some of them -- no experience in this domain, not quite familiar with recent work, no labmates or close friends working on it -- while relevant papers (I thought) were not assigned to me.
Got 5 papers to review for ARR today, all from different AEs, the due date is Dec 16! Logged into the system, there's no option to reject the assignment or discuss with AEs to extend the deadline/find a replacement. I wonder what's the average review load for Novš¤@ReviewAcl.
1
1
12
Interested in how VLMs represent spatial relations and why thatās super hard? This is *the* paper to read.
Do Vision-Language Models represent space, and how?. Introducing šļøCOnsistent Multilingual Frame Of Reference Test (COMFORT), an evaluation protocol to assess the spatial reasoning in VLMs under ambiguities. š š Moreš
0
0
13
I look forward to contributing more to the NLP/CL community, and to pushing forward CS based in Waterloo, Ontario, and Canada!.
1
0
11
1. Chain-of-thought prompting is essential for the reasoning performance for both GPT-3 and PaLM; and notably, reasoning steps in English (EN-CoT) almost always outperform the ones in the same language as the problem (Native-CoT). (5/n)
1
1
10
In this work, we introduce the Multilingual Grade School Math (MGSM) dataset, by manually translating 250 English GSM8K test examples to 10 typologically diverse languages, and investigate language modelsā reasoning ability with it. (2/n).
1
0
10
The multilingual reasoning abilities of language models also extend to other tasks: on XCOPA, a multilingual commonsense reasoning dataset, PaLM-540B sets a new state of the art (89.9% average accuracy) using only 4 examples, outperforming the prior best by 13.8%. (7/n).
1
0
9
Prior work has mainly categorized grounding into 2 types: semantic grounding (finding meanings for forms; Harnad, 1990) and communicative grounding (finding common ground in dialogue; Clark and Brennan, 1991 + earlier work in pragmatics). 3/
1
0
10
Joint work with Mirac Suzgun, @markuseful, Xuezhi Wang, Suraj Srivats, @CrashTheMod3, @hwchung27, @yitayml, @seb_ruder, @denny_zhou, @dipanjand, @_jasonwei (9/9).
0
0
8
I'll be talking about Visually Grounded Neural Syntax Acquisition, one of the listed papers, on Monday 4:00 pm at Session 3E! This is a joint work with Jiayuan Mao, @kevingimpel and Karen Livescu. Paper: .Project page:
We are delighted to announce the list of papers that have been nominated as candidates for ACL 2019 Best Paper Awards! Check the list at #acl2019nlp.
0
1
10
2 great surveys centered around the above 2 senses of grounding, respectively:.In the Harnad (1990) sense, by @ybisk, @universeinanegg, @_jessethomason_ and colleagues.In the Clark & Brennan (1991) sense, by folks incl. @ybisk.4/
1
0
10
In my thesis, I discuss a family of tasks: learning language structures from supervision in other sources (through grounding) and corresponding methods to deal with each considered task. As many have recognized, grounding is a highly ambiguous term. More in š§µ.2/.
1
0
10
Thanks to @McAllesterDavid, our anti-grounding prof at @TTIC_Connect: thank you for all the inspiring conversations and writings that push back the idea of grounding, e.g., I hope (and believe) my grounding above is not what you are against :).10/.
1
0
9
Both Heinrich (who didnāt wanna share seat with others) and I enjoyed your excellent defense talk ā huge congrats Dr. Kanishka Misra!.
Oh and my favorite photo from the defense was taken by @fredahshi -- I hope everyone here enjoys it as much as I did (what a great cat!). 5/6
1
0
9
An interesting and counterintuitive example of grounding under this formalization is GroupViT by @Jerry_XU_Jiarui, @xiaolonw, and folks, where an image segmentation model is trained from textual supervision---vision can be grounded in language, too!.8/
1
0
9
@ybisk @universeinanegg @_jessethomason_ In my thesis, I proposed the following definition of grounding, unifying all cases above. Grounding means processing the primary data X with supervision from source Y (the ground), where the mutual information I(X; Y) > 0, so we can find meaningful connections between them. 6/.
1
0
9
@ybisk @universeinanegg @_jessethomason_ One exception is acoustically grounded word embeddings (e.g., Settle et al., 2019), where they encode acoustic knowledge into word embeddings. Perhaps no one thinks the pronunciation of a word is its meaning, but still, this is an acceptable usage of "grounding.".5/
1
0
9
@ybisk @universeinanegg @_jessethomason_ In real-world scenarios, the conditional entropy H(Y|X) almost always> 0, meaning that the ground is usually more complicated than what is to be grounded from certain perspectives. 7/.
1
0
8
Really an interesting paper and well-deserved award! Congratulations!.
Excited and honored to receive a Best Paper Award for our work on the inductive bias of the transformer architecture with @broccolitwit šš #ACL2024NLP
1
0
8
Najoung and her lab always work on problems that attract me a lot. Go and apply to be her student! . p.s. From the logo, I think I'll claim the most important nature of language is recursion.
tinlab at Boston University (with a new logo! šŖ) is recruiting PhD students for F25 and/or a postdoc! Our interests include meaning, generalization, evaluation design, and the nature of computation/representation underlying language and cognition, in both humans and machines. ā¬ļø
1
0
8
2. When example problems in the same language as the problem of interest are available, use them for prompting. If not, use examples from a diverse set of languages. (6/n)
1
0
6
I'm extremely thankful to my advisors Karen and @kevingimpel & my committee members and mentors @lukezettlemoyer and @roger_p_levy, for the great questions and suggestions on my thesis. 12/.
1
0
7
In addition, we analyze the effect of the choices on prompting examples and prompting techniques, and highlight the following takeaways. (4/n).
1
0
6
@denny_zhou @kchonyc I believe both explanations are valid, although marginalizing over reasoning paths that share the same result is probably the most natural way to think about it. My thesis (P123) discusses three explanations of SC and MBR-Exec ().
1
0
6
@McAllesterDavid @TTIC_Connect Also here's a quick guide for readers interested in the additional content covered by my thesis. 11/
1
0
6
To my friends, mentors, coauthors, and everyone who has offered direct or indirect help in the past years, please read my thanks in acknowledgment, which is probably the most exciting part of every PhD thesis. 14/14.
0
0
6
Excited to have the work on tree-based neural sentence modeling (joint with my excellent collaborators Hao Zhou, Jiaze Chen and Lei Li) accepted by #EMNLP2018.
0
0
5
Sida is beyond amazing! Go work with him!.
I'm hiring a PhD intern for the FAIR CodeGen (Code Llama) team. Do research on Code LLMs, execution feedback, evaluation, etc. Apply here:
0
0
5
2. LogogramNLP: Comparing Visual and Textual Representations of Ancient Logographic Writing Systems for NLP (led by @danlu_ai).Aug 12 16:00 Poster Session 3.Visual representation-based system for NLP on ancient logographic languages outperforms conventional Latin transliteration!.
0
1
6
Special thanks to @MichaelHBowling, Dale Schuurmans, and @nidhihegde for the wonderful discussion on grounding at a dinner a year ago. The conversation has made the term grounding (in my mind) more articulable. 9/.
1
0
5
@yuntiandeng We have multilingual GSM here ( and by @davlanade and colleagues! Curious about the performance across languages ;).
1
1
5
I had some difficulty figuring out the horizontal scroll (ęØŖę¹; hĆ©ng pÄ«)āwhile eventually I realize in this case it should be read from left to right, we typically write it from right to left in China :) Happy New Year to my friends who are celebrating!.
The difficulty of expressing "nothing": This is a clever attempt to write aĀ spring couplet (chÅ«nliĆ”n ę„čÆ), not in the usual Sinoglyphs / Chinese characters, but in pictographs: (source) I could figure out about half of the character equivalents (rebusesā¦
0
0
5
@universeinanegg You should probably try .Helped me find a few good cogsci, (non-comp) linguistics and social sci papers. Support on CS papers wasnāt good enough though at the last time I checked it (.5 year ago).
1
0
5
Hearty thanks to organizers @ttk_kuribayashi @g_rambelli @ecekt2, Philipp Wicke, Yohei Oseki and @jixingli for the wonderful event! .2/2.
2
0
5
If not conditioned on their logo, though, I will claim grounding is more important š¤.
1
0
5
Check out Kanishkaās talk this Friday! Highly recommended!.
In case people recovering from colm want more lm content:.
0
0
5
My brain LM is still favo(u)ring āfavoriteā - you should seriously consider coming to Canada š.
Definitely one of my top 3 favourite papers :) It marries deep learning with a minimal set of universal grammar rules for grounded language learning. It draws inspiration from lexicalist linguistics and cognitive science (bootstrapping from core knowledge).
0
0
4
Can't agree more. I voted for "that's syntax", but I wouldn't be happy to see a paper using "syntactic features" to refer to POS tags only, and I've been not so happy for >3 times.
@emilymbender At the very least it's a misleading use of the term. To me it's like doing linear regression and calling it a neural approach. technically true (linear regression can be seen as a 1-neuron neural network) but I don't see why anyone would say it (w/o context) if not to oversell.
0
0
4
Our work got the same result on sentence encoder!.
#EMNLP2018 "A Tree-based Decoder for NMT", a framework for incorporating trees in target side of MT systems. We compare constituency/dependency/non-syntactic binary trees, find surprising result that non-syntactic trees perform best, and try to explain why
1
0
4
Madhur's course is really nice! I'd recommend it to everyone who wishes to review/learn some fundamental mathematical concepts related to machine learning.
@EugeneVinitsky Madhur Tulsiani runs a very similar course every other year (this has links to iterations of the class, the latter ones have more refined notes).
0
0
4
Of course, the work covered in this thesis is built on the foundation of the literatureāmy thanks go to the authors of the papers I cited. I hope I've discussed your work in a fair way. 13/.
1
0
4
And great to see. Yoav brings back the cute llama arts!.
The @COLM_conf reviewing period has started. Reviewers should now receive emails, and all papers are now assigned. Thanks to all our ACs who adjusted assignments in the last few days. Happy reviewing all!
1
0
4
@MorrisAlper @moranynk @ElorHadar @RGiryes Excited to see more work on quantifying visual concreteness! Our ACL'19 work on quantifying text span concreteness and using it for syntactic parsing might also be of interest:
0
0
4
(and guess which is me w/o exact matching on either first or last name! :).
Honored to receive the 2021 Google PhD fellowship in natural language processing. Thanks @GoogleAI for the support! Kudos to my advisors and mentors: thanks for teaching me everything over the past years, and for showing me concrete examples of best researchers---yourselves!.
2
0
4
@maojiayuan @jiajunwu_cs @roger_p_levy As in a CCG, each lexicon entry has its syntactic type and semantic representation. We induce the syntax and semantics of the questions, execute the neuro-symbolic semantic program with visual input, and reward the parser if the execution result is correct. (4/)
1
0
3
1
0
3
@tallinzen Thatās part of the reason why I started using GitHub to manage my working papers. Another part is the nice combination of VSCode & LaTeX workshop.
0
0
3
@UndefBehavior Sorry to hear this! As an alternative, my coauthors and I tried to publish at ML conferences (for our case, NeurIPS) on highly linguistic topics. We got constructive feedback from reviewers, but very little attention for our presentation.
0
0
2
@sharonlevy21 Congrats on the excellent work! I found the Table 1 example very interesting: these four sentences are clearly negative to me, and I can't imagine if anyone would label any of them positive---wonder if more data could fix this?.
2
0
3
Happy days!.
Midwest Speech and Language Days | Day 2. Thank you to all participants, speakers and organizers!
0
0
2
@universeinanegg I occasionally came across this before āā perhaps itās something close to what youāre looking for?.
1
0
2
@kanishkamisra I started to use onenote (not supposed for managing todos though). Just start a new page for all todos this week each Monday morning, and copy leftovers from the prior week.
0
0
2
@joycjhsu This is cool! Can I ask a quick question - why would humans say "no" to the teaser question? From a quick glance, it could perfectly be a "wug" to me :).
3
0
2
Definitely reach out to @WenhuChen @yuntiandeng and/or me if youād like to learn more about Waterloo NLP (and probably catch @hllo_wrld & @lintool next time :).
0
0
2
@SonglinYang4 Thanks Sonta! Now I know who to blame if someone calls me Dr Meow at conferencesš¼.
0
0
2
Great paper and impressive results. Very excited to see it!.
Now with paper link: And code: New results on unsupervised parsing: +6.5 F1 compared to ON-LSTM (2019), +6 F1 compared to PRLG (2011).
0
0
2
We are also aware that the method comes with efficiency issues in complicated real-world settings, and thatās an exciting direction to explore in the future! (13/13).
0
0
2
In summary, we show what will happen when neuro-symbolic models meet grammars (generalized CCGs), where we achieve significantly improved performance on compositional generalization. (12/).
1
0
2