Cui Ding

Cui Ding

@cuiding.bsky.social

44 followers82 following19 posts

Cui Ding
@cuiding.bsky.social

昨天是emotional的一天。中午Reinhold问我，什么时候会完成dessertation，以后有什么计划。我忙着说对不起，说我总是想证明自己，证明自己其实比profile中展示出来的更好，因为博士前两年…… Reinhold说：don't worry. People have already formed opinions. 一直以来都觉得要做到“人不知而不愠”好难好难。那刻忽然像有了答案，很释怀了。

Cui Ding reposted

Jan Chromý
@janchromy.bsky.social

New paper out in Cognitive Science! We analyzed various predictors of local ambiguity resolution on a heterogeneous set of garden-path sentences, namely surprisal, frequency, plausibility and cloze scores (measuring the likelihood of misanalysis of the ambiguous region). 1/2

onlinelibrary.wiley.comCloze, Frequency, Surprisal, or Plausibility? A Comparative Analysis of Predictors for Local Ambiguity ResolutionThis study investigated the cognitive mechanisms underlying the processing of garden-path sentences by examining the influence of verb/structural bias, cloze probability, surprisal, and plausibility....

Cui Ding reposted

Rico Sennrich
@ricosennrich.bsky.social

My lab is recruiting one PhD student and one post-doctoral researcher for a start as soon as this Summer! Apply by April 1 / March 31 to be among first candidates considered. jobs.uzh.ch/job-vacancie... jobs.uzh.ch/job-vacancie...

jobs.uzh.chUZH: PhD in Language AI / Natural Language ProcessingYou will be joining the Department of Computational Linguistics, which has 6 Research Groups and around 70 postdoctoral and student researchers in the areas of Text Technologies, Phonetics and Speech ...
Cui Ding
@cuiding.bsky.social

So happy to have presented our project on **Individual signatures of spillover** at CPL2025. Grateful to my supervisor Lena Jaeger, for her great ideas and for her kind support! Thanks to my coauthors and the audience at CPL2025. The award was a democratic vote by the audience!

Computational Linguistics @ UZH@cl-uzh.bsky.social

Congratulations to our PhD student @cuiding.bsky.social for winning the Best Talk Award in the Computational Psycholinguistics Meeting 2025 (CPL 2025) lnkd.in/euTX89gi!

Cui Ding reposted

Tamar Regev
@tamaregev.bsky.social

New preprint on prosody in the brain! tinyurl.com/2ndswjwu HeeSoKim NiharikaJhingan SaraSwords @hopekean.bsky.social @coltoncasto.bsky.social JenniferCole @evfedorenko.bsky.social Prosody areas are distinct from pitch, speech, and multiple-demand areas, and partly overlap with lang+social areas→🧵

tinyurl.comA distinct set of brain areas process prosody--the melody of speechHuman speech carries information beyond the words themselves: pitch, loudness, duration, and pauses--jointly referred to as 'prosody'--emphasize critical words, help group words into phrases, and conv...

Cui Ding reposted

Yevgeni Berzak
@whylikethis.bsky.social

#NeurIPS2025 Check out EyeBench 👀, a mega-project which provides a much needed infrastructure for loading & preprocessing eye-tracking for reading datasets, and addressing super exciting modeling challenges: decoding linguistic knowledge 👩 and reading interactions 👩+📖 from gaze! eyebench.github.io

Cui Ding
@cuiding.bsky.social

Wow! Thank you so much!!

BogaertsLab@bogaertslab.bsky.social

Very much enjoying #IndiREAD workshop on #IndividualDifferences in #Reading! While individual differences have brought my lab both excitement and headaches in recent years, I liked the optimistic take of @cuiding.bsky.social.

View full thread
Cui Ding
@cuiding.bsky.social

Key Result 2: Information from Models Using fine-tuned Qwen2.5-VL and TransOCR, we estimated the MI between images and word identity. MI systematically drops: Full > Upper > Lower — perfectly mirroring human reading patterns! 🤯

Cui Ding
@cuiding.bsky.social

Take-home Message 🔹 We formalize input quality in reading as mutual information. 🔹 We link it to measurable human behavior. 🔹 We show multimodal LLMs can model this effect quantitatively. Bottom-up information matters — and now we can measure how much it matters.

View full thread
Cui Ding
@cuiding.bsky.social

We model reading time as proportional to the number of visual “samples” needed to reduce uncertainty below a threshold ϕ. Higher mutual information → fewer samples → faster reading.

Cui Ding
@cuiding.bsky.social

Key Result 1: Human Reading Reading times show a clear pattern: Full visible< Upper visible < Lower visible in both English & Chinese. 👉 Upper halves are more informative (and easier to read).

Cui Ding
@cuiding.bsky.social

We propose a formal model where reading is a Bayesian update integrating top-down expectations and bottom-up evidence. When bottom-up input is noisy (e.g., words are partially occluded), comprehension becomes harder and slower.

Cui Ding
@cuiding.bsky.social

📊 We quantify this using mutual information (MI) between visual input and word identity. To test the theory, we created a reading experiment using the MoTR (Mouse-Tracking-for-Reading) paradigm 🖱️📖 We ran the study in both English and Chinese.

Cui Ding reposted

Hanxu Hu
@hanxuhu.bsky.social

💥Introducing new paper: arxiv.org/pdf/2510.17715, QueST — train specialized generators to create challenging coding problems. From Qwen3-8B-Base ✅ 100K synthetic problems: better than Qwen3-8B ✅ Combining with human written problems: matches DeepSeek-R1-671B 🧵(1/5)

Cui Ding reposted

Henrik Singmann
@singmann.bsky.social

Exciting #rstats news for Bayesian model comparison: bridgesampling is finally ready to support cmdstanr, see screenshot. Help us by installing the development version of bridgesampling and letting us know if it works for your model(s): pak::pkg_install("quentingronau/bridgesampling#44")

Cui Ding
@cuiding.bsky.social

Some of my colleagues are already very excited about this work!

Tiago Pimentel@tpimentel.bsky.social

If you use LLMs, tokenisation bias probably affects you: * Text generation: tokenisation bias ⇒ length bias 🤯 * Psycholinguistics: tokenisation bias ⇒ systematically biased surprisal estimates 🫠 * Interpretability: tokenisation bias ⇒ biased logits 🤔

Cui Ding
@cuiding.bsky.social

I am so proud of this work. My first NLP experience. I learned a lot from this amazing team!!!!

Ethan Gotlieb Wilcox@wegotlieb.bsky.social

⭐🗣️New preprint out: 🗣️⭐ “Using Information Theory to Characterize Prosodic Typology: The Case of Tone, Pitch-Accent and Stress-Accent” with @cuiding.bsky.social , Giovanni Acampa, @tpimentel.bsky.social , @alexwarstadt.bsky.social ,Tamar Regev: arxiv.org/abs/2505.07659

arxiv.orgUsing Information Theory to Characterize Prosodic Typology: The Case of Tone, Pitch-Accent and Stress-AccentThis paper argues that the relationship between lexical identity and prosody -- one well-studied parameter of linguistic variation -- can be characterized using information theory. We predict that lan...

Cui Ding reposted

Ethan Gotlieb Wilcox
@wegotlieb.bsky.social

⭐🗣️New preprint out: 🗣️⭐ “Using Information Theory to Characterize Prosodic Typology: The Case of Tone, Pitch-Accent and Stress-Accent” with @cuiding.bsky.social , Giovanni Acampa, @tpimentel.bsky.social , @alexwarstadt.bsky.social ,Tamar Regev: arxiv.org/abs/2505.07659

arxiv.orgUsing Information Theory to Characterize Prosodic Typology: The Case of Tone, Pitch-Accent and Stress-AccentThis paper argues that the relationship between lexical identity and prosody -- one well-studied parameter of linguistic variation -- can be characterized using information theory. We predict that lan...
View full thread
Cui Ding
@cuiding.bsky.social

Participants must move their mouse over the text to reveal the words, while their cursor movements are recorded (similar to how eye movements are recorded in eye tracking). See below for an example MoTR trial.

Cui Ding
@cuiding.bsky.social

The biggest advantage of MoTR over alternative methods is that it is very cheap and fast compared to its alternatives, while still provides very sensitive and accurate measurements. Our online data collection from 60 Russian speakers took less than 24 hours!!

Cui Ding
@cuiding.bsky.social

2- We use MoTR (Mouse Tracking for Reading) as a cheaper but reliable alternative to in-person eye tracking. MoTR is a new experimental tool, where participants screen is blurred except for a small region around the tip of the mouse pointer

You are up to date.