Byung-Doh Oh

@byungdoh

byungdoh.github.io

Joined June 2009

136 Following

128 Followers

Replies Reposts Media

Byung-Doh Oh
@byungdoh
Dec 16, 2025
Sorry for the spam, we fixed the URL! The workshop is still on May 16, 2026 in Mallorca
CMCL workshop
@CMCL_workshop
Dec 16, 2025
The CMCL Call for Papers is out! 🌴We are looking forward to receiving many interesting submissions!✨ (Deadline: February 25, 2026) sites.google.com/view/cmclworks… @byungdoh @ttk_kuribayashi @g_rambelli Philipp Wicke, @jixingli , @ecekt2 @yoshiryo0617
Byung-Doh Oh
@byungdoh
Dec 16, 2025
CMCL will be on May 16, 2026 in Mallorca (alongside LREC)! The webpage is actually sites.google.com/view/cmclworks…!
This post is unavailable.
Byung-Doh Oh
@byungdoh
Nov 14, 2025
LM surprisal was thought to just grossly underpredict human garden-path effects, but eye-tracking data tells a different story! Read (and reread) about it here! osf.io/eq2ra
Will Timkey
@wtimkey8
Nov 14, 2025
New Preprint: osf.io/eq2ra Reading feels effortless, but it's actually quite complex under the hood. Most words are easy to process, but some words make us reread or linger. It turns out that LLMs can tell us about why, but only in certain cases... (1/n)
GIF
Byung-Doh Oh
@byungdoh
Jun 9, 2025
Is the slowdown in reading at low-frequency words separable from predictability effects? Studies using LM surprisal like doi.org/10.1162/opmi_a… suggest so. Our new JML paper doi.org/10.1016/j.jml.… instead argues it may depend on the LM, which is itself sensitive to frequency.
1
Byung-Doh Oh
@byungdoh
Jun 5, 2025
Subword token granularity influences 1) word probabilities at initialization and 2) token co-occurrences during training. How does this impact LM surprisal? We tackled this using Mamba-2 models, which can be flexibly trained on sequences of different lengths. #ACL2025 #ACL2025NLP
1