New study on LMs and discourse sensitivity! We evaluate 25 LMs on their ability to prioritize discourse-relevant info, and find that (1) smaller & dialogue-trained models align closer to human patterns, while (2) larger/instruction models overuse structural or discourse cues.
๐จ I am on the faculty job market this year ๐จ
I will be presenting at #NeurIPS2024 and am happy to chat in-person or digitally!
I work on developing AI agents that can collaborate and communicate robustly with us and each other.
My work covers 3 key problems๐
1โฃ Multi-agent + pragmatic interaction and communication
2โฃ Grounding (multimodal inputs, actions)
3โฃ Handling uncertainty, underspecification, and ambiguity
Iโm currently a postdoc with @mohitban47 at @uncnlp and did my Ph.D. with @ben_vandurme at @jhuclsp. If you think I would be a good fit for your department, please reach out:
https://t.co/Ij4QdOVs8D
๐งต๐
Work with @UChicagoLangLab on processing appositive (ARCs) vs. restrictive relative clauses (RRCs) is out: https://t.co/QwvsK3XmOG ARCs typically contain side-commentary info; does a distractor in an ARC lead to an absence of agreement attraction effect? (1/8)
@MH_Christiansen@CSL_Lab@pcontrerask Is there a link to the talk or to a paper/abstract? Thereโs work on agreement errors in English models (partially) overlapping with humans without RLHF or โfeedbackโ. Is the argument they align more closely after fine tuning ?
Would like to disseminate our new paper at #CogSci2024 with Titus von der Malsburg (@tmalsburg) titled "Work Smarter...Not Harder: Efficient Minimization of Dependency Length in SOV Languages". @cogsci_soc
Link: https://t.co/Z1AeAIdGx5
A quick walk-through: ๐งต (1/n)
I'm hoping to find emergency reviewers for the upcoming CogSci conference. If you work in the intersection of NLP/AI and cognitive science and can review a paper by Friday March 15th. Please reach out to me.
I'm at LSA. Always interested in making new friends and catching up with old ones. Happy to chat about the intersection of linguistics and NLP and/or cats. Also, presenting work on adjective acquisition and language models with @mkgotowski on Jan 6! #LSA2024