Interested in fact-checking? Just announced: Next FEVERworkshop at EMNLP shared task: fever.ai/task.html aiming to evaluate systems verifying real-world claims with evidence from the Web, based on the AVERITEC dataset: arxiv.org/abs/2305.13117 With the amazing team:

Happy to share my meta internship work, to be presented at FAccT ‘24

Diversity exists, even w/in classes!! Ignoring it leads to biased performance 😨 Embracing it can improve performance AND transparency, for free 😌

arxiv.org/abs/2404.16717

Today we're launching PRISM, a new resource to diversify the voices contributing to alignment. We asked 1500 people around the world for their stated preferences over LLM behaviours, then we observed their contextual preferences in 8000 convos with 21 LLMs arxiv.org/abs/2404.16019

🗣️🌏Submit your work on spoken language translation to IWSLT by April 29!
In addition to shared task papers, you can submit regular scientific papers. And we have a special highlights track to reflect on emerging trends, or share work published elsewhere with the IWSLT community

Happy to share our recent EACL 2024 paper #TENOR . Unlike LDA, BERTopic, Contextualized topic model... that generate a list of abstract keywords for each topic: TENOR leverages topic models with user-centered topic generation and document analysis with minimal user effort.

I've finally uploaded the thesis on arXiv: arxiv.org/abs/2404.12150

It ties together a bunch of papers exploring some alternatives to RL for finetuning LMs, including pretraining with human preferences and minimizing KL divergences from pre-defined target distributions.

Dear #NAACL Members, to better explain some of the arguments for a possible name change, several members of our community who reside or originate from the Americas outside of US/Canada have written an open letter 👉 naacl.org/posts/2024-04-…

Original survey forms.gle/r8SWiu8goG79kw…

Many AI safety methods rely on AIs to provide oversight and supervision. The fact that models are interacting with one another alone should be a safety concern. We provide some initial evidence to support this intuition. More to follow soon!

I'm one of the co-organisers of this workshop this year at ACL -- submit your papers on gender bias! A lot of really cool work comes out of this, and I want to see yours!

Super cool that even smaller LLMs can self-improve in MT, great work!!

Kendrick is gonna find something absolutely disgusting to rhyme with Millie Bobby Brown

Is ChatGPT Transforming Academics’ Writing Style?

People are beginning to use the word 'significantly' significantly more frequently, thanks to ChatGPT

arxiv.org/abs/2404.08627

Clever SRS algorithm optimizer paper uses BERT-based similarity metric on item text to better approximate forgetting curves for new items (via Jarrett Ye)

Paper URL: arxiv.org/abs/2402.12291

Via this very interesting SRS benchmarking post: reddit.com/r/Anki/comment…

“What would an AI research & education mission that crossed the whole campus look like?”

This is the question posed to me & a working group by
Univ. of Maryland's Provost Rice summer'23.

Now we’re announcing the answer: #AIMaryland , the AI Interdisciplinary Institute at Maryland.
>

.Univ. of Maryland launches an institute focused on ethical AI development. It supports interdisciplinary research and prepares the future workforce for an AI-infused world. CS Professor Hal Daumé III is its inaugural director.

Read more:
go.umd.edu/AIM

✨ Can we detect style in the generated images? Our recent work takes a step towards understanding this question. We train a style-focused vision feature extractor built on top of CLIP, which we call a Contrastive Style Descriptior (CSD).

paper: arxiv.org/abs/2404.01292

Style

📣 Attention all participants of trustnlp!
We're excited to announce that the deadline for our workshop has been extended!
🗓️ New Deadline: April 2nd (Tue), AOE
This is your chance to fine-tune your projects and submissions. Let's make the most of this extra time! 🌟

[1/7] Pre-trained LMs can do in-context learning, but this is unexpected given the distribution shift between pre-training data and ICL prompts. What structures of pre-training data yield ICL? Check out our work “Parallel Structures in Pre-training Data Yield In-Context Learning”

We sadly found out our CTM paper (ICLR24) was plagiarized by TCD! It's unbelievable😢—they not only stole our idea of trajectory consistency but also comitted 'verbatim plagiarism,' literally copying our proofs word for word! Please help me spread this.

📢 I'm recruiting a postdoc at UMD
🗓️ Starting in 2024
🖥️🙋 Topics related to human-AI interaction
🔜 Apply by May 31, but earlier if possible!
🤑 $75-$80k+benefits, unencumbered (ie not 'big project') funding

Link to full post and form for applications in thread.

>