ConvAI had a great NeurIPS season with four accepted papers to the main conference🎉 Find all the authors in San Diego this December ☀️
20.09.2025 15:17
👍 3
🔁 1
💬 1
📌 0
ConvAI had a great NeurIPS season with four accepted papers to the main conference🎉 Find all the authors in San Diego this December ☀️
Thrilled to announce our new survey that explores the exciting possibilities and troubling risks of computational persuasion in the era of LLMs 🤖💬
📄Arxiv: arxiv.org/pdf/2505.07775
💻 GitHub: github.com/beyzabozdag/...
We won a Senior Area Chair Award at NAACL!! Many thanks again to my amazing coauthors Gaurav Kamath and @sivareddyg.bsky.social :-)
Congratulations to Mila members @adadtur.bsky.social , Gaurav Kamath and @sivareddyg.bsky.social for their SAC award at NAACL! Check out Ada's talk in Session I: Oral/Poster 6. Paper: arxiv.org/abs/2502.05670
Super excited that this is finally out! We evaluated leading LLM-based web agents from OpenAI, Anthropic, and more, on our new benchmark SafeArena and found that many are surprisingly compliant with malicious requests. Check out the leaderboard here: huggingface.co/spaces/McGil...
This work was done by an awesome team of authors: @adadtur.bsky.social, Nick, @arkil.bsky.social, @karstanczak.bsky.social, Esin, @spandanagella.bsky.social, and @sivareddyg.bsky.social.
It's also important to recognize the incredible works that helped us build SafeArena:
Agents like OpenAI Operator can solve complex computer tasks, but what happens when users use them to cause harm, e.g. spread misinformation?
To find out, we introduce SafeArena (safearena.github.io), a benchmark to assess the capabilities of web agents to complete harmful web tasks. A thread 👇
While persuasive models are promising for social good, they can also be misused towards harmful behavior. Recent work by @beyzabozdag.bsky.social and @shuhaib.bsky.social aims to assess LLM persuasiveness and susceptibility towards persuasion.
[1/6] Can LLMs out-persuade each other? 🤖🧠💬
Introducing Persuade Me If You Can (PMIYC)—a new framework to evaluate (1) how persuasive LLMs are and (2) how easily they can be persuaded! 🚀
📄Arxiv: arxiv.org/abs/2503.01829
🌐Project Page: beyzabozdag.github.io/PMIYC/
Overview figure for paper, showing creation of constituent movement data, in addition to three step experimentation: "Model Shifting Preference", "Motivating Factors of Model Preference", "Human-Model Preference Correlation"
Super excited to finally announce our NAACL 2025 main conference paper “Language Models Largely Exhibit Human-like Constituent Ordering Preferences”!
We examine constituent ordering preferences between humans and LLMs; we present two main findings… 🧵
🚀Very excited about my new paper!
NN-CIFT slashes data valuation costs by 99% using tiny neural nets (205k params, just 0.0027% of 8B LLMs) while maintaining top-tier performance!
The secret sauce for this work is the ReAct style training data preparation: “User-Thought1-Action/API-Observation-Thought2-Response”. We transformed public dialogue datasets into this format for training. Congratulations to @emrecanacikgoz and the @convai_uiuc and Oumi teams!
Instruction data can also be synthesized using feedback based on reference examples. Please check our recent work for more information. Thanks to @shuhaib.bsky.social, Xiusi Chen, and Heng Ji!
💡 Introducing Reference-Level Feedback: A new paradigm for using feedback to improve synthetic data!
🌐 shuhaibm.github.io/refed/
🧵 [1/n]
AI over-reliance is an important issue for conversational agents. Our work supported mainly by the DARPA FACT program proposes introducing positive friction to encourage users to think critically when making decisions. Great team-work, all!
@convai-uiuc.bsky.social @gokhantur.bsky.social
‼️ Ever wish LLMs would just... slow down for a second?
In our latest work, "Better Slow than Sorry: Introducing Positive Friction for Reliable Dialogue Systems", we delve into how strategic delays can enhance dialogue systems.
Paper Website: merterm.github.io/positive-fri...
Seeing this year's ACL Fellows is like walking through the hallways of Microsoft Research, Building 99 in 2016. Congratulations @dilekh.bsky.social Scott Yih, Jianfeng Gao, and Lucy Vanderwende 👏
Nice overview of the ReSpAct framework for conversational task completion agents @convai-uiuc.bsky.social
cobusgreyling.medium.com/building-con...
@chrupala.me please add me in to the SLP pack
