[#ACL2026 Paper Alert] Real-world requests are often underspecified. So when should an AI agent ask a clarification question? Not always. Not never. We introduce Value of Information (VoI): a decision-theoretic framework for deciding when to ask, when to act, and when to stop.

SimBench now at #ICLR2026! Often in social simulations, the goal is not to predict what one specific person will do. It is to estimate how a group will respond, whether in pre-testing a real polling question, or in stress-testing a policy or intervention before running it in the real world.

1/7 🧵 The GPT-4 technical report featured detailed calibration curves. Since then, not a single major model release has reported calibration. The field quietly stopped measuring whether models know what they don't know. Our new position paper argues this is a mistake. Here's why.

Proud to contribute to the new International AI Safety Report chaired by @YoshuaBengio, with a fantastic international team! Every word was weighed to ensure a rigorous, evidence-based view of current AI capabilities and the risks they pose. A short summary of my section below.

Yoshua Bengio@yoshuabengio.bsky.social · 6mo ago

Today we’re releasing the International AI Safety Report 2026: the most comprehensive evidence-based assessment of AI capabilities, emerging risks, and safety measures to date. 🧵 (1/19)

I’m pleased to share the Second Key Update to the International AI Safety Report, which outlines how AI developers, researchers, and policymakers are approaching technical risk management for general-purpose AI systems. (1/6)

Personalization certainly needs boundaries and we show how that could look like!

Debora Nozza@deboranozza.bsky.social · 9mo ago

🚨 New main paper out at #EMNLP2025! 🚨 ⚡ We show that personalization of content moderation models can be harmful and perpetuate hate speech, defeating the purpose of the system and hurting the community. We argue that personalized moderation needs boundaries, and we show how to build them.

Instruction tuning unlocks incredible skills in LLMs, but at a cost: they become dangerously overconfident. You face a choice: a well-calibrated base model or a capable but unreliable instruct model. What if you didn't have to choose? What if you could navigate the trade-off? (1/8)

River, Yinhong and I will all be in person and we look forward to the discussions!

Cohere Labs@cohereforai.bsky.social · 9mo ago

Cohere Labs x EMNLP 2025 "When Personalization Meets Reality: A Multi-Faceted Analysis of Personalized Preference Learning" Congrats to authors Yijiang River Dong, @tiancheng.bsky.social, Yinhong Liu, Ahmet Üstün, Nigel Collier. 📜 arxiv.org/abs/2502.19158

See you next week at EMNLP! We will be presenting our work: Scaling Low-Resource MT via Synthetic Data Generation with LLMs 📍 Poster Session 13 📅 Fri, Nov 7, 10:30-12:00 - Hall C 📖 Check it out! arxiv.org/abs/2505.14423 @helsinki-nlp.bsky.social @cambridgenlp.bsky.social @emnlpmeeting.bsky.social

Scaling Low-Resource MT via Synthetic Data Generation with LLMs

We investigate the potential of LLM-generated synthetic data for improving low-resource Machine Translation (MT). Focusing on seven diverse target languages, we construct a document-level synthetic co...

arxiv.org

Can AI simulate human behavior? 🧠 The promise is revolutionary for science & policy. But there’s a huge "IF": Do these simulations actually reflect reality? To find out, we introduce SimBench: The first large-scale benchmark for group-level social simulation. (1/9)

Excited to share a "Key Update" from the International AI Safety Report, which I was proud to contribute to. We took a rigorous, evidence-based look at the latest AI developments. If you want a clear view of where things stand, this is a must-read. 👇

Yoshua Bengio@yoshuabengio.bsky.social · 10mo ago

AI is evolving too quickly for an annual report to suffice. To help policymakers keep pace, we're introducing the first Key Update to the International AI Safety Report. 🧵⬇️ (1/10)

Heading to Vienna today to attend #ACL2025NLP! Let's chat if you are interested in LLM social simulation, personalization, character training and human-centered AI!

I will be at #acl2025 to present "Beyond Demographics: Fine-tuning Large Language Models to Predict Individuals’ Subjective Text Perceptions" ✨ Huge thank you to my collaborators Jiaxin Pei @paul-rottger.bsky.social Philipp Cimiano @davidjurgens.bsky.social @dirkhovy.bsky.social 🍰 more below

Working on LLM social simulation and need data? Excited to announce our iNews paper is accepted to #ACL2025! 🥳 It's a large-scale dataset for predicting individualized affective responses to real-world, multimodal news. Paper: arxiv.org/abs/2503.03335 Data: huggingface.co/datasets/pit...

Tiancheng Hu@tiancheng.bsky.social · last yr.

Ever notice how something that makes your blood boil barely registers with your friend? Our emotional reactions aren't universal at all—they're deeply personal. And AI needs to understand that. Excited to share our new paper: "iNews" 🧵 (1/8) arxiv.org/abs/2503.03335

We created Approximate Likelihood Matching, a principled (and very effective) method for *cross-tokenizer distillation*! With ALM, you can create ensembles of models from different families, convert existing subword-level models to byte-level and a bunch more🧵

Ever notice how something that makes your blood boil barely registers with your friend? Our emotional reactions aren't universal at all—they're deeply personal. And AI needs to understand that. Excited to share our new paper: "iNews" 🧵 (1/8) arxiv.org/abs/2503.03335

iNews: A Multimodal Dataset for Modeling Personalized Affective Responses to News

Current approaches to emotion detection often overlook the inherent subjectivity of affective experiences, instead relying on aggregated labels that mask individual variations in emotional responses. ...

arxiv.org

Great work by @riverdong.bsky.social - we dug deep into existing datasets & algorithms and found quite some surprising stuff

@riverdong.bsky.social · last yr.

🚨New Paper Alert🚨 Many personalization methods optimize performance but ignore real-world impact. We examine its effects on: ✅ Performance ⚖️ Fairness: Can it represent minorities fairly? ⚠️ Unintended Effects: Does it harm safety? 🔄 Adaptability: Quickly adapt to new users?

Happy to write this News & Views piece on the recent audit showing LLMs picking up "us versus them" biases: www.nature.com/articles/s43... (Read-only version: rdcu.be/d5ovo) Check out the amazing (original) paper here: www.nature.com/articles/s43...

Large language models act as if they are part of a group - Nature Computational Science

An extensive audit of large language models reveals that numerous models mirror the ‘us versus them’ thinking seen in human behavior. These social prejudices are likely captured from the biased conten...

nature.com

1/9 🧵 New paper alert (now in Nature Computational Science)! As polarisation continues to shape our world, we asked: Do social and political biases transfer to our AI? I.e. do LLMs show ingroup and outgroup bias? www.nature.com/articles/s43...

Generative language models exhibit social identity biases - Nature Computational Science

Researchers show that large language models exhibit social identity biases similar to humans, having favoritism toward ingroups and hostility toward outgroups. These biases persist across models, trai...

nature.com

Such an amazing work in so many ways! Well done! Great to see convergent evidence that the more persona information you have about someone, the more accurate your simulation would be. aclanthology.org/2024.acl-lon... Is there a scaling law for simulation based on persona detailedness?

Joon Sung Park@joon-s-pk.bsky.social · 2y ago

Simulating human behavior with AI agents promises a testbed for policy and the social sciences. We interviewed 1,000 people for two hours each to create generative agents of them. These agents replicate their source individuals’ attitudes and behaviors. 🧵 arxiv.org/abs/2411.10109

🚨New Preprint: "Generative language models exhibit social identity biases" Did you know LLMs mirror human-like biases, showing human-levels of ingroup solidarity & outgroup hostility? @profsanderlinden.bsky.social @steverathje.bsky.social 📄 arxiv.org/abs/2310.15819