RAG enables LLMs to access external info 📖. But when this info is multiple languages 🌐, can LLMs reconcile differing viewpoints 🧐? We introduce BordIRlines, a dataset to study the robustness of cross-lingual RAG. 📃arxiv.org/abs/2410.01171 🗃️ huggingface.co/datasets/borde… 1/4 🧵

📣 New Paper 📣: AI-generated fake news is flooding social media, making detection crucial. Introducing MiRAGeNews—a large dataset of image+caption pairs for detecting fake news. Our MiRAGe model outperforms both humans and SOTA VLMs across various generators.

✨Updates✨: • Dolomites was accepted to TACL: dolomites-benchmark.github.io! - Our data is now also up on HuggingFace: huggingface.co/datasets/cmala…. • I will be talking about Dolomites at EMNLP'24 in Miami (Session 11 on Nov 14 at 10:45 ET). Please say hi if you're around!

🤔What model explanation method should you use? How to ensure it reflects the model’s true reasoning? 🌟 In our CL survey, Towards Faithful Model Explanation in NLP, we review 110+ explainability methods through the lens of faithfulness. Check out my presentation at #EMNLP2024!

Excited to share ✨ Contextualized Evaluations ✨! Benchmarks like Chatbot Arena contain underspecified queries, which can lead to arbitrary eval judgments. What happens if we provide evaluators with context (e.g who's the user, what's their intent) when judging LM outputs? 🧵↓

I'm at #EMNLP2024 presenting ✨TinyStyler✨, an efficient, effective, and fast method for few-shot text style transfer! Paper: aclanthology.org/2024.findings-… Demo: huggingface.co/spaces/tinysty… Code: github.com/zacharyhorvitz…

Excited to share our first preprint on a comprehensive analysis of withdrawn papers from arXiv spanning its entire history through Sept 2024, in collaboration with Thomas G. Dietterich and Jonathan Young from the arXiv.org team! A quick summary and link to the paper in this thread:

Teach GPT-4o to edit on charts and tables to ReFocus 🔍 and facilitate reasoning 🧠! 🔥 We introduce ReFocus, which edits input table and chart images to better reason visually zeyofu.github.io/ReFocus/ 🤔 Can we teach smaller models to learn such visual CoT reasoning? 🚀 Yes --

🗣️ New Paper 🗣️ Can a single AI text detector generalize to a fixed set of LLMs and domains? Our shared task results suggest yes! Winners Pangram Labs and Leidos got over 99% TPR across 467k documents spanning 11 LLMs, 8 domains, and 4 decoding strategies See thread 🧵

🚨 LLMs must grasp implied language to reason about emotions, social cues, etc. Our Google DeepMind paper presents the Implied NLI dataset. Targeting social norms 🌎 and conversational dynamics 💬, we enhance LLM understanding of real-world implication! arxiv.org/abs/2501.07719

We share Code-Guided Synthetic Data Generation: using LLM-generated code to create multimodal datasets for text-rich images, such as charts📊, documents📄, etc., to enhance Vision-Language Models. Website: yueyang1996.github.io/cosyn/ Dataset: huggingface.co/datasets/allen… Paper:

New study with a billion words! Here’s the 60-second version. ⏲️ nature.com/articles/s4159… Nature Portfolio Sharath Guntuku The University of Chicago

Luca Soldaini 🎀 Heard good things about DataDreamer from UPenn NLP arxiv.org/pdf/2402.10379

#ICLR2025 Oral LLMs often struggle with reliable and consistent decisions under uncertainty 😵‍💫 — largely because they can't reliably estimate the probability of each choice. We propose BIRD 🐦, a framework that significantly enhances LLM decision making under uncertainty. BIRD

🚀 Introducing Science-T2I - Towards bridging the gap between AI imagination and scientific reality in image generation! [CVPR 2025] 📜 Paper: arxiv.org/abs/2504.13129 🌐 Project: jialuo-li.github.io/Science-T2I-Web 💻 Code: github.com/Jialuo-Li/Scie… 🤗 Dataset: huggingface.co/collections/Ji… 🔍

#NAACL2025 How to compare cultural differences with social media data in scale? Our work uses lexica to annotate X 🇺🇸 & Weibo 🇨🇳 posts with valence (😄☹️) & arousal (🔥❄️) scores, revealing cross-cultural differences in emotional expression. aclanthology.org/2025.findings-…

🤖💬 Herding instincts… in AIs? Yes, even LLMs can follow the crowd! • 📉 Conformity ↑ when agents lack confidence but trust peers • 🧠 Presentation format shapes peer influence • 🎯 Controlled herding can boost collaboration outcomes 👉 Read more: arxiv.org/abs/2505.21588

Ever wondered what makes language models generate overly verbose, vague, or sycophantic responses? Our new paper investigates these and other idiosyncratic biases in preference models, and presents a simple post-training recipe to mitigate them! Thread below 🧵↓

In a world of geopolitical conflicts, how can AI help us navigate? Our #ACL2025-F work studies RAG robustness across 49 languages. TL;DR: 📈 boost robustness w/ multilingual RAG, 🤔 take care w/ low-resource citations 📜arxiv.org/abs/2410.01171 🤗huggingface.co/datasets/borde… 1/4 🧵