CambridgeLTL

📏 "UNCLE" by @caiqizh - New benchmark for evaluating uncertainty expression in long-form generation aclanthology.org/2025.emnlp-mai…

🌐 "ReCoVeR the Target Language" by @h_sterz et al - Reducing language confusion while maintaining task performance aclanthology.org/2025.findings-…

And that's a wrap! At the #EMNLP2025 our lab had an incredible showing with 8 papers! Congratulations to all our amazing students! 🎉 Links to papers: [🧵below]

AI that is “forced to be good” v “genuinely good” Should we care about the difference? (yes!) We’re releasing the first open implementation of character training. We shape the persona of AI assistants in a more robust way than alternatives like prompting or activation steering.

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)

Check out this amazing work from Tiancheng!

Our LTL team had an amazing time at @M2lSchool in Croatia 🇭🇷! 🏆 @h_sterz won Best Poster Award! → arxiv.org/html/2509.1481… 🎤 @licwu delivered an amazing lecture → drive.google.com/file/d/1feqnLe…

- Large Language Models are Miscalibrated In-Context Learners @li_chengzu - WinSpot: GUI Grounding Benchmark with Multimodal Large Language Models Zack Hui

- Retrofitting Large Language Models with Dynamic Tokenization @licwu @bminixhofer - Culturally Aware and Adapted NLP: A Taxonomy and a Survey of the State of the Art @ChenLiu47008770 - DARE: Diverse Visual Question Answering with Robustness Evaluation Hannah Sterz, @licwu

A 🧵of our papers at @aclmeeting 2025 in Vienna: - Conformity in Large Language Models @caiqizh - LoGU: Long-form Generation with Uncertainty Expressions @caiqizh - iNews: A Multimodal Dataset for Modeling Personalized Affective Responses to News @tiancheng_hu @nigelhcollier

🔥 We teach LLMs to say how confident they are on-the-fly during long-form generation. 🤩No sampling. No slow post-hoc methods. Not limited to short-form QA! ‼️Just output confidence in a single decoding pass. ✅Better calibration! 🚀 20× faster runtime. arXiv:2505.23912 👇

We achieved the first instance of successful subword-to-byte distillation in our (just updated) paper. This enables creating byte-level models at a fraction of the cost of what was needed previously. As a proof-of-concept, we created byte-level Gemma2 and Llama3 models. 🧵

🚀Let’s Think Only with Images. No language and No verbal thought.🤔 Let’s think through a sequence of images💭, like how humans picture steps in their minds🎨. We propose Visual Planning, a novel reasoning paradigm that enables models to reason purely through images.