Gregor Geigle
@GregorGeigle
PhD student @Uni_WUE| NLP, Multimodal Vision+Language
Introducing MVL-SIB, a massively multilingual vision-language benchmark for cross-modal topic matching in 205 languages! 🤔Tasks: Given images (sentences), select topically matching sentence (image). Arxiv: arxiv.org/abs/2502.12852 HF: huggingface.co/datasets/WueNL… Details👇
Thanks to a GPU grant by @huggingface , you can try out Centurio Aya here: huggingface.co/spaces/WueNLP/… (code shamelessly adapted from @mervenoyann demo of Llava-Next)
Want to train a *multilingual* LVLM but not sure how? Or looking for a strong model to use? Presenting "Centurio: On Drivers of Multilingual Ability of Large Vision-Language Model"! Arxiv: arxiv.org/abs/2501.05122 HF Collection: huggingface.co/collections/Wu…
📣Happy to (pre-)release my Fleurs-SLU benchmark to evaluate massively multilingual spoken language understanding on SIB & Belebele. Work done at @Mila_Quebec with @davlanade @gg42554 @licwu Datasets: huggingface.co/datasets/WueNL… huggingface.co/datasets/WueNL… Details to follow👇
Excited to present NLLB-LLM2Vec at @emnlpmeeting Tuesday 2pm! Drop by our poster to chat about multilingual & multimodal research. NLLB-LLM2Vec can now easily be used with @huggingface AutoModels — try it esp. for embedding low-resource languages! 🌐 huggingface.co/fdschmidt93/NL…
Introducing NLLB-LLM2Vec! 🚀 We fuse the NLLB encoder & Llama 3 8B trained w/ LLM2Vec to create NLLB-LLM2Vec which supports cross-lingual NLU in 200+ languages🔥 Joint work w/ Philipp Borchert, @licwu, and @gg42554 during my great research stay at @cambridgeltl
Awesome work! I don't know why but it feels strange to see my University logo in the same figure as these big labs & groups😅
🌍 I’ve always had a dream of making AI accessible to everyone, regardless of location or language. However, current open MLLMs often respond in English, even to non-English queries! 🚀 Introducing Pangea: A Fully Open Multilingual Multimodal LLM supporting 39 languages! 🌐✨…
We recently tested public LVLMs for fine-grained object classification (arxiv.org/abs/2406.14496). Models pretraining with >1B examples crushed it compared to LLaVA & co. PaliGemma was excellent, too, despite its size and this part of the report explains now why.
