Pinned
New Paper Alert!‼️
🌍 Culture Affordance Atlas: Teaching AI to See the World as People Actually Live It
(Joint work with Longju Bai @Longju_Bai, Oana Ignat @OanaIgnatRo, and Rada Mihalcea @radamihalcea)
Joan Nwatu
@_agirlyengineer
- Ecstatic 😁 to share that our work on improving vision-language model performance for lower-income communities was accepted to #NAACL2025 main track!! Check out our paper: arxiv.org/abs/2407.02623Vision-language models often miss the mark on non-Western, low-income data, reinforcing AI's lack of diversity.🧵(1/8)
- Vision-language models often miss the mark on non-Western, low-income data, reinforcing AI's lack of diversity.🧵(1/8)
- I am grateful for the opportunity to have contributed to this cool project that takes a step in the right direction towards improving representation in AI. Please check out our paper. arxiv.org/pdf/2403.07687…What models we can build and what datasets we can construct often comes down to 💰 -- can one afford the budget? Our work proposes strategies to collect and use geo-diverse annotations on a budget, to balance model performance and annotation costs.
- Last week, I gave an oral presentation for the paper, Bridging the Digital Divide: Performance Variation across Socio-Economic Factors in Vision-Language Models at the #emnlp2023 conference in Singapore! I had an amazing experience sharing this work with others at the conference
