SEACrowd

We’re thrilled that SEA-VL has been accepted to the ACL 2025 (Main)!

Thank you to everyone who contributed to this project 🥳

Paper: arxiv.org/abs/2503.07920
Project: seacrowd.github.io/seavl-launch/

#ACL2025NLP #SEACrowd #ForSEABySEA

Let’s build a VLM that sees and celebrates Southeast Asia—together. 💪

@josephimperial.bsky.social @samuel-cahyawijaya.bsky.social @jcblaise.bsky.social @ruochenzhang.bsky.social @rianadamr.bsky.social @antonrufino.bsky.social

🚨 SEA-VL Phase 2 - Building Vision-Language Models for Southeast Asia: Call for Contributors Welcome! SEA-VL is a global community project organized by the SEACrowd community to push the boundaries of vision and language research in Southeast Asia (SEA). We recently completed Phase 1 of this ...

Whether you’re a researcher, developer, artist, linguist, photographer, student, or simply someone who loves Southeast Asia, your voice and skills matter. Join us!

📥 Apply now: seacrowd.github.io//seavl-phase...

💬 Questions? Join the conversation on Discord: discord.gg/XXRHFuvkTA

Why contribute?
🤝 Work with an international team of passionate researchers
🏅 Earn points for every contribution—with opportunities for a certificate, exclusive merch (t-shirt & keychain), and even co-authorship on our final paper

We are looking for contributors who can:
🔹Submit culturally relevant images from SEA
🔹Annotate image submissions
🔹Translate existing benchmarks to SEA languages
🔹Create high-quality questions for multicultural images from SEA
🔹Create high-quality prompts for image generation with our VLM

We want build the first open-source vision-language model (VLM) that fully captures Southeast Asia’s rich cultures, languages, and everyday life!

📢 Calling all SEA-passionate individuals!

SEACrowd is excited to launch our contributor call for SEA-VL Phase 2: Building Visual Language Models for Southeast Asia! 🌏

After the success of Phase 1, we're now taking on a bigger mission (see thread)👇

The ACL Special Interest Group on SEA NLP Southeast Asia

Interested in pushing research for Southeast Asian languages? We're happy to welcome you in SEACrowd and SIGSEA! See links below:

SIGSEA: www.sigsea.org/home
Discord: discord.gg/XXRHFuvkTA

Introducing SEA-VL with 1.3M culturally relevant images—50x larger than existing datasets!
🔍 Key insights:
✅ Crowdsourcing: good accuracy but slow & costly
✅ Image Crawling: ~85% cultural relevance
❌ Image Generation fails to capture SEA nuances & faces licensing issues

Why is this important?
✅ AI models trained on culturally relevant data can better understand local contexts, traditions, and languages.
✅ Community contributions ensure AI does not misrepresent local identities.
✅ We empower local communities in AI development.

💡 That’s why we created SEA-VL, an open-source initiative designed to bridge the resource gap and provide AI models with more accurate, culturally relevant data from SEA. But we couldn’t have done it alone!

#NoLanguageLeftBehind #SoutheastAsia

AI is shaping the future, but how often does it reflect the cultures, languages, and traditions of Southeast Asia? Not enough!

Most VL datasets used to train AI are dominated by Western-centric data, leaving Southeast Asian cultures largely underrepresented.

SEA-VL: Building AI for Southeast Asian Research 🌏

We release SEA-VL, the largest vision-language dataset tailored for SEA’s diverse culture.

📜 arXiv: arxiv.org/abs/2503.07920
🤗 Data: huggingface.co/collections/...

Check the thread 🧵