Woongyeong Yeo

@wgcyeo

M.S. Student @KAIST_AI

wgcyeo.github.io

Joined September 2024

137

Following

Followers

Replies Reposts Media

Pinned
Woongyeong Yeo
@wgcyeo
Sep 29
💡 Success under uncertainty is hard to repeat, while confident failures tend to recur. How can we learn from both to improve exploration in LLM reasoning? Introducing EAPO (Entropic Advantage Policy Optimization), entropy-guided credit assignment for RLVR that treats success
Minki Kang and 5 others
1
Woongyeong Yeo
@wgcyeo
Sep 25
🥳 Happy to share that our new work on alignment for contextual integrity, SelfCI, has been accepted to #NeurIPS2026! So excited to attend three different (but somehow similar now) conferences —CVPR, ACL, and NeurIPS—in a single year! See you at somewhere (🇦🇺 or 🇺🇸) 😆
Woongyeong Yeo
@wgcyeo
May 21
📢 New preprint out on contextual integrity (CI) and a new Product-of-Experts (PoE) view of self-distillation! Introducing SelfCI, a novel self-distillation framework that operationalizes CI by optimizing for the intersection of task utility and minimal disclosure. 🧵👇
Woongyeong Yeo
@wgcyeo
Sep 8
Unbelievable. Very excited (and also shocked) about what the next generation of AI is becoming capable of. Maybe AI has already reached, and even surpassed, the frontiers of human capability 🫢
OpenAI
@OpenAI
Sep 8
We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics. The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra. The problem
Woongyeong Yeo
@wgcyeo
Aug 28
Happy to share that HINT-SD has been accepted to Findings of #EMNLP2026! 🎉 Long-horizon agent training suffers from relevance sparsity: rewards are sparse, and only a few action spans need corrections. HINT-SD uses full-trajectory hindsight to target and distill selected spans.
2
Woongyeong Yeo
@wgcyeo
Jul 7
My first oral presentation is happening today in just a few hours! If you’re interested in any-to-any RAG, please stop by our talk on UniversalRAG (universalrag.github.io) at #ACL2026. I’d love to see you there! 🎙️ Oral Session G: Retrieval-Augmented Language Models 🕙 11:00 AM
Woongyeong Yeo
@wgcyeo
Apr 8
🔍 Is a single embedding space really enough for multimodal RAG? Excited to share that UniversalRAG has been accepted to the #ACL2026 main conference! 🥳 We introduce the first any-to-any multimodal RAG framework, enabling retrieval across diverse modalities and granularities.