✨ Thrilled to share that our paper, "Evaluating Input Feature Explanations through a Unified Diagnostic Evaluation Framework," has been accepted at NAACL 2025—with an Oral presentation! 🎉 A huge thank you to my incredible supervisors @Isabelle and @atanasovapepa . #NAACL2025
Jingyi Sun
@JingyiSun_
PhD Fellow @CopeNLU @uni_copenhagen. Explainable AI, Natural Language Processing, ML.
6 Following 37 Followers 9 Posts
Latest posts 3
⚡️ Excited to share our new paper on improving the faithfulness of natural language explanations with graph-guided highlight explanations!
⚡️ Read it here: arxiv.org/pdf/2412.12318.
Grateful for the backing of @CopeNLU, @Sca_DS, and @NLLG_lab! 🚀
#NLProc #XAI
⚡️ Read it here: arxiv.org/pdf/2412.12318.
Grateful for the backing of @CopeNLU, @Sca_DS, and @NLLG_lab! 🚀
#NLProc #XAI
🚀📄 New Paper!View quoted post on X
Ever wondered how to make Natural Language Explanations (NLEs) more faithful to a model’s reasoning? 🤔
Introducing our Graph-Guided Framework G-TEx! 🌟 Using GNNs + ✨highlight explanations✨, we improve NLE faithfulness. 🔍
📄 Read more: arxiv.org/pdf/2412.12318
🎉Exciting News !🎉 What's the ultimate type of input feature explanation? We've created a unified framework to determine the best regarding different diagnostic properties!
See our paper at arxiv.org/abs/2406.15085. Work done with the amazing @IAugenstein and @atanasovapepa.
See our paper at arxiv.org/abs/2406.15085. Work done with the amazing @IAugenstein and @atanasovapepa.