
PhD student at National Taiwan University @ntu_spml
Taiwan
Joined August 2021
- Tweets39
- Following99
- Followers20
- Likes56
🚨 New paper alert! 🎙️
From INSPIRE to IRIS-Embed!
Can speech retrieval understand what users want to find, beyond what was said?
Excited to share IRIS-Embed, our new approach to instruction-aware speech embeddings. 🧵
📄 arxiv.org/abs/2611.09122
📊 Beyond INSPIRE:
On 15 MAEB audio tasks, IRIS-Embed improves macro-average scores by 9.4–19.8 points over its corresponding LALM baselines.
The gains extend beyond instruction-aware speech retrieval.
1
26
From benchmarking the challenge with INSPIRE to building a model with IRIS-Embed, excited about this direction!
Many thanks to my advisors Hung-yi Lee @HungyiLee2, and to my amazing collaborator Cheng-Han Chiang @dcml0714!
💻 Code & Model: github.com/lca0503/IRIS
24
Chen-An Li retweeted
New paper❗
Does understanding how AI models work actually make them safer?
This assumption often underlines interpretability research, but different subfields operationalise this intuition in very different ways. Our survey of mechanistic interpretability (MI) for safety provides an opinionated guide to the field of MI, detailing conceptual approaches and techniques under this broad banner and how different types of work relates to the mission of making AI model safer. 🧵
🚨New paper alert, Our #Interspeech2026 paper is now on arXiv! 🎙️🎉
INSPIRE: A Benchmark for Instruction-Aware Speech Retrieval
What if the same speech query should return different results depending on user intent? 🤔
Speaker? Content? Style? Background sound? 🧵👇
📄 Paper: arxiv.org/abs/2608.16203
💻 Code: github.com/lca0503/INSPIRE
📊 Data: github.com/lca0503/INSPIRE
1
23
🙏 Great thanks to my advisor, Hung-yi Lee @HungyiLee2, for his guidance and support throughout this work! ✨
See you at #Interspeech2026! 🎙️
7
Chen-An Li retweeted
🎊 Excited to share that our paper MI-Fuse will be presented at ICASSP 2026!
🔗 Paper Link: arxiv.org/pdf/2509.20706
Come see us tomorrow!
⏳ Time: Thursday, May 7 | 10:20 – 10:40
📍 Session: SLP-L14 Speech Emotion Recognition and Language Models (Oral)
✨ Location: Room 114
Heading to #ICASSP2026? 🇪🇸🚀
Stop by our poster: “When Silence Matters: The Impact of Irrelevant Audio on Text Reasoning in Large Audio-Language Models”.
We found that irrelevant audio reduces accuracy & increases prediction volatility in LALMs.
⏰ Wed, 6 May | 14:00 - 16:00
Our results highlight cross-modal interference as a major robustness challenge, emphasizing the need for more efficient fusion strategies.
📄 Paper: arxiv.org/pdf/2510.00626
💻 Code: github.com/lca0503/AudioInte…
See you tomorrow in Barcelona! 👋🇪🇸
🍀 Many thanks to my advisor Hung-yi Lee @HungyiLee2, and to my amazing collaborator Tzu-Han Lin @tzuhan_0316!
1
46



