Explainability
Methods and studies that help explain how audio and speech language models make decisions, including post-hoc interpretability, attribution, reasoning analyses, and faithfulness.
ICASSP 2027 Workshop
Explainability, Robustness, and Controllability of Language Modeling for Speech and Audio
Explainability, robustness, and controllability for audio and speech language models.
Scope
Language modeling for speech and audio
Robustness
Work on noise, adversarial inputs, prompt injections, hallucinations, temporal shifts, safety, security, and behavior under distribution shift.
Controllability
Architectures and representations that enable control over speech, music, and audio generation or editing while preserving unrelated content.
Call for Papers
Submissions
We invite four-page submissions in the ICASSP format, with an optional appendix. Submissions will be double blind and peer reviewed on OpenReview.
Paper submission deadline
Paper acceptance notification
Camera-ready paper submission
Invited Speakers
Confirmed invited speakers
Tentative Program
One-day workshop
The program will combine invited talks, contributed oral presentations, posters, and a moderated panel with audience questions.
- Opening remarks
- Plenary speaker 1
- Contributed oral presentations
- Plenary speaker 2
- Poster session and lunch break
- Plenary speaker 3
- Contributed oral presentations
- Plenary speaker 4
- Coffee break
- Plenary speaker 5
- Poster session
- Panel discussion
- Closing remarks
Organizers