
YP
Yen-Ting Piao, Jay Chiehen Liao, Wei-Tang Chien, Toshiki Ogimoto, Shang-Tse Chen, Yun-Nung Chen, Chun-Yi Lee, Shao-Yuan Lo
· 1 min read
ResearcharXiv cs.CL
When Demonstrations Fail: Diagnosing the Limits of In-Context Learning in Large Audio-Language Models with Progressive Cue Removal
arXiv:2603.20433v2 Announce Type: replace-cross
Abstract: While Large Audio-Language Models (LALMs) have been shown to exhibit degraded instruction-following capabilities, their ability to infer task patterns from in-context examples with audio remains understudied. To address this gap, we design a three-stage evaluation pipeline that progressively reduces textual guidance to systematically evaluate LALMs' in-context learning ability in the audio modality. Evaluating six LALMs across four audio understanding tasks under two output constraint categories, we uncover a consistent asymmetry across LALMs: in-context demonstrations reliably improve format compliance but fail to improve the core task performance. This suggests that LALMs can glean surface-level formatting patterns from demonstrations but may struggle to leverage cross-modal semantic grounding to reliably infer task objectives from examples with audio, highlighting potential limitations in current cross-modal integration. We further probe how demonstrations are used through two complementary analyses, demonstration label shuffling and attention knockout on demonstration spans, both showing that LALMs leverage in-context examples primarily to establish the output label space and format rather than to learn a meaningful input-output correspondence.
Original source
This story was published by arXiv cs.CL and written by Yen-Ting Piao, Jay Chiehen Liao, Wei-Tang Chien, Toshiki Ogimoto, Shang-Tse Chen, Yun-Nung Chen, Chun-Yi Lee, Shao-Yuan Lo. SyncAI.news shows a preview; the complete article is on the publisher's site.
Read the full story on arxiv.org


