Reproduce “Zero-Shot Spoken Language Understanding with SLAM-ASR: What Do Monolingual ASR-Trained Linear Projectors Enable?” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Edwin Arkel Rios, Sneh Sourabh, Jocelyn Zaruma, Girish Ewoorkar, Julian Mack, Bo-Cheng Lai
The authors test whether a transcription-trained bridge lets a frozen language model answer questions about speech. For 11 languages, they train separate projectors on transcription data between frozen Whisper-medium and Tiny Aya Global, without SLU examples. Mean seven-topic balanced accuracy is 23.2% from audio versus 71.5% from reference text, implicating the audio-to-decoder pathway but not isolating the projector.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.