MB using only 10% of the target-domain speech achieves word error rates comparable to, or better than, conventional ASR fine-tuning with the full dataset, indicating that small amounts of speech provide a strong modality-alignment signal.
This work analyzes three datasets: ANDROIDS, DAIC-WOZ, E-DAIC and identifies a systematic bias from interviewer prompts in semi-structured interviews that highlights a cross-dataset, architecture-agnostic bias and emphasizes the need for analyses that localize decision evidence by time and speaker to ensure models lear...
Hasindri Watawana, Sergio Gastón Burdisso, Diego A. Moreno-Galv'an et al.· International Conference on...· 1 citation
A discriminative adaptation that reads the final prompt token's hidden state through a classification head, producing a label in one forward pass without modifying the backbone, gives a controlled comparison of generative and discriminative inference in an otherwise identical speechLLM.
The results show that encoder-based metrics remain highly competitive, while generative LLMs perform strongly in hypothesis comparison and improve the interpretability of ASR evaluation.
Thibault Bañeras-Roux, Shashi Kumar, Driss Khalil et al.· 0 citations
This work shows that Group Relative Policy Optimization (GRPO) extracts far more from the same synthetic speech than SFT, and traces the gain to behavior rather than representation: GRPO reduces insertion errors by improving stopping calibration and speech-to-text alignment by better anchoring attention to audio, leavi...
Shashi Kumar, Yanis Labrak, Hasindri Watawana et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
