DoWhatISay provides spoken and written prompt variants across tasks and languages for SLLM evaluation, showing text prompts outperform spoken ones except in speech-output tasks.
Twists, humps, and pebbles: Multilingual speech recognition models exhibit gender performance gaps
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Do What I Say: A Spoken Prompt Dataset for Instruction-Following
DoWhatISay provides spoken and written prompt variants across tasks and languages for SLLM evaluation, showing text prompts outperform spoken ones except in speech-output tasks.