Saved in:
Bibliographic Details
Main Authors: Juvekar, Kush, Manohar, Kavya, Menon, Aditya Srinivas, Bhattacharya, Arghya, Nethil, Kumarmanas
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.13087
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914562753691648
author Juvekar, Kush
Manohar, Kavya
Menon, Aditya Srinivas
Bhattacharya, Arghya
Nethil, Kumarmanas
author_facet Juvekar, Kush
Manohar, Kavya
Menon, Aditya Srinivas
Bhattacharya, Arghya
Nethil, Kumarmanas
contents Fine-tuning multilingual ASR models like Whisper for low-resource languages often improves read speech but degrades spontaneous audio performance, a phenomenon we term studio-bias. To diagnose this mismatch, we introduce Vividh-ASR, a complexity-stratified benchmark for Hindi and Malayalam across four tiers: studio, broadcast, spontaneous, and synthetic noise. Through a controlled study of learning-rate timing and curriculum ordering, we find that early large parameter updates improve global WER by 12 absolute points, while a hard-to-easy curriculum adds gains for spontaneous speech. These findings motivate reverse multi-stage fine-tuning (R-MFT), a training recipe that enables a parameter-efficient 244M Whisper model to match or exceed conventionally fine-tuned 769M counterparts. Representational analysis via CKA and SVD reveals effective schedules concentrate adaptation in the decoder, preserving the pre-trained encoder's acoustic geometry. We release the benchmark and models.
format Preprint
id arxiv_https___arxiv_org_abs_2605_13087
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Vividh-ASR: A Complexity-Tiered Benchmark and Optimization Dynamics for Robust Indic Speech Recognition
Juvekar, Kush
Manohar, Kavya
Menon, Aditya Srinivas
Bhattacharya, Arghya
Nethil, Kumarmanas
Computation and Language
Artificial Intelligence
Fine-tuning multilingual ASR models like Whisper for low-resource languages often improves read speech but degrades spontaneous audio performance, a phenomenon we term studio-bias. To diagnose this mismatch, we introduce Vividh-ASR, a complexity-stratified benchmark for Hindi and Malayalam across four tiers: studio, broadcast, spontaneous, and synthetic noise. Through a controlled study of learning-rate timing and curriculum ordering, we find that early large parameter updates improve global WER by 12 absolute points, while a hard-to-easy curriculum adds gains for spontaneous speech. These findings motivate reverse multi-stage fine-tuning (R-MFT), a training recipe that enables a parameter-efficient 244M Whisper model to match or exceed conventionally fine-tuned 769M counterparts. Representational analysis via CKA and SVD reveals effective schedules concentrate adaptation in the decoder, preserving the pre-trained encoder's acoustic geometry. We release the benchmark and models.
title Vividh-ASR: A Complexity-Tiered Benchmark and Optimization Dynamics for Robust Indic Speech Recognition
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2605.13087