Prasanth Yadla

Cross-Platform Natural Language Processing Framework for Electoral Misinformation Detection

Compressing Streaming Neural Audio Encoders via Latent-Space Distillation

This work studies how to compress a tokenizer by distillation, taking as the supervision target neither the discrete token nor the output distribution but the pre-quantizer latent the model actually consumes - the last representation the two token interfaces share.

A survey of parameter-efficient fine-tuning techniques for audio foundation models

This survey systematically reviews PEFT methods that update only 0.1–5% of model parameters while achieving good performance across diverse audio tasks, the first survey to comprehensively examine PEFT techniques specifically for audio foundation models.