Sreyan Ghosh

A look into the pitfalls of instruction tuning for AI language models.

RECAP uses advanced techniques to generate accurate audio captions without retraining.

This article discusses hallucinations in LVLMs and proposes methods to tackle them.

ABEX uses Abstract-and-Expand to enhance training data for natural language understanding tasks.

GAMA improves audio processing by merging sound and language insights.

ReCLAP enhances audio classification with detailed prompts for better accuracy.