Rubric-Guided Fine-tuning of SpeechLLMs for Multi-Aspect, Multi-Rater L2 Reading-Speech Assessment

TL;DR

This paper presents a rubric-guided, uncertainty-aware fine-tuning method for SpeechLLMs that improves alignment with human ratings in multi-aspect L2 speech assessment, emphasizing interpretability and reliability.

Contribution

It introduces a novel rubric-guided reasoning framework with uncertainty calibration for SpeechLLMs, enhancing interpretability and alignment with human judgments.

Findings

Outperforms baseline regression and classification methods.

Achieves strong alignment with human ratings for fluency and prosody.

Highlights challenges in automatically assessing accuracy.

Abstract

Reliable and interpretable automated assessment of second-language (L2) speech remains a central challenge, as large speech-language models (SpeechLLMs) often struggle to align with the nuanced variability of human raters. To address this, we introduce a rubric-guided reasoning framework that explicitly encodes multi-aspect human assessment criteria: accuracy, fluency, and prosody, while calibrating model uncertainty to capture natural rating variability. We fine-tune the Qwen2-Audio-7B-Instruct model using multi-rater human judgments and develop an uncertainty-calibrated regression approach supported by conformal calibration for interpretable confidence intervals. Our Gaussian uncertainty modeling and conformal calibration approach achieves the strongest alignment with human ratings, outperforming regression and classification baselines. The model reliably assesses fluency and prosody while highlighting the inherent difficulty of assessing accuracy. Together, these results demonstrate that rubric-guided, uncertainty-calibrated reasoning offers a principled path toward trustworthy and explainable SpeechLLM-based speech assessment.