From Scoring to Explanations: Evaluating SHAP and LLM Rationales for Rubric-based Teaching Quality Assessment
Quick Answer
The study evaluates SHAP and LLM rationales for rubric-based teaching quality assessment, revealing that fine-tuned pretrained language models outperform LLMs in accuracy but struggle with label compression.
Quick Take
The study evaluates SHAP and LLM rationales for rubric-based teaching quality assessment, revealing that fine-tuned pretrained language models outperform LLMs in accuracy but struggle with label compression. SHAP attributions provide more reliable explanations than LLM-generated rationales, making it a better choice for educational assessments.
Key Points
- Fine-tuned pretrained language models outperform LLMs in prediction accuracy across 6k transcript segments.
- SHAP identifies key sentences driving model predictions, yielding larger prediction shifts than LLM rationales.
- LLM rationales show limited and inconsistent influence compared to robust SHAP attributions.
- The framework aids in systematic evaluation of scoring models and their explanations in education.
- Findings suggest SHAP is preferable for high-stakes rubric-based assessments.
Article Content
From source RSS / original summaryarXiv:2606. 05180v1 Announce Type: new Abstract: Automated scoring models are increasingly used to assign rubric-based quality ratings to complex language performances, including classroom transcripts, yet they typically provide little insight into why a particular score is produced. We propose a general framework for sentence-level interpretability of rubric-based scoring that combines model-agnostic Shapley-value attributions with rationales generated by large language models (LLMs).
Instantiated on the Quality of Feedback dimension of the CLASS framework using the NCTE corpus, the framework enables systematic comparison of fine-tuned pretrained language models (PLMs) and prompted LLMs on both scoring performance and explanation faithfulness. Across 6k annotated transcript segments, fine-tuned PLMs outperform LLMs in prediction accuracy but exhibit label compression toward mid-scale scores.
Deletion-based tests show that SHAP identifies sentences that reliably drive model predictions, producing typically larger and more coherent prediction shifts than LLM-generated rationales. Cross-model analyses further reveal that SHAP attributions transfer robustly across architectures, whereas LLM rationales exert limited and inconsistent influence.
Overall, the findings demonstrate that SHAP provides more faithful and transferable explanations for rubric-based scoring, and that the proposed framework offers a principled basis for evaluating both scoring models and their explanations in high-stakes educational settings and other rubric-based language assessment tasks.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents?
The REFLECT benchmark reveals that current LLM judges are unreliable, achieving below 55% accuracy in evaluating reasoning and evidence use, highlighting the need for improved evaluation methods for deep research agents.