Auditing Multimodal LLM Raters: Central Tendency Bias in Clinical Ordinal Scoring

Hugging Face Daily Papers Papers

Summary

This paper investigates central tendency bias in multimodal LLMs used for clinical ordinal scoring of the Clock Drawing Test, finding that LLMs compress predictions toward the middle of the scale, disproportionately affecting critical extremes. The study extends the LLM-as-judge bias literature to clinical assessment, highlighting the need for calibration-aware evaluation before deployment.

Multimodal large language models (LLMs) are increasingly explored as automated evaluators in clinical settings, yet their scoring behavior on ordinal clinical scales remains poorly understood. We benchmark three frontier LLM families against supervised deep learning models for scoring Clock Drawing Test (CDT) images on two public datasets using the Shulman rubric. While fully fine-tuned Vision Transformers achieve the best calibration (MAE 0.52, within-1 accuracy 91%), zero-shot LLMs remain competitive on tolerance-based agreement (GPT-5 MAE 0.67, within-1 accuracy 92%) despite higher absolute error. However, per-score analysis reveals that all three LLM families exhibit a pronounced central tendency effect (systematic endpoint compression): predictions are systematically compressed toward the middle of the scale, with over-prediction at the low end (score 0 to 1) and under-prediction at the high end (score 5 to 4). This effect disproportionately affects the clinically critical extremes where accurate scoring most impacts screening decisions for cognitive impairment. Targeted ablations show that neither few-shot exemplars spanning the full score range nor removing clinical terminology from the prompt eliminates the effect. Our findings extend the LLM-as-a-judge bias literature from NLP evaluation to clinical assessment, and highlight the need for calibration-aware evaluation and post-hoc calibration before deploying LLM-based raters in high-stakes screening workflows.
Original Article
View Cached Full Text

Cached at: 05/19/26, 06:33 PM

Paper page - Auditing Multimodal LLM Raters: Central Tendency Bias in Clinical Ordinal Scoring

Source: https://huggingface.co/papers/2605.16386

Abstract

Large language models exhibit systematic bias toward central tendency when evaluating clinical assessments, particularly affecting critical score extremes important for cognitive impairment screening.

Multimodal large language models (LLMs) are increasingly explored as automated evaluators in clinical settings, yet their scoring behavior onordinal clinical scalesremains poorly understood. We benchmark three frontier LLM families against supervised deep learning models for scoringClock Drawing Test(CDT) images on two public datasets using theShulman rubric. While fully fine-tunedVision Transformersachieve the bestcalibration(MAE 0.52, within-1 accuracy 91%), zero-shot LLMs remain competitive on tolerance-based agreement (GPT-5 MAE 0.67, within-1 accuracy 92%) despite higher absolute error. However, per-score analysis reveals that all three LLM families exhibit a pronouncedcentral tendency effect(systematic endpoint compression): predictions are systematically compressed toward the middle of the scale, with over-prediction at the low end (score 0 to 1) and under-prediction at the high end (score 5 to 4). This effect disproportionately affects the clinically critical extremes where accurate scoring most impacts screening decisions for cognitive impairment. Targeted ablations show that neither few-shot exemplars spanning the full score range nor removing clinical terminology from the prompt eliminates the effect. Our findings extend theLLM-as-a-judge biasliterature from NLP evaluation toclinical assessment, and highlight the need forcalibration-aware evaluation and post-hoccalibrationbefore deploying LLM-based raters in high-stakes screening workflows.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2605\.16386

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.16386 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.16386 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.16386 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Do Benchmarks Underestimate LLM Performance? Evaluating Hallucination Detection With LLM-First Human-Adjudicated Assessment

arXiv cs.CL

This paper investigates whether standard benchmarks underestimate LLM performance by re-evaluating hallucination detection datasets using an LLM-first, human-adjudicated assessment method. The study finds that incorporating LLM reasoning into the adjudication process improves agreement and suggests that model-assisted re-evaluation yields more reliable benchmarks for ambiguity-prone tasks.