The rapid rise of large language models (LLMs) is reshaping the landscape of automatic assessment in education. While these systems demonstrate substantial advantages in adaptability to diverse question types and flexibility in output formats, they also introduce new challenges related to output uncertainty, stemming from the inherently probabilistic nature of LLMs. Output uncertainty is an inescapable challenge in automatic assessment, as assessment results often play a critical role in informing subsequent pedagogical actions, such as providing feedback to students or guiding instructional decisions. Unreliable or poorly calibrated uncertainty estimates can lead to unstable downstream interventions, potentially disrupting students' learning processes and resulting in unintended negative consequences. To systematically understand this challenge and inform future research, we benchmark a broad range of uncertainty quantification methods in the context of LLM-based automatic assessment. Although the effectiveness of these methods has been demonstrated in many tasks across other domains, their applicability and reliability in educational settings, particularly for automatic grading, remain underexplored. Through comprehensive analyses of uncertainty behaviors across multiple assessment datasets, LLM families, and generation control settings, we characterize the uncertainty patterns exhibited by LLMs in grading scenarios. Based on these findings, we evaluate the strengths and limitations of different uncertainty metrics and analyze the influence of key factors, including model families, assessment tasks, and decoding strategies, on uncertainty estimates. Our study provides actionable insights into the characteristics of uncertainty in LLM-based automatic assessment and lays the groundwork for developing more reliable and effective uncertainty-aware grading systems in the future.
Figures & tables
Question
Size
Score
Subject
Question
Size
Score
Subject
ASAP
500
0-3
Essay Writing
Semveal
500
0-3
Mixture
T1-DCI
124
0-3
Chemistry
T1-SEP
124
0-3
Chemistry
T2-DCI
141
0-3
Chemistry
T2-SEP
141
0-3
Chemistry
T1-CK
208
0-3
Math Pedagogy
T1-PCK
188
0-3
Math Pedagogy
T2-CK
202
0-3
Math Pedagogy
T2-PCK
188
0-3
Math Pedagogy
U41Q1
159
0-4
Earth Science
U41Q2
184
0-4
Earth Science
Table 1. List of questions evaluated by our benchmark.
Proprietary Models
Model
Model ID
Scale
Calude Haiku 4.5 ( Anthropic, 2025b )
claude-haiku-4-5-20251001
/
Calude Sonnet 4.5 ( Anthropic, 2025a )
claude-sonnet-4-5-20250929
/
Gemini 2.5 Pro ( Team et al., 2023 )
gemini-2.5-pro
/
Gemini 3 Flash ( Team et al., 2023 )
gemini-3-flash-preview
/
GPT-4o mini ( OpenAI, 2024 )
gpt-4o-mini
/
Table 2. List of LLMs evaluated by our benchmark.
Method
AUROC
C-index
AUARC
AUERC
CE
4.71
5.44
4.91
5.50
FSD
5.20
5.67
5.24
5.61
MAR
4.82
5.50
5.01
5.51
Numset
5.86
6.55
5.75
6.32
JS_NAD
6.97
6.69
7.03
6.78
JS_GE
8.00
7.40
8.04
7.56
Table 3. Average ranks of methods across questions, datasets, and generation strategies on effectiveness metrics.
Figure 1. Distribution of various in-group average rank for each model, question and generation strategies variants.
Figure 2. Method Comparison between stablelm and gemini over different metrics on the ASAP question.
Figure 3. Method Comparison between stablelm and gemini over different metrics on the SemEval question.
Method
Delta
Spearmanr
Method
Delta
Spearmanr
CE
5.30
4.51
NLI_NAD
8.50
7.01
FSD
7.69
6.58
NLI_GE
8.05
9.93
MAR
8.65
5.37
NLI_Eigen
12.10
10.56
Numset
2.47
6.26
NLI_DSE
13.57
12.84
JS_NAD
2.84
3.86
Embed_NAD
6.00
4.92
JS_GE
1.89
7.23
Embed_GE
5.02
7.89
Table 4. Average ranks of methods across questions, datasets, and generation strategies on stability metrics.
Figure 4. Distribution of various in-group average rank in stability across model, question and strategy.
Figure 5. Mean and variance of aggregated pairwise correlation values among methods across questions, models, and generation strategies.
Automatic Short Answer Grading (ASAG) with generative large language models (LLMs) has recently demonstrated strong performance without task-specific fine-tuning, while also enabling the generation of synthetic feedback for educational assessment. Despite these advances, LLM-based grading remains imperfect, making reliable confidence estimates essential for safe and effective human-AI collaboration in educational decision-making. In this work, we investigate confidence estimation for ASAG with LLMs by jointly considering model-based confidence signals and dataset-derived uncertainty. We systematically compare three model-based confidence estimation strategies, namely verbalizing, latent, and consistency-based confidence estimation, and show that model-based confidence alone is insufficient to reliably capture uncertainty in ASAG. To address this limitation, we propose a hybrid confidence framework that integrates model-based confidence signals with an explicit estimate of dataset-derived aleatoric uncertainty. Aleatoric uncertainty is operationalized by clustering semantically embedded student responses and quantifying within-cluster heterogeneity. Our results demonstrate that the proposed hybrid confidence measure yields more reliable confidence estimates and improves selective grading performance compared to single-source approaches. Overall, this work advances confidence-aware LLM-based grading for human-in-the-loop assessment, supporting more trustworthy AI-assisted educational assessment systems.
Longwei Cong, Sonja Hahn, Sebastian Gombert +3
DIPF | Leibniz Institute for Research and Information in Education, 60323 Frankfurt am Main, Germany · Faculty of Computer Science, Goethe University Frankfurt, 60629 Frankfurt am Main, Germany · Chemnitz University of Technology, 09111 Chemnitz, Germany
Large Language Models (LLMs) show promise for automated grading, but their outputs can be unreliable. Rather than improving grading accuracy directly, we address a complementary problem: predicting when an LLM grader is likely to be correct. This enables selective automation where high-confidence predictions are processed automatically while uncertain cases are flagged for human review. We compare three confidence estimation methods (self-reported confidence, self-consistency voting, and token probability) across seven LLMs of varying scale (4B to 120B parameters) on three educational datasets: RiceChem (long-answer chemistry), SciEntsBank, and Beetle (short-answer science). Our experiments reveal that self-reported confidence consistently achieves the best calibration across all conditions (avg ECE 0.166 vs 0.229 for self-consistency). Surprisingly, self-consistency remains 38% worse despite requiring 5× the inference cost. Larger models exhibit substantially better calibration though gains vary by dataset and method (e.g., a 28% ECE reduction for self-reported), with GPT-OSS-120B achieving the best calibration (avg ECE 0.100) and strong discrimination (avg AUC 0.668). We also observe that confidence is strongly top-skewed across methods, creating a ``confidence floor'' that practitioners must account for when setting thresholds. These findings suggest that simply asking LLMs to report their confidence provides a practical approach for identifying reliable grading predictions. Code is available here.
Scaling educational assessment with large language models requires not just accuracy, but the ability to recognize when predictions are trustworthy. Instruction-tuned models tend to be overconfident, and their reliability deteriorates as curricula evolve, making fully autonomous deployment unsafe in high-stakes settings. We introduce CHiL(L)Grader, the first automated grading framework that incorporates calibrated confidence estimation into a human-in-the-loop workflow. Using post-hoc temperature scaling, confidence-based selective prediction, and continual learning, CHiL(L)Grader automates only high-confidence predictions while routing uncertain cases to human graders, and adapts to evolving rubrics and unseen questions. Across three short-answer grading datasets, CHiL(L)Grader automatically scores 35-65% of responses at expert-level quality (QWK >= 0.80). A QWK gap of 0.347 between accepted and rejected predictions confirms the effectiveness of the confidence-based routing. Each correction cycle strengthens the model's grading capability as it learns from teacher feedback. These results show that uncertainty quantification is key for reliable AI-assisted grading.
Pranav Raikote, Korbinian Randl, Ioanna Miliou +2
Department of Computer and Systems Sciences, Stockholm University, 164 25 Kista, Sweden