LLMs have been increasingly used to catch data quality issues automatically, but we know very little about how consistent these judgments actually are. This study tests an LLM on two e-commerce data quality tasks, entity matching and brand mislabeling, against rule based baselines and human verified ground truth, under both zero-shot and few-shot prompting. On entity matching while using the Abt Buy benchmark (2,194 labeled pairs), a simple rule based baseline (F1=0.950) performed about as well as LLM zero shot prompting (F1=0.948). Moreover, a few-shot prompt revision that looked effective on a small validation sample reduced full-scale performance to F1=0.914. This showed that small sample prompt evaluation can be misleading. On brand mislabeling detection, using 500 Amazon product listings with synthetically injected labeling errors, the LLM clearly outperformed a naive rule based baseline (F1=0.833 vs 0.721), because it could draw on background knowledge of brand product relationships that a simple rule could not access. Testing consistency across repeated runs (200 pairs, 5 runs at temperature 0.7) showed the model agreeing with itself 99.7% of the time on average, with 99% of pairs giving identical answers across all 5 runs. Using majority voting across these runs only improved F1 by 0.005, at 5 times the inference cost. These results suggest that the value of using an LLM over traditional methods depends heavily on the task. LLMs offer little advantage when strong lexical signals already exist, but a clear advantage when the task requires background knowledge, all while remaining highly consistent across repeated queries.
Figures & tables
Fig. 1: Example prompts and responses for the two prompting strategies tested. Panel (a) shows a correctly resolved match under zero-shot prompting; panel (b) shows a true match incorrectly rejected under few-shot SKU-guidance prompting.
Method
Precision
Recall
F1
Accuracy
Rule-based (Jaccard, threshold=0.2)
0.994
0.910
0.950
0.952
LLM zero-shot
0.999
0.902
0.948
0.951
LLM few-shot + SKU guidance
0.994
0.846
0.914
0.920
TABLE I: Precision, recall, F1, and accuracy for the three entity matching methods evaluated on the Abt-Buy benchmark (2,194 labeled pairs).
Fig. 2: Comparison of precision, recall, F1, and accuracy across the three entity matching methods on the Abt-Buy benchmark.
Method
Precision
Recall
F1
Accuracy
Rule-based (manufacturer-title)
0.564
1.000
0.721
0.622
LLM zero-shot
0.827
0.840
0.833
0.836
TABLE II: Precision, recall, F1, and accuracy for the two brand mislabeling detection methods evaluated on 500 Amazon product listings.
Fig. 4: Consistency of LLM predictions across five repeated runs at temperature 0.7 on a 200-pair sample (left), and F1 score for a single deterministic run versus a five-run majority vote on the same sample (right).
Large language models (LLMs) are increasingly used in annotation tasks with varying success. While several studies have demonstrated the opportunities and shortcomings of LLMs for annotation, a critical question remains unanswered: is it possible to predict how well they will perform on a new dataset? To answer this and related questions, we construct two task families which induce 461 text classification datasets. We find that it is possible to predict LLM performance - to a certain degree - and that contrary to other work which highlights the usefulness of LLM verbalizations of self-confidence, internal log-probabilities are the most predictive signal.
Large language models (LLMs) achieve strong entity matching performance without task-specific training data, but applying them to large sets of candidate pairs is slow and costly. Matchers built on pretrained language models (PLMs), such as BERT, offer faster inference but require training data. We systematically study knowledge-distillation workflows in which an LLM teacher labels training pairs for a smaller student matcher. We vary pair selection, labeling budget, teacher model, correspondence post-processing, and student model across eight benchmarks, including unseen entities and non-English data. We compare students trained on machine-labeled data with matchers trained on the original benchmark training sets. In most cases, PLM-based matchers trained on LLM-labeled data perform similarly to those trained on benchmark sets. Pair selection matters most for small labeling budgets, where active learning is often most effective. An open-weight teacher trains competitive students, so distillation requires no closed-weight models. Compact PLM-based students compete with much larger LLM students on most tasks while requiring 34 to 459 times less inference time than direct LLM matching. On the two benchmarks with high shares of unseen products, PLM-based students substantially underperform their teachers, as do students trained on benchmark data. Under GPT-5.2 pricing, LLM labeling costs per training set average $5.86 to $8.11. These findings support knowledge distillation as a practical approach to reduce the effort of labeling task-specific training data while enabling efficient inference.
Aaron Steiner, Christian Bizer
Data and Web Science Group University of Mannheim Mannheim, Germany
Reliable evaluation of large language models (LLMs) is critical as their deployment rapidly expands, particularly in high-stakes domains such as business and finance. The LLM-as-a-Judge framework, which uses prompted LLMs to evaluate response quality, is appealing due to its scalability, low cost, and strong correlations with human stylistic preferences. However, it remains unclear how accurately these methods can assess response quality in domains where correctness matters more than style. To address this gap, we introduce the Business and Finance Fundamentals Benchmark (BFF-Bench), a dataset of 160 challenging questions and long-form responses authored by financial professionals. These experts subsequently evaluated the correctness of 1,200 responses generated by a diverse set of LLMs on both BFF-Bench and a challenging subset of MT-Bench. With this expert-annotated dataset of judgments (VERDICTS), we analyze the agreement between a suite of automated grading methods and human experts. While we observe that LLM Judges are more reliable than other grading methods, our findings reveal a clear pattern in LLM Judge performance: when not provided with a correct reference, judges show high agreement with human experts only on questions the judges were able to correctly answer themselves. We demonstrate that providing the judges with expert-written references largely mitigates this issue, highlighting the limits of using LLM-as-a-Judge without any form of human verification.
Michael Krumdick, Charles Lovering, Varshini Reddy +2