Comparing Developer and LLM Biases in Code Evaluation

• URL: http://arxiv.org/abs/2603.24586v1
• Date: Wed, 25 Mar 2026 17:56:55 GMT
• Title: Comparing Developer and LLM Biases in Code Evaluation
• Abstract summary: We present TRACE, a framework that evaluates judges' ability to predict human preferences.<n>Among 13 different models, the best judges underperform human annotators by 12-23%.<n>We find significant misalignment on the majority of existing code quality dimensions.
• Score: 30.695314123307764
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: As LLMs are increasingly used as judges in code applications, they should be evaluated in realistic interactive settings that capture partial context and ambiguous intent. We present TRACE (Tool for Rubric Analysis in Code Evaluation), a framework that evaluates LLM judges' ability to predict human preferences and automatically extracts rubric items to reveal systematic biases in how humans and models weigh each item. Across three modalities -- chat-based programming, IDE autocompletion, and instructed code editing -- we use TRACE to measure how well LLM judges align with developer preferences. Among 13 different models, the best judges underperform human annotators by 12-23%. TRACE identifies 35 significant sources of misalignment between humans and judges across interaction modalities, the majority of which correspond to existing software engineering code quality criteria. For example, in chat-based coding, judges are biased towards longer code explanations while humans prefer shorter ones. We find significant misalignment on the majority of existing code quality dimensions, showing alignment gaps between LLM judges and human preference in realistic coding applications.

Related papers

• Bias in the Loop: Auditing LLM-as-a-Judge for Software Engineering [0.0]
  Large Language Models are increasingly used as judges to evaluate code artifacts.<n>Current practice lacks a principled account of reliability and bias.<n>This paper studies LLM-as-a-Judge for code through a measurement-first lens.
  arXiv  Detail & Related papers  (2026-04-18T02:35:05Z)
• Human-Aligned MLLM Judges for Fine-Grained Image Editing Evaluation: A Benchmark, Framework, and Analysis [95.89328387635176]
  We introduce a fine-grained Multimodal Large Language Model (MLLM)-as-a-Judge framework for image editing.<n>We present a new human-validated benchmark that integrates human judgments, MLLM-based evaluations, model outputs, and traditional metrics.
  arXiv  Detail & Related papers  (2026-02-13T15:34:32Z)
• JudgeBoard: Benchmarking and Enhancing Small Language Models for Reasoning Evaluation [13.831735556002426]
  Small language models (SLMs) have shown promise on various reasoning tasks.<n>Their ability to judge the correctness of answers remains unclear compared to large language models (LLMs)
  arXiv  Detail & Related papers  (2025-11-20T01:14:39Z)
• BigCodeArena: Unveiling More Reliable Human Preferences in Code Generation via Execution [68.95247403447051]
  We introduce BigCodeArena, an open human evaluation platform for code generation backed by a comprehensive and on-the-fly execution environment.<n>We collected over 14,000 raw code-centric conversation sessions across 10 widely used LLMs, spanning 10 languages and 8 types of execution environments.<n>For BigCodeReward, we post-processed the 4,700 conversations and evaluated the consistency between reward models and human preferences.
  arXiv  Detail & Related papers  (2025-10-09T18:01:47Z)
• Quantitative LLM Judges [60.773734899532336]
  We propose quantitative LLM judges, which align evaluation scores of existing LLM judges to humans in a given domain.<n>The models are trained to improve the score of the original judge using its rationale and score.<n>Our experiments show that quantitative judges can improve the predictive power of existing judges through post-hoc modeling.
  arXiv  Detail & Related papers  (2025-06-03T14:44:23Z)
• An LLM-as-Judge Metric for Bridging the Gap with Human Evaluation in SE Tasks [15.820416019287622]
  SE-Jury is the first evaluation metric for LLM-as-Ensemble-Judge.<n>We evaluate SE-Jury across a diverse set of software engineering (SE) benchmarks.
  arXiv  Detail & Related papers  (2025-05-27T08:04:34Z)
• Can LLMs Replace Human Evaluators? An Empirical Study of LLM-as-a-Judge in Software Engineering [18.766132076075365]
  Large language models (LLMs) have been deployed to tackle various software engineering (SE) tasks like code generation.<n>Pass@k metric requires extensive unit tests and configured environments, and is not suitable for evaluating LLM-generated text.<n>Conventional metrics like BLEU, which measure only lexical rather than semantic similarity, have also come under scrutiny.
  arXiv  Detail & Related papers  (2025-02-10T06:49:29Z)
• Can Large Language Models Serve as Evaluators for Code Summarization? [47.21347974031545]
  Large Language Models (LLMs) serve as effective evaluators for code summarization methods.<n>LLMs prompt an agent to play diverse roles, such as code reviewer, code author, code editor, and system analyst.<n> CODERPE achieves an 81.59% Spearman correlation with human evaluations, outperforming the existing BERTScore metric by 17.27%.
  arXiv  Detail & Related papers  (2024-12-02T09:56:18Z)
• JudgeBench: A Benchmark for Evaluating LLM-based Judges [61.048125269475854]
  JudgeBench is a benchmark for evaluating LLM-based judges on challenging response pairs spanning knowledge, reasoning, math, and coding.<n>Our comprehensive evaluation on a collection of prompted judges, fine-tuned judges, multi-agent judges, and reward models shows that JudgeBench poses a significantly greater challenge than previous benchmarks.
  arXiv  Detail & Related papers  (2024-10-16T17:58:19Z)
• Calibrating LLM-Based Evaluator [92.17397504834825]
  We propose AutoCalibrate, a multi-stage, gradient-free approach to calibrate and align an LLM-based evaluator toward human preference. Instead of explicitly modeling human preferences, we first implicitly encompass them within a set of human labels. Our experiments on multiple text quality evaluation datasets illustrate a significant improvement in correlation with expert evaluation through calibration.
  arXiv  Detail & Related papers  (2023-09-23T08:46:11Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.