verbosity-bias

Here are 3 public repositories matching this topic...

shahoismael / prompt-quality-vs-response-compliance

Code, per-item results and figures for a study dissociating prompt quality from response compliance in automated prompt-engineering assessment. Four-agent MATLAB evaluator on a locally hosted Qwen 2.5-7B judge, over 498 prompts from IFEval, LMSYS-Chat-1M and WildChat.

matlab automated-assessment ai-in-education instruction-following equivalence-testing construct-validity prompt-engineering ollama qwen llm-evaluation educational-assessment llm-as-a-judge open-weights ifeval verbosity-bias

Updated Sep 19, 2026
MATLAB

tkarim45 / llm-judge-bias

Star

Measure position/verbosity/assertiveness bias in an LLM-as-judge (Claude) by judging pairs in both orders. Finding: no position bias, but 75% verbosity bias and 100% assertiveness bias — judge scores gameable by length + tone.

evaluation bedrock bias claude llm-as-judge position-bias verbosity-bias

Updated Jul 25, 2026
Python

jo-sh-varughese / Toward-Interaction-Aware-Bias-Correction-in-LLM-as-a-Judge-Evaluation

Star

SSIT: a label-free, gold-free test for whether LLM-judge position x verbosity bias corrections actually compose. No human labels, no model of the judge. Code, 7-judge/6-family pilot data, and pre-registered protocol for the NeurIPS 2026 JUDGe workshop paper.

bias-correction difference-in-differences ssit llm-evaluation llm-as-judge neurips-2026 position-bias verbosity-bias

Updated Sep 16, 2026
Python

Add this topic to your repo

To associate your repository with the verbosity-bias topic, visit your repo's landing page and select "manage topics."

Learn more