CodeInsights Code Efficiency

HELM scenario that asks a model to write C++ matching one student's runtime style, then compares wall-clock times on unit tests.

Also known as: CodeInsightsCodeEfficiencyScenario, codeinsights code efficiency

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categorycoding
SubcategoryHELM C++ runtime alignment against a CodeInsight student sample
Page statusunknown
Metriccomposite: functional_correctness, runtime_efficiency_ratio, efficiency_alignment_score
Directionhigher_is_better
Dataset size19
PublisherStanford CRFM (HELM scenario); CodeInsight dataset from VNU-HCM University of Technology CS

What it measures

codeinsights_code_efficiency is a Stanford HELM run spec on sampled CodeInsight logs. For each retained student it shows three earlier C++ submissions and a fourth Vietnamese problem, and asks the model to solve that problem in the same personal style, including the student's inefficiency when the examples are slow. Scoring compiles both the model body and the student's target response, then compares mean runtime on tests that pass. It is runtime imitation, not generic Big-O ranking, and not the CodeInsight paper's next-attempt predictor.

Task format

English HELM instruction plus a Vietnamese/English problem prompt; model returns a fenced C++ fragment for {{ STUDENT_ANSWER }}. Default temperature 0, max_tokens 4000, num_testcases=1. Efficiency metric uses 5 timed runs and a 10 s timeout.

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub