Efficiency Hallucination: Formalizing and Measuring Behavioral Calibration in LLM-Based Code Optimization

arXiv:2609.14839 · cs.SE, cs.AI · Submitted 2026-09-13 · Read on arXiv

cs.SE, cs.AI

Submitted: 2026-09-13

Updated: 2026-09-13

Code: https://github.com/sarah-wilsxn/efficiency-hallucinations

License: http://creativecommons.org/licenses/by/4.0/

The gist: The integration of Large Language Models (LLMs) into automated code optimization introduces a critical reliability risk we term the Efficiency Hallucination: an LLM's tendency to issue non-functional

Terminology

Abstract

The integration of Large Language Models (LLMs) into automated code optimization introduces a critical reliability risk we term the Efficiency Hallucination: an LLM's tendency to issue non-functional mutations with unsubstantiated performance claims on already-optimized code. This is driven by the Evaluation Trap, wherein binary benchmarks incentivize unnecessary modifications over safely abstaining. We present a validation framework using classification penalty methods, evaluated across 180 optimization runs on nine models (GPT, Claude, Gemini) using EffiBench. Under standard prompts, models exhibit a 100% over-edit rate on optimal code. Our guardrail raises correct abstention from 0% to to 44.4%, preserving a 100% edit rate on sub-optimal code with zero false abstentions. Calibration is uneven: GPT-5.4 Mini approaches near-perfect abstention, and simple code is recognized more reliably than complex code. Our framework offers a training-free mechanism to mitigate LLM overconfidence before deployment in production.

Sources

Related papers