mlsanityData to 7 Oct 2026

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

arXiv 2609.14839 ↗First posted 13 Sep 2026cs.SE cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

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.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 14 of its 20 references that are in the map. First posted on arXiv 13 Sep 2026.

Builds on

14 of 20 references.

References · 14

  1. 2025AlphaEvolve: A coding agent for scientific and algorithmic discoverycited by 763
  2. 2025Why Language Models Hallucinatecited by 205
  3. 2023Learning Performance-Improving Code Editscited by 103
  4. 2024LLM Hallucinations in Practical Code Generation: Phenomena, Mechanism, and Mitigationcited by 92
  5. 2024EffiBench: Benchmarking the Efficiency of Automatically Generated Codecited by 62
  6. 2024LLaMoCo: Instruction Tuning of Large Language Models for Optimization Code Generationcited by 45
  7. 2024How Efficient is LLM-Generated Code? A Rigorous & High-Standard Benchmarkcited by 39
  8. 2025SWE-Perf: Can Language Models Optimize Code Performance on Real-World Repositories?cited by 33
  9. 2024CodeMirage: Hallucinations in Code Generated by Large Language Modelscited by 23
  10. 2025Efficient Hallucination Detection for LLMs Using Uncertainty-Aware Attention Headscited by 14
  11. 2026Controlled Self-Evolution for Algorithmic Code Optimizationcited by 11
  12. 2023PerfRL: A Small Language Model Framework for Efficient Code Optimizationcited by 8
2 more
  1. 2025Evolving Excellence: Automated Optimization of LLM-based Agentscited by 4
  2. 2025How Do Agents Perform Code Optimization? An Empirical Studycited by 4

Built on by

Not cited yet.