← 研踪 PaperTrailarXiv:2605.19537 · v3
The Silent Hyperparameter: Quantifying the Impact of Inference Backends on LLM Reproducibility
首次提交 2026-05-19 16:37
版本更新 2026-11-06 19:03
雷达观测 2026-11-09 11:19
摘要
Progress in LLMs is increasingly measured through standardized benchmarks, where state-of-the-art improvements are often separated by fractions of a percentage point. At the same time, the computational cost of evaluating modern LLMs has driven widespread adoption of specialized inference backends, software systems that execute trained models efficiently at inference time. While critical for scalability, system-level optimizations, such as custom CUDA kernels and reduced-precision arithmetic, can alter token probabilities and introduce non-determinism, possibly cascading into divergent generation. In this work, we first survey the inference landscape, identifying $200$ distinct engines, and analyze $35,000$ ML publications, finding that the specific inference stack is rarely reported despite this widespread diversity. We then present a systematic empirical study of how inference backends affect LLM benchmark results. Holding model weights, decoding parameters, and hardware constant, we evaluate five widely used inference engines, including vLLM, SGLang, and llama.cpp, across multiple open-weight models spanning 4B-30B parameters and established benchmarks. We show that the choice of backend alone can shift benchmark scores by up to $16.6$ percentage points and induce high rates of output disagreement. By isolating backend optimizations and tracing the execution pipeline, we find this divergence is driven by system-level optimizations such as prefix caching and CUDA graphs, custom kernels, and engine-specific defaults in logit processing. Our findings identify the inference backend as a previously unreported yet consequential hyperparameter in LLM evaluations and advocate standardized reporting of inference stacks to improve the reproducibility and interpretability of benchmark comparisons.
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v3 · 2026-11-06 19:03 · 最新观测版本
The Silent Hyperparameter: Quantifying the Impact of Inference Backends on LLM Reproducibility
Progress in LLMs is increasingly measured through standardized benchmarks, where state-of-the-art improvements are often separated by fractions of a percentage point. At the same time, the computational cost of evaluating modern LLMs has driven widespread adoption of specialized inference backends, software systems that execute trained models efficiently at inference time. While critical for scalability, system-level optimizations, such as custom CUDA kernels and reduced-precision arithmetic, can alter token probabilities and introduce non-determinism, possibly cascading into divergent generation. In this work, we first survey the inference landscape, identifying $200$ distinct engines, and analyze $35,000$ ML publications, finding that the specific inference stack is rarely reported despite this widespread diversity. We then present a systematic empirical study of how inference backends affect LLM benchmark results. Holding model weights, decoding parameters, and hardware constant, we evaluate five widely used inference engines, including vLLM, SGLang, and llama.cpp, across multiple open-weight models spanning 4B-30B parameters and established benchmarks. We show that the choice of backend alone can shift benchmark scores by up to $16.6$ percentage points and induce high rates of output disagreement. By isolating backend optimizations and tracing the execution pipeline, we find this divergence is driven by system-level optimizations such as prefix caching and CUDA graphs, custom kernels, and engine-specific defaults in logit processing. Our findings identify the inference backend as a previously unreported yet consequential hyperparameter in LLM evaluations and advocate standardized reporting of inference stacks to improve the reproducibility and interpretability of benchmark comparisons.
查看此版本解读 →查看此版本原文 ↗