多语言量化代价：边缘端小型语言模型（SLM）中的结构崩溃与类型学脆弱性
The Multilingual Quantization Tax: Structural Collapse and Typological Fragility in Edge SLMs
浏览论文内容
中文总结 AI 辅助
本研究针对Gemma 4、Qwen 3.5架构开展4比特量化的零样本多语言评估，发现量化存在类型学脆弱性等四类现象，揭示了量化代价的多语言差异。
中文摘要 AI 辅助
尽管4比特权重量化对于将小型语言模型（SLM）部署到边缘设备至关重要，但对由此产生的性能下降（即量化代价）的评估仍以英语为中心。我们针对Gemma 4和Qwen 3.5架构，开展了4比特量化的零样本多语言评估。使用MMLU ProX Lite和GlobalPIQA在8种类型学差异显著的语言上进行评估，结果表明参数截断暴露了预训练中的深层不平等。我们识别出四种现象：（1）类型学脆弱性：低资源语言和特定非拉丁文字会因架构特有的双重解离而出现表征崩溃，无法生成有效的任务对数；（2）母语脆弱性悖论：基础预训练路径提供的精度损失保护有限；（3）领域特定遗忘：多步骤跨语言路由能力下降，而关联软科学记忆仍保持稳健；（4）量化抗性：高度饱和、类型学匹配的领域可抵抗确定性退化，量化后性能提升受限于统计噪声。
英文摘要
While 4-bit weight quantization is critical for deploying Small Language Models (SLMs) on edge devices, evaluations of the resulting performance degradation-the quantization tax-remain overwhelmingly English-centric. We present a zero-shot multilingual evaluation of 4-bit quantization across the Gemma 4 and Qwen 3.5 architectures. Evaluating on eight typo-logically diverse languages using MMLU ProX Lite and GlobalPIQA, we show parameter truncation exposes deep pre-training inequalities. We identify four phenomena: (1) Typological Fragility: low-resource and specific non-Latin scripts suffer representational collapse via architecture-specific double dissociations, failing to generate valid task logits; (2) Home Language Fragility Paradox: foundational pre-training pathways provide limited precision loss protection; (3) Domain-Specific Forgetting: multi-step cross-lingual routing degrades while associative soft-science recall remains robust; and (4) Quantization Resistance: highly saturated, typologically aligned domains resist deterministic degradation, with post-quantization performance gains bounded by statistical noise.
