mlsanityData to 13 Oct 2026

Language-Conditional Dequantization: Recovering What Quantization Steals from Non-English Languages

arXiv 2608.11786 ↗First posted 12 Aug 2026cs.CLcited by 0

Not on the map. Why?

Abstract (arXiv)

Aggressive quantization disproportionately harms multilingual capability: in the sub-4B INT3 GPTQ regime, we measure 2-4x larger perplexity degradation on non-English languages than on English. We propose Language-Conditional Dequantization (LCD), a post-hoc method that attaches per-language rank-2 LoRA corrections to the linear layers of an already-quantized model, adding 0.12% parameters per language and training in under 20 minutes on a single GPU. Across Qwen2.5-3B and Llama-3.2-3B, LCD recovers 70-83% of the perplexity gap for non-Latin script languages and 17-28% of the GlobalMMLU accuracy gap, outperforming a language-agnostic correction of equal capacity by 3-9 points on typologically distant languages and a data-free low-rank baseline (LQER) by an order of magnitude. We further identify a perplexity-accuracy disconnect and trace it to where quantization concentrates damage: early-depth errors (Llama) propagate downstream and resist local correction, while late-depth errors (Qwen) do not. A layer-restricted variant of LCD validates this mechanism directly.

Key facts. No citing papers in the map yet (data through 13 Oct 2026). Builds on 7 of its 17 references that are in the map. First posted on arXiv 12 Aug 2026.

Builds on

7 of 17 references.

References · 7

  1. 2024The Llama 3 Herd of Modelscited by 14,486
  2. 2022GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformerscited by 1,884
  3. 2024Global MMLU: Understanding and Addressing Cultural and Linguistic Biases in Multilingual Evaluationcited by 182
  4. 2024LQER: Low-Rank Quantization Error Reconstruction for LLMscited by 41
  5. 2024ResQ: Mixed-Precision Quantization of Large Language Models with Low-Rank Residualscited by 32
  6. 2024LRQ: Optimizing Post-Training Quantization for Large Language Models by Learning Low-Rank Weight-Scaling Matricescited by 11
  7. 2024RILQ: Rank-Insensitive LoRA-based Quantization Error Compensation for Boosting 2-bit Large Language Model Accuracycited by 9

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Calibrating Beyond English: Language Diversity for Better Quantized Multilingual LLM
  2. 2024Cherry on Top: Parameter Heterogeneity and Quantization in Large Language Models
  3. 2026The Multilingual Quantization Tax: Structural Collapse and Typological Fragility in Edge SLMs
  4. 2026The Structure of Quantization Damage in LLMs: Why the Next Bit Should Be Spent Globally
  5. 2024Exploring the Trade-Offs: Quantization Methods, Task Difficulty, and Model Size in Large Language Models From Edge to Giant
  6. 2026Perplexity Can Miss SAE Feature Damage Under Quantization
  7. 2024What Makes Quantization for Large Language Models Hard? An Empirical Study from the Lens of Perturbation
  8. 2026Target-Aware Calibration Data Selection for Preserving Uncertainty in Quantized Language Models