cs.CLNov 18, 2026

Compression Effects on Low-Resource Multilingual Translation

Authors: Abbas Sani

Organizations: Independent Researcher / Arewa Data Science / Aboki AI, Kano, Nigeria

Abstract

We present a pilot study characterising the effect of post-training quantisation on multilingual machine translation quality and inference cost for three African languages - Hausa, Swahili, and Yoruba - across three model families: a small general-purpose LLM (Qwen2.5-1.5B), a larger multilingual instruction-tuned LLM (Aya-23-8B), and a dedicated multilingual MT model (NLLB-200-distilled-1.3B). Evaluating on the FLORES-200 devtest set (1,012 sentences per direction) under FP16, bitsandbytes INT8, and NF4 INT4 precisions on consumer-grade NVIDIA T4 GPUs, we find: (i) the specialised MT model outperforms both LLMs by 32.7-43.3 chrF on the Hausa and Swahili directions, and by 12.9-24.7 chrF on Yoruba, despite being roughly a sixth the size of the 8B instruction-tuned LLM; (ii) post-training quantisation preserves aggregate translation quality within approximately 1 chrF across all model families and language directions, with a single worst case of 2.19 chrF; (iii) on consumer GPUs, INT8 and INT4 quantisation provide substantial memory savings (49-61% at INT4) but do not reduce inference latency relative to FP16 - in fact, latency increases by 1.3x to 5.3x; (iv) a clear performance gradient tracking training-resource availability is visible across languages (Swahili > Hausa > Yoruba). These findings motivate a research agenda focused on hardware-aware compression and error-type analysis beyond aggregate metrics for low-resource deployment.

Figures & tables

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

Jul 31, 2026

Studying quantization trade-offs for efficient inference deployment in machine translation

Deploying large language models in realistic server environments poses challenges, as the system needs to provide high-quality responses with low latency. Quantization is a common approach to reduce the memory footprint and improve inference efficiency, yet its impact on latency and throughput is rarely evaluated under controlled, orchestration-level workloads. In this work we study the quantization trade-offs of two translation model families, EuroLLM \citep{martins2025eurollm} and Hy-MT2 \citep{zheng2026hy} across five models ranging from 1.7B to 22B for efficient deployment on a single A100 or H100 GPU. We demonstrate that combining a document-chunking strategy with W4A8 or W8A8 quantization improves the latency-throughput Pareto-curve under a wide range of workloads. Furthermore, since standard machine translation (MT) benchmarks rely on isolated sentences and fail to capture long-context dynamics, we introduce a document-level evaluation from WMT24++ to assess how text chunking strategies affect translation quality under quantization. Our results reveal that standard segment-level evaluation can fail to predict the interaction between quantization and long-context document translation. While Hy-MT2 remains robust under quantization, EuroLLM shows strong sensitivity and translation quality collapses rapidly for all considered quantization formats. Overall, our experiments show that the trade-off between inference efficiency and translation quality depends not only on the quantization format, but also on the choice of text chunking strategy.
Oct 13, 2026

Lightweight but Linguistically Fair? Quantisation Robustness across Bavarian, German, and English

Post-training quantisation reduces the memory cost of large language models, but it remains unclear whether compression preserves language skills of regional varieties as reliably as their standard-language counterparts. We study this question for Bavarian, Standard German, and English using BAVGROUND, a multiple-choice benchmark, where 206 source questions are aligned across all three languages. Across 14 open-weight 7B--10B instruction models, we compare BF16 with 8-bit GPTQ and two 4-bit settings, GPTQ and AWQ. Eight-bit GPTQ remains close to BF16 and largely preserves the relative behaviour of the three languages, while 4-bit quantisation was found to be much less stable. The two 4-bit methods reduce average accuracy by about 1.4--1.5 percentage points and change 11.9--16.0% of BF16 decisions, with changes concentrated on examples where the BF16 model is already uncertain. This instability does not translate into a consistent Bavarian disadvantage. Although Bavarian, a low-resource dialect, has lower baseline accuracy than German and English, the language most affected at 4 bits varies across models and quantisation methods. Calibration further shapes these cross-language effects: the relative Bavarian--German effect changes with calibration language, while changing the calibration source has little average effect on GPTQ-4 but can affect AWQ-4 much more strongly. Free-form evaluation on five models further shows that 4-bit compression can degrade generated responses even in settings where multiple-choice accuracy changes only modestly, while 8-bit GPTQ remains comparatively stable. Overall, our results show that quantisation robustness for regional language varieties depends on the model, quantisation method, calibration data, and evaluation setting, and cannot be measured by aggregate multilingual accuracy alone.
Oct 30, 2026

Compression Technique Interactions in Small Language Models: A Factorial Ablation Study

Deploying small language models on resource-constrained hardware requires compression, yet practitioners lack systematic guidance on how quantization, pruning, and post-compression recovery interact when combined. We present a factorial ablation study evaluating 72 compression configurations across four SLM architectures (2.7B-7B parameters), extended with a 48-configuration ordering analysis, benchmark accuracy validation on 30 compressed models, and robustness checks including multi-seed validation and full-corpus perplexity evaluation. INT8 quantization is effectively lossless (<0.5% accuracy degradation), while NF4 quantization combined with L1 magnitude pruning causes model collapse in all 16 tested combinations. LoRA-based fine-tuning reduces degraded perplexity by up to 97% after aggressive pruning; an ablation against standard fine-tuning without a teacher model shows the recovery is driven by LoRA adaptation itself rather than the distillation signal. Compression ordering shows no significant effect among viable configurations (ANOVA p = 0.96) but determines viability for NF4+pruning combinations. INT8 quantization alone provides 40-45% memory reduction with no quality loss. On-device CPU deployment via llama.cpp confirms these findings transfer off the GPU: 4-bit quantization runs 46-54% faster than 8-bit for a perplexity difference under 0.5 points, while un-structured pruning yields no CPU latency or memory benefit, consistent with its dense storage format.