mlsanityData to 14 Oct 2026

Workload-Driven Optimization for On-Device Real-Time Subtitle Translation

arXiv 2607.09957 ↗First posted 10 Jul 2026cs.CL cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

This report studies on-device English-to-Traditional-Chinese subtitle translation for Taiwan under short inputs, short outputs, batch-size-one inference, low latency, and privacy constraints. These conditions limit the value of optimizations designed for long-context or high-throughput language-model serving. Starting from LMT-60-0.6B, preliminary profiling suggests that vocabulary projection becomes a more important decode-time cost after GGUF quantization reduces the relative cost of Transformer blocks. We replace the original 151k-token vocabulary with a 64k-token subtitle-domain tokenizer, migrate the embedding space, and adapt the model through embedding calibration followed by full supervised fine-tuning. On an OpenSubtitles2024 test set, LocalSubs achieves a 59.2% tie-excluded win rate against Google Translate under GPT-4o pairwise judging. Performance is strongest on short cues and declines as cue length increases. In a separate preliminary Apple M2 Metal profiling run, LocalSubs shows a 1.63x speedup over a 151k-vocabulary baseline. The code is available on https://github.com/aiden1020/localsubs .

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 8 of its 19 references that are in the map. First posted on arXiv 10 Jul 2026.

Builds on

8 of 19 references.

References · 8

  1. 2025Qwen3 Technical Reportcited by 8,820
  2. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  3. 2015Neural Machine Translation of Rare Words with Subword Unitscited by 2,918
  4. 2021Better than Average: Paired Evaluation of NLP Systemscited by 30
  5. 2024An Analysis of BPE Vocabulary Trimming in Neural Machine Translationcited by 7
  6. 2026Mediocrity is the key for LLM as a Judge Anchor Selectioncited by 7
  7. 2025NiuTrans.LMT: Toward Inclusive and Scalable Multilingual Machine Translation with LLMscited by 4
  8. 2024A Case Study on Contextual Machine Translation in a Professional Scenario of Subtitlingcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026From Utterance to Vividity: Training Expressive Subtitle Translation LLM via Adaptive Local Preference Optimization
  2. 2025The Uneven Impact of Post-Training Quantization in Machine Translation
  3. 2026Studying quantization trade-offs for efficient inference deployment in machine translation
  4. 2024Evaluating the Generalization Ability of Quantized LLMs: Benchmark, Analysis, and Toolbox
  5. 2025Efficient Speech Translation through Model Compression and Knowledge Distillation
  6. 2026Optimizing Korean-Centric LLMs via Token Pruning
  7. 2024On the Compressibility of Quantized Large Language Models
  8. 2025FuxiMT: Sparsifying Large Language Models for Chinese-Centric Multilingual Machine Translation