Skip to content
OpenTrain AIFor AI Companies

HFEPX · Eval paper review

Context-Aware Subtitle Translation on CPU: Deployment-Driven Training, Evaluation, and Latency Design

Tsz-To Wong, Shyang-En Weng

Published

Jul 10, 2026

Citations

0

Trust level

Moderate

Usefulness score

65/100 (Medium)

Extraction confidence

70% (Moderate)

Derived from extracted protocol signals and abstract evidence.

Rater population

Not reported

Signals refreshed

Oct 22, 2026

Open arXiv Find implementation Shortlist (0) RSS feed

Tags

Automatic Metrics Coding Multilingual Pairwise Preference Web Browsing

Should you rely on this paper?

This paper has useful evaluation signal, but protocol completeness is partial; pair it with related papers before deciding implementation strategy.

Use this for comparison and orientation, not as your only source.

Best use

Secondary protocol comparison source

Use if you need

A secondary eval reference to pair with stronger protocol papers.

What to verify

Validate the evaluation procedure and quality controls in the full paper before operational use.

Main weakness

No major weakness surfaced.

Human feedback signal
Detected
From extracted signals
Evaluation signal
Detected
Eval setup described
Usefulness for eval research
65/100
Moderate-confidence candidate

Useful as a secondary reference; validate protocol details against neighboring papers.

If you are doing eval pipeline work, start here

Human Eval Papers LLM-as-Judge Papers Pairwise Preference Papers Tool-Use Evaluation Search

Abstract

We study English-to-Traditional-Chinese (Taiwan) subtitle translation for a free browser extension, where the model must run on the viewer's CPU in about 500 ms per subtitle. We train two models built on the LMT-60 base models, Sotto-1.7B and the deployed Sotto-0.6B, with a task-specific vocabulary, continued pretraining, supervised fine-tuning, weight interpolation, logit distillation and on-policy preference optimization. Because reference translations of subtitles are unreliable, we evaluate by pairwise comparison against one fixed baseline, Google Translate, with each pair judged in both orders by a panel of two LLM judges. On 1,820 held-out subtitles, Sotto-1.7B has a win rate of 0.576 against the baseline and wins head-to-head against four open translation models, and is slightly below a 27B model. We split latency into prefill and decoding: idle-time prefill, which prefills the known part of the prompt while waiting for the next subtitle, cuts mean end-to-end latency by 31% on a CPU, and the task-specific vocabulary encodes the same Chinese text in 28% fewer tokens. The deployed model stays within 500 ms for every subtitle in our latency test. The code is available at https://github.com/aiden1020/localsubs .

What we could verify

These are the protocol signals we could actually recover from the available paper metadata. Use them to decide whether this paper is worth deeper reading.

Human Feedback Types

strong

Pairwise Preference

Directly usable for protocol triage.

"We study English-to-Traditional-Chinese (Taiwan) subtitle translation for a free browser extension, where the model must run on the viewer's CPU in about 500 ms per subtitle."

Evaluation Modes

strong

Automatic Metrics

Includes extracted eval setup.

"We study English-to-Traditional-Chinese (Taiwan) subtitle translation for a free browser extension, where the model must run on the viewer's CPU in about 500 ms per subtitle."

Quality Controls

missing

Not reported

No explicit QC controls found.

"We study English-to-Traditional-Chinese (Taiwan) subtitle translation for a free browser extension, where the model must run on the viewer's CPU in about 500 ms per subtitle."

Benchmarks / Datasets

missing

Not extracted

No benchmark anchors detected.

"We study English-to-Traditional-Chinese (Taiwan) subtitle translation for a free browser extension, where the model must run on the viewer's CPU in about 500 ms per subtitle."

Reported Metrics

strong

Win rate

Useful for evaluation criteria comparison.

"On 1,820 held-out subtitles, Sotto-1.7B has a win rate of 0.576 against the baseline and wins head-to-head against four open translation models, and is slightly below a 27B model."

Benchmarks and datasets

No benchmark or dataset names were extracted from the available abstract.

Reported metrics

win rate
Human feedback details
Uses human feedback
Yes
Feedback types
Pairwise Preference
Rater population
Not reported
Unit of annotation
Pairwise
Expertise required
Coding, Multilingual
Evaluation details
Evaluation modes
Automatic Metrics
Agentic eval
Web Browsing
Quality controls
Not reported
Evidence quality
Moderate
Use this page as
Secondary protocol comparison source

Research brief

Metadata summary

We study English-to-Traditional-Chinese (Taiwan) subtitle translation for a free browser extension, where the model must run on the viewer's CPU in about 500 ms per subtitle.

Based on abstract + metadata only. Check the source paper before making high-confidence protocol decisions.

Key takeaways

  • We study English-to-Traditional-Chinese (Taiwan) subtitle translation for a free browser extension, where the model must run on the viewer's CPU in about 500 ms per subtitle.
  • We train two models built on the LMT-60 base models, Sotto-1.7B and the deployed Sotto-0.6B, with a task-specific vocabulary, continued pretraining, supervised fine-tuning, weight interpolation, logit distillation and on-policy preference optimization.
  • Because reference translations of subtitles are unreliable, we evaluate by pairwise comparison against one fixed baseline, Google Translate, with each pair judged in both orders by a panel of two LLM judges.

Researcher actions

  • Compare this paper against nearby papers in the same arXiv category before using it for protocol decisions.
  • Check the full text for explicit evaluation design choices (raters, protocol, and metrics).
  • Use related-paper links to find stronger protocol-specific references.

Caveats

  • Generated from abstract + metadata only; no PDF parsing.
  • Signals below are heuristic and may miss details reported outside the abstract.

Recommended queries

Pairwise preference evaluation Win rate

Contribution summary

  • We train two models built on the LMT-60 base models, Sotto-1.7B and the deployed Sotto-0.6B, with a task-specific vocabulary, continued pretraining, supervised fine-tuning, weight interpolation, logit distillation and on-policy preference…
  • Because reference translations of subtitles are unreliable, we evaluate by pairwise comparison against one fixed baseline, Google Translate, with each pair judged in both orders by a panel of two LLM judges.
  • On 1,820 held-out subtitles, Sotto-1.7B has a win rate of 0.576 against the baseline and wins head-to-head against four open translation models, and is slightly below a 27B model.

Why it matters for eval

  • We train two models built on the LMT-60 base models, Sotto-1.7B and the deployed Sotto-0.6B, with a task-specific vocabulary, continued pretraining, supervised fine-tuning, weight interpolation, logit distillation and on-policy preference…
  • Because reference translations of subtitles are unreliable, we evaluate by pairwise comparison against one fixed baseline, Google Translate, with each pair judged in both orders by a panel of two LLM judges.

Researcher checklist

  • Human feedback protocol is explicit

    Detected: Pairwise Preference

  • Evaluation mode is explicit

    Detected: Automatic Metrics

  • Quality control reporting appears

    No calibration/adjudication/IAA control explicitly detected.

  • Benchmark or dataset anchors are present

    No benchmark/dataset anchor extracted from abstract.

  • Metric reporting is present

    Detected: win rate