Last Translation Benchmark

Hugging Face Daily Papers Papers

Summary

The Last Translation Benchmark introduces a live dataset of peer-reviewed, multimodal examples designed to evaluate and break leading machine translation models, with handcrafted verification rules for reliable assessment. It addresses the saturation of current benchmarks and the unreliability of automatic metrics.

For scientific progress, we need benchmarks that test the limits of state-of-the-art models, and evaluation methods that inform us about failure cases. As models get stronger, standard benchmarks for machine translation are approaching saturation. Further, automatic translation metrics are unreliable, vulnerable to reward-hacking, and provide unactionable assessments. Even gold human evaluation is not problem-free, because it often lacks reproducibility, objectivity, and scalability. Overall, this prevents us from tracking objective progress in the field and identifying pathways for improvement. We introduce the Last Translation Benchmark, a collection of human-authored and peer-reviewed examples (texts, images, audio, videos) that break leading machine translation models. We also present a new evaluation approach: each example comes with handcrafted verification rules describing concrete failure cases on that example, therefore allowing reliable and actionable future evaluation. The Last Translation Benchmark is a live dataset that accepts ongoing contributions. The latest version is LTBv1, containing accepted contributions prior to September 1st 2026, with future releases planned as new data is continuously collected.
Original Article
View Cached Full Text

Cached at: 09/04/26, 03:58 PM

Paper page - Last Translation Benchmark

Source: https://huggingface.co/papers/2609.04173 Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

The Last Translation Benchmark introduces peer-reviewed, multimodal examples that break leading translation models alongside handcrafted verification rules for reliable, actionable evaluation.

For scientific progress, we need benchmarks that test the limits of state-of-the-art models, and evaluation methods that inform us about failure cases. As models get stronger, standard benchmarks formachine translationare approaching saturation. Further,automatic translation metricsare unreliable, vulnerable toreward-hacking, and provide unactionable assessments. Even gold human evaluation is not problem-free, because it often lacks reproducibility, objectivity, and scalability. Overall, this prevents us from tracking objective progress in the field and identifying pathways for improvement. We introduce theLast Translation Benchmark, a collection of human-authored and peer-reviewed examples (texts, images, audio, videos) that break leadingmachine translationmodels. We also present a new evaluation approach: each example comes with handcraftedverification rulesdescribing concrete failure cases on that example, therefore allowing reliable and actionable future evaluation. TheLast Translation Benchmarkis a live dataset that accepts ongoing contributions. The latest version is LTBv1, containing accepted contributions prior to September 1st 2026, with future releases planned as new data is continuously collected.

View arXiv pageView PDFProject pageGitHub45Add to collection

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.04173 in a model README.md to link it from this page.

Datasets citing this paper1

#### zouhar/last-translation-benchmark Viewer• Updatedabout 11 hours ago • 3.46k • 33 • 17

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.04173 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

OpenSTBench: Beyond Semantic Evaluation for Speech Translation

Hugging Face Daily Papers

OpenSTBench is a unified multidimensional evaluation framework for speech translation systems that jointly assesses translation quality, speech quality, speaker preservation, emotion fidelity, and latency across both S2TT and S2ST systems in offline and streaming settings. The framework addresses the gap left by fragmented evaluation protocols and provides a reproducible benchmark for comparing heterogeneous speech translation systems.

Benchmarking Speech-to-Speech Translation Models

arXiv cs.CL

COMPASS is a unified benchmarking framework for speech-to-speech translation (S2ST) that integrates 46 metrics across eight dimensions, evaluated on 1,248 model-language configurations. It identifies complementary architecture strengths and proposes reduced metric subsets that preserve rankings while cutting evaluation time.