arXiv subjects

Shenbin Qian

Publications and source records attributed to Shenbin Qian.

17 recordsLinked to original sources

arXiv2026

TransClean: A Benchmark for Detecting and Extracting Clean Translations from Large Language Model Outputs

Large language models (LLMs) are increasingly used for machine translation, yet their outputs often contain additional text beyond the translation itself, such as language labels, explanations or bilingual repetitions, which we term translation noise. Despite its prevalence, this problem lacks dedicated benchmarks and systematic study. We analyze over 790,000 translation outputs from 12 LLMs across 22 language pairs (LPs) and identify 12 recurring noise patterns, which we group into formatting and content noise. Building on the observed patterns, we construct TransClean, a controlled benchmark of 9,900 pairs of noisy and clean translation outputs, comprising 8,800 synthetically generated instances and 1,100 manually curated authentic instances. We evaluate two extraction approaches on the TransClean benchmark: 1) a span-based extraction method leveraging translation quality estimation models for span detection, and 2) an LLM-based extraction method that prompts an LLM to isolate the translation. Our benchmark and analysis provide the first systematic framework to evaluate and improve the cleanliness of LLM translation outputs.

cs.CL↗

arXiv2026

Last Translation Benchmark

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.

cs.CL↗

arXiv2026

IndicQE-APE: A Benchmark for Quality Estimation and Automatic Post-Editing for Indic Languages

Indic quality estimation (QE) and automatic post-editing (APE) data is spread across separate releases, so no single resource supports training and evaluation across tasks and language pairs on one footing. We consolidate the WMT 2020--2024 shared-task lineage with an extended English--Malayalam resource into \indicqe: $126{,}754$ instances over nine directional pairs, with up to four label types aligned on the same segment, a direct assessment, a human post-edit, word-level OK/BAD tags and an error explanation, and a test set stratified over four difficulty axes. On it we benchmark six prompted LLMs and three COMET metrics on segment-level QE, and three systems on APE. Two of the axes are defined partly on the direct assessment and select a compressed slice of it, so each axis is compared against a control drawn from the same language pair with the same score distribution. Only one survives that control: segments whose holistic and token-level quality signals conflict are ranked worse than equally-scored segments of the same language, for all nine systems and all seven pairs that carry the axis. Annotator disagreement, which looks second-hardest without the control, has no effect with it. Few-shot prompting costs every model $\leq$ $3.4$B both correlation and output-format compliance. Within-language accuracy does not make scores comparable across pairs: of the three trained metrics, the one with the best within-language correlation loses most when the pairs are pooled. The benchmark (https://huggingface.co/datasets/surrey-nlp/IndicQE-APE) and code (https://github.com/surrey-nlp/IndicQE-APE) are released.

cs.CL↗

arXiv2026

MultiSynt/MT: Trillion-Token Multi-Parallel Pre-Training Data Translated Across 36 Languages

Open web-scale pre-training corpora remain concentrated in English, limiting multilingual LLM development. We introduce MultiSynt/MT, an open synthetic parallel corpus with approximately 4.8 trillion target-language tokens across 36 languages, produced by translating 100 billion high-quality Nemotron-CC tokens with Tower+ and OPUS-MT/HPLT-MT systems. For many medium- and lower-resource European languages, this is the largest openly available pre-training resource. Across five high- and medium-resource languages, reference LLMs trained on MultiSynt/MT reach the final score of HPLT 2.0, a native-data baseline, using roughly 72% fewer pre-training tokens, and outperform it by approximately 15% relative at a matched 100B-token training budget. Our analyses also identify evaluation blind spots: standard multiple-choice benchmarks miss translation-quality differences that a fluency-sensitive LLM-as-judge protocol recovers on the trained LLMs without detecting a deficit relative to its native-data baseline, while Norwegian idiomatic and culturally grounded tasks remain better served by native data. We release the corpus, including row-aligned translations from multiple systems, to support controlled research on multilingual pre-training data and evaluation.

cs.CL↗

arXiv2026

Why do Large Language Models Fail in Low-resource Translation? Unraveling the Token Dynamics of Large Language Models for Machine Translation

Large Language Models (LLMs) have recently demonstrated strong performance in machine translation (MT). However, most prior work focuses on improving or benchmarking translation quality, offering limited insight into when and why LLM-based translation fails. In this work, we systematically analyze failure modes of LLMs in MT by evaluating 15 models, including four reasoning LLMs, across 22 language pairs (LPs) with varying resource levels. We find that non-English-centric LPs consistently yield lower COMET scores than English-centric pairs. To investigate the underlying causes, we introduce Token Activation Rate (TAR), a metric that captures how effectively a model utilizes language-specific tokens in its vocabulary during generation. We validate TAR as a proxy for language representation using models with known language distributions in the training data, and show that lower TAR is strongly associated with poorer translation performance. Furthermore, reasoning LLMs tend to generate more tokens when translating into low-TAR languages, suggesting a compensatory mechanism, although its impact on translation quality varies across models. Overall, our findings emphasize the importance of token-level dynamics in understanding MT performance of LLMs.

cs.CL↗

arXiv2026

Beyond Scalar Scores: Reinforcement Learning for Error-Aware Quality Estimation of Machine Translation

Quality Estimation (QE) aims to assess the quality of machine translation (MT) outputs without relying on reference translations, making it essential for real-world, large-scale MT evaluation. Large Language Models (LLMs) have shown significant promise in advancing the field of quality estimation of machine translation. However, most of the QE approaches solely rely on scalar quality scores, offering no explicit information about the translation errors that should drive these judgments. Moreover, for low-resource languages where annotated QE data is limited, existing approaches struggle to achieve reliable performance. To address these challenges, we introduce the first segment-level QE dataset for English to Malayalam, a severely resource-scarce language pair in the QE domain, comprising human-annotated Direct Assessment (DA) scores and Translation Quality Remarks (TQR), which are short, contextual, free-form annotator comments that describe translation errors. We further introduce ALOPE-RL, a policy-based reinforcement learning framework that trains efficient adapters based on policy rewards derived from DA score and TQR. Integrating error-aware rewards with ALOPE-RL, enables LLMs to reason about translation quality beyond numeric scores. Despite being trained on a small-scale QE dataset, ALOPE-RL achieves state-of-the-art performance on English to Malayalam QE using compact LLMs (<=4B parameters}) fine-tuned with LoRA and 4-bit quantization, outperforming both larger LLM-based baselines and leading encoder-based QE models. Our results demonstrate that error-aware, policy-based learning can deliver strong QE performance under limited data and compute budgets. We release our dataset, code, and trained models to support future research.

cs.CL↗

arXiv2025

Human-Centred Evaluation of Text-to-Image Generation Models for Self-expression of Mental Distress: A Dataset Based on GPT-4o

Effective communication is central to achieving positive healthcare outcomes in mental health contexts, yet international students often face linguistic and cultural barriers that hinder their communication of mental distress. In this study, we evaluate the effectiveness of AI-generated images in supporting self-expression of mental distress. To achieve this, twenty Chinese international students studying at UK universities were invited to describe their personal experiences of mental distress. These descriptions were elaborated using GPT-4o with four persona-based prompt templates rooted in contemporary counselling practice to generate corresponding images. Participants then evaluated the helpfulness of generated images in facilitating the expression of their feelings based on their original descriptions. The resulting dataset comprises 100 textual descriptions of mental distress, 400 generated images, and corresponding human evaluation scores. Findings indicate that prompt design substantially affects perceived helpfulness, with the illustrator persona achieving the highest ratings. This work introduces the first publicly available text-to-image evaluation dataset with human judgment scores in the mental health domain, offering valuable resources for image evaluation, reinforcement learning with human feedback, and multi-modal research on mental health communication.

q-bio.NC↗

arXiv2025

The Mind's Eye: A Multi-Faceted Reward Framework for Guiding Visual Metaphor Generation

Visual metaphor generation is a challenging task that aims to generate an image given an input text metaphor. Inherently, it needs language understanding to bind a source concept with a target concept, in a way that preserves meaning while ensuring visual coherence. We propose a self-evaluating visual metaphor generation framework that focuses on metaphor alignment. Our self-evaluation approach combines existing metrics with our newly proposed metaphor decomposition score and a meaning alignment (MA) metric. Within this setup, we explore two novel approaches: a training-free pipeline that explicitly decomposes prompts into source-target-meaning (S-T-M) mapping for image synthesis, and a complementary training-based pipeline that improves alignment using our proposed self-evaluation reward schema, without any large-scale retraining. On the held-out test set, the training-free approach surpasses strong closed baselines (GPT-4o, Imagen) on decomposition, CLIP, and MA scores, with the training-based approach close behind. We evaluate our framework output using a user-facing study, and observed that participants preferred GPT-4o overall, while our training-free pipeline led open-source methods and edged Imagen on abstract metaphors. Our analyses show S-T-M prompting helps longer or more abstract metaphors, with closed models excelling on short, concrete cases; we also observe sensitivity to sampler settings. Overall, structured prompting and lightweight RL perform metaphor alignment well under modest compute, and remaining gaps to human preference appear driven by aesthetics and sampling.

cs.CL↗

arXiv2025

ALOPE: Adaptive Layer Optimization for Translation Quality Estimation using Large Language Models

Large Language Models (LLMs) have shown remarkable performance across a wide range of natural language processing tasks. Quality Estimation (QE) for Machine Translation (MT), which assesses the quality of a source-target pair without relying on reference translations, remains a challenging cross-lingual task for LLMs. The challenges stem from the inherent limitations of existing LLM-based QE systems, which are pre-trained for causal language modelling rather than regression-specific tasks, further elevated by the presence of low-resource languages given pre-training data distribution. This paper introduces ALOPE, an adaptive layer-optimization framework designed to enhance LLM-based QE by restructuring Transformer representations through layer-wise adaptation for improved regression-based prediction. Our framework integrates low-rank adapters (LoRA) with regression task heads, leveraging selected pre-trained Transformer layers for improved cross-lingual alignment. In addition to the layer-specific adaptation, ALOPE introduces two strategies-dynamic weighting, which adaptively combines representations from multiple layers, and multi-head regression, which aggregates regression losses from multiple heads for QE. Our framework shows improvements over various existing LLM-based QE approaches. Empirical evidence suggests that intermediate Transformer layers in LLMs provide contextual representations that are more aligned with the cross-lingual nature of the QE task. We make resultant models and framework code publicly available for further research, also allowing existing LLM-based MT frameworks to be scaled with QE capabilities.

cs.CL↗

arXiv2025

NEAR$^2$: A Nested Embedding Approach to Efficient Product Retrieval and Ranking

E-commerce information retrieval (IR) systems struggle to simultaneously achieve high accuracy in interpreting complex user queries and maintain efficient processing of vast product catalogs. The dual challenge lies in precisely matching user intent with relevant products while managing the computational demands of real-time search across massive inventories. In this paper, we propose a Nested Embedding Approach to product Retrieval and Ranking, called NEAR$^2$, which can achieve up to $12$ times efficiency in embedding size at inference time while introducing no extra cost in training and improving performance in accuracy for various encoder-based Transformer models. We validate our approach using different loss functions for the retrieval and ranking task, including multiple negative ranking loss and online contrastive loss, on four different test sets with various IR challenges such as short and implicit queries. Our approach achieves an improved performance over a smaller embedding dimension, compared to any existing models.

cs.IR↗

arXiv2025

Automatically Generating Chinese Homophone Words to Probe Machine Translation Estimation Systems

Evaluating machine translation (MT) of user-generated content (UGC) involves unique challenges such as checking whether the nuance of emotions from the source are preserved in the target text. Recent studies have proposed emotion-related datasets, frameworks and models to automatically evaluate MT quality of Chinese UGC, without relying on reference translations. However, whether these models are robust to the challenge of preserving emotional nuances has been left largely unexplored. To address this gap, we introduce a novel method inspired by information theory which generates challenging Chinese homophone words related to emotions, by leveraging the concept of self-information. Our approach generates homophones that were observed to cause translation errors in emotion preservation, and exposes vulnerabilities in MT systems and their evaluation methods when tackling emotional UGC. We evaluate the efficacy of our method using human evaluation for the quality of these generated homophones, and compare it with an existing one, showing that our method achieves higher correlation with human judgments. The generated Chinese homophones, along with their manual translations, are utilized to generate perturbations and to probe the robustness of existing quality evaluation models, including models trained using multi-task learning, fine-tuned variants of multilingual language models, as well as large language models (LLMs). Our results indicate that LLMs with larger size exhibit higher stability and robustness to such perturbations. We release our data and code for reproducibility and further research.

cs.CL↗

arXiv2025

GCDance: Genre-Controlled Music-Driven 3D Full Body Dance Generation

Music-driven dance generation is a challenging task as it requires strict adherence to genre-specific choreography while ensuring physically realistic and precisely synchronized dance sequences with the music's beats and rhythm. Although significant progress has been made in music-conditioned dance generation, most existing methods struggle to convey specific stylistic attributes in generated dance. To bridge this gap, we propose a diffusion-based framework for genre-specific 3D full-body dance generation, conditioned on both music and descriptive text. To effectively incorporate genre information, we develop a text-based control mechanism that maps input prompts, either explicit genre labels or free-form descriptive text, into genre-specific control signals, enabling precise and controllable text-guided generation of genre-consistent dance motions. Furthermore, to enhance the alignment between music and textual conditions, we leverage the features of a music foundation model, facilitating coherent and semantically aligned dance synthesis. Last, to balance the objectives of extracting text-genre information and maintaining high-quality generation results, we propose a novel multi-task optimization strategy. This effectively balances competing factors such as physical realism, spatial accuracy, and text classification, significantly improving the overall quality of the generated sequences. Extensive experimental results obtained on the FineDance and AIST++ datasets demonstrate the superiority of GCDance over the existing state-of-the-art approaches.

cs.GR↗

arXiv2025

When LLMs Struggle: Reference-less Translation Evaluation for Low-resource Languages

This paper investigates the reference-less evaluation of machine translation for low-resource language pairs, known as quality estimation (QE). Segment-level QE is a challenging cross-lingual language understanding task that provides a quality score (0-100) to the translated output. We comprehensively evaluate large language models (LLMs) in zero/few-shot scenarios and perform instruction fine-tuning using a novel prompt based on annotation guidelines. Our results indicate that prompt-based approaches are outperformed by the encoder-based fine-tuned QE models. Our error analysis reveals tokenization issues, along with errors due to transliteration and named entities, and argues for refinement in LLM pre-training for cross-lingual tasks. We release the data, and models trained publicly for further research.

cs.CL↗

arXiv2024

Are Large Language Models State-of-the-art Quality Estimators for Machine Translation of User-generated Content?

This paper investigates whether large language models (LLMs) are state-of-the-art quality estimators for machine translation of user-generated content (UGC) that contains emotional expressions, without the use of reference translations. To achieve this, we employ an existing emotion-related dataset with human-annotated errors and calculate quality evaluation scores based on the Multi-dimensional Quality Metrics. We compare the accuracy of several LLMs with that of our fine-tuned baseline models, under in-context learning and parameter-efficient fine-tuning (PEFT) scenarios. We find that PEFT of LLMs leads to better performance in score prediction with human interpretable explanations than fine-tuned models. However, a manual analysis of LLM outputs reveals that they still have problems such as refusal to reply to a prompt and unstable output while evaluating machine translation of UGC.

cs.CL↗

arXiv2024

A Multi-task Learning Framework for Evaluating Machine Translation of Emotion-loaded User-generated Content

Machine translation (MT) of user-generated content (UGC) poses unique challenges, including handling slang, emotion, and literary devices like irony and sarcasm. Evaluating the quality of these translations is challenging as current metrics do not focus on these ubiquitous features of UGC. To address this issue, we utilize an existing emotion-related dataset that includes emotion labels and human-annotated translation errors based on Multi-dimensional Quality Metrics. We extend it with sentence-level evaluation scores and word-level labels, leading to a dataset suitable for sentence- and word-level translation evaluation and emotion classification, in a multi-task setting. We propose a new architecture to perform these tasks concurrently, with a novel combined loss function, which integrates different loss heuristics, like the Nash and Aligned losses. Our evaluation compares existing fine-tuning and multi-task learning approaches, assessing generalization with ablative experiments over multiple datasets. Our approach achieves state-of-the-art performance and we present a comprehensive analysis for MT evaluation of UGC.

cs.CL↗

arXiv2024

What do Large Language Models Need for Machine Translation Evaluation?

Leveraging large language models (LLMs) for various natural language processing tasks has led to superlative claims about their performance. For the evaluation of machine translation (MT), existing research shows that LLMs are able to achieve results comparable to fine-tuned multilingual pre-trained language models. In this paper, we explore what translation information, such as the source, reference, translation errors and annotation guidelines, is needed for LLMs to evaluate MT quality. In addition, we investigate prompting techniques such as zero-shot, Chain of Thought (CoT) and few-shot prompting for eight language pairs covering high-, medium- and low-resource languages, leveraging varying LLM variants. Our findings indicate the importance of reference translations for an LLM-based evaluation. While larger models do not necessarily fare better, they tend to benefit more from CoT prompting, than smaller models. We also observe that LLMs do not always provide a numerical score when generating evaluations, which poses a question on their reliability for the task. Our work presents a comprehensive analysis for resource-constrained and training-less LLM-based evaluation of machine translation. We release the accrued prompt templates, code and data publicly for reproducibility.

cs.CL↗

arXiv2023

Evaluation of Chinese-English Machine Translation of Emotion-Loaded Microblog Texts: A Human Annotated Dataset for the Quality Assessment of Emotion Translation

In this paper, we focus on how current Machine Translation (MT) tools perform on the translation of emotion-loaded texts by evaluating outputs from Google Translate according to a framework proposed in this paper. We propose this evaluation framework based on the Multidimensional Quality Metrics (MQM) and perform a detailed error analysis of the MT outputs. From our analysis, we observe that about 50% of the MT outputs fail to preserve the original emotion. After further analysis of the errors, we find that emotion carrying words and linguistic phenomena such as polysemous words, negation, abbreviation etc., are common causes for these translation errors.

cs.CL↗