Machine translation based on large language models (LLMs) has achieved remarkable progress in recent years. However, relying on a single translation model and fixed translation instructions often limits translation quality, especially for multilingual and multi-domain scenarios. To address this issue, we propose HW-TSC-Agent, a multi-stage translation agent framework that integrates translation model fine-tuning, dual-model candidate generation, and instruction-aware translation refinement into a unified pipeline. First, we perform Supervised Fine-Tuning (SFT) and Contrastive Preference Optimization (CPO) on openPangu-Embedded-7B using high-quality bilingual data to enhance its multilingual translation capabilities. The fine-tuned openPangu-Embedded-7B and HY-MT2-7B are then used to generate candidate translations. Finally, source-language analysis, candidate translations, and task-specific instructions are incorporated into a dynamic prompt, which guides DeepSeek-V4-Flash to compare, refine, and fuse the candidate translations to produce the final output. Our system is submitted to the WMT2026 General Machine Translation Shared Task, covering 13 translation directions. On the WMT2025 English-to-Chinese development test set, enriching the dynamic prompt generally improves translation quality over an instruction-only DeepSeek-V4-Flash baseline.

We describe HW-TSC’s submissions to both tracks and all language pairs of the WMT26 Terminology Translation Shared Task. Our primary system uses Qwen3.7-Max for source-term extraction, bilingual term alignment, and document translation. In Track 1, it matches the current document or translation chunk against the supplied dictionary and places at most 200 applicable term pairs in the translation prompt. In Track 2, it first induces a bilingual glossary from the seed bitexts and then reuses the Track 1 translation module. Induction combines open source-term extraction with seed-source expression matching, followed by LLM alignment to the seed targets. We optimize the extraction prompt with SkillOpt on a train/validation/test split of the WMT25 English–Chinese finance data and use Qwen3.7-Max to rewrite it for directions without labeled optimization data. Extraction recall rises from 0.6250 to 0.6344 on validation and from 0.6451 to 0.6674 on the held-out test split. On the same held-out test documents with Qwen3.7-Max, glossary induction raises TSR from 0.5212 (no glossary) to 0.7910 with the initial prompt and to 0.8026 after SkillOpt, with chrF++ of 57.45, 67.05, and 67.36; the official glossary yields TSR 0.9019 and chrF++ 63.94.

This paper describes the HW-TSC’s submission to the IWSLT 2026 Offline Speech Translation Task, specifically for the English-to-Chinese and English-to-German unconstrained tracks. Our system adopts a robust cascade architecture optimized for long-form, unsegmented audio. To mitigate the hallucination and inconsistency issues common in long-sequence processing, we propose a two-pass transcription strategy: an initial streaming ASR with a 12-second context buffer for sentence-level coherence, followed by Qwen3-ForcedAligner for precise timestamping. Based on these alignments, a second-pass refinement is conducted using Qwen3-Omni on re-segmented 30-second chunks to ensure high-fidelity transcriptions. For the translation module, we employ a context-aware segment merging strategy (up to 150 tokens) to empower the Qwen3 llm with sufficient semantic context. Experimental results on the tst-2022 benchmark demonstrate the effectiveness of our pipeline, achieving COMET scores of 0.8462 (En-Zh) and 0.7854 (En-De), significantly outperforming the standard cascade baselines.

This paper presents HW-TSC’s submission to the IWSLT 2026 Cross-Lingual Voice Cloning Track. The Cross-Lingual Voice Cloning Track includes three target languages: Arabic, Chinese, and French. We take part in two language tasks of this track, namely Chinese and French. We employ the Qwen3-TTS-12Hz-1.7B-Base multilingual model as the core voice cloning model. To tackle problems such as excessively long duration of the original reference audio and scattered features, we design a sliding-window audio segmentation preprocessing method, which continuously splits long audio into standardized short segments with overlapping redundancy. This method avoids feature attenuation caused by overly long audio and maximizes the preservation of complete timbre information through step overlap. To select the outputs with the highest timbre similarity from numerous synthetic results, this study conducts voiceprint recognition based on the Enhanced Context-Dependent Adversarial Time Delay Neural Network (ECAPA-TDNN), with cosine similarity as the core quantitative evaluation metric, and selects the result with the highest similarity as the optimal output.

This paper introduces HW-TSC’s submission to the IWSLT 2026 Subtitling track. For automatic subtitle generation, we employ a cascaded strategy under unconstrained conditions. First, we construct a large-model-based streaming speech recognition framework, which incorporates VAD voice activity detection, sliding-window context caching, long audio chunking, and the Qwen3 forced alignment model to achieve high-precision transcription and timestamping from English speech to text. Next, we perform text translation using a Qwen3-based translation model. Finally, according to subtitle constraints such as characters per second (CPS) and characters per line (CPL), we identify translation segments that exceed compliance thresholds via quantitative evaluation, and rewrite them using a large language model while preserving core semantic meaning, ultimately producing subtitle files that meet the required standards.

Pricing automation in large-scale tourism is challenging because travel orders are highly unstructured, while pricing policies are complex, rapidly evolving, and inherently open-ended. Traditional rule engines are brittle and costly to maintain, whereas unconstrained LLM agents lack the reliability and auditability required for financial decisions. We present a production-grade LLM-powered pricing system with a strict decision boundary: LLMs perform structured extraction and bounded policy/path selection, while all numeric pricing, including total-price computation, is executed deterministically. Policies are compiled into interpretable condition trees, enabling open-ended support for new clauses and evolving rules without code changes, while exposing auditable artifacts for human-in-the-loop control. Periodic fine-tuning on logged traces further improves tree induction and path matching. Deployed at a municipal state-owned tourism enterprise across 7 scenic sites and 12 business categories with 1,500+ operators and 1,000+ active policies, the system processed 3,960 orders in six months, reduced the order management team from 15-20 to 3, and cut per-order handling time from 10 minutes to <2 minutes.

Large language models (LLMs) excel in natural language generation but also exhibit biases, particularly in gender, race, and religion, which can be amplified with widespread use. However, research on biases in specific domains, such as finance, remains limited. To address this gap, we conducted a comprehensive evaluation of 23 leading LLMs and found varying degrees of financial bias, including more pronounced biases in financial-specific LLMs (FinLLMs). In response, we propose the Financial Bias Indicators (FBI) framework, which includes components like the Bias Unveiler, Bias Detective, Bias Tracker, and Bias Antidote, designed to identify, detect, analyze, and mitigate financial biases. Our analysis explores the root causes of these biases and introduces a debiasing method based on financial causal knowledge, alongside three other debiasing techniques. For the most biased model, we successfully reduced bias by 68% according to key metrics. This study advances our understanding of LLM biases in finance and highlights the need for greater scrutiny in their application within this critical domain.

While current tasks of converting natural language to SQL (NL2SQL) using Foundation Models have shown impressive achievements, adapting these approaches for converting natural language to Graph Query Language (NL2GQL) encounters hurdles due to the distinct nature of GQL compared to SQL, alongside the diverse forms of GQL. Moving away from traditional rule-based and slot-filling methodologies, we introduce a novel approach, R³-NL2GQL, integrating both small and large Foundation Models for ranking, rewriting, and refining tasks. This method leverages the interpretative strengths of smaller models for initial ranking and rewriting stages, while capitalizing on the superior generalization and query generation prowess of larger models for the final transformation of natural language queries into GQL formats. Addressing the scarcity of datasets in this emerging field, we have developed a bilingual dataset, sourced from graph database manuals and selected open-source Knowledge Graphs (KGs). Our evaluation of this methodology on this dataset demonstrates its promising efficacy and robustness.

Desire is a strong wish to do or have something, which involves not only a linguistic expression, but also underlying cognitive phenomena driving human feelings. As the most primitive and basic human instinct, conscious desire is often accompanied by a range of emotional responses. As a strikingly understudied task, it is difficult for machines to model and understand desire due to the unavailability of benchmarking datasets with desire and emotion labels. To bridge this gap, we present MSED, the first multi-modal and multi-task sentiment, emotion and desire dataset, which contains 9,190 text-image pairs, with English text. Each multi-modal sample is annotated with six desires, three sentiments and six emotions. We also propose the state-of-the-art baselines to evaluate the potential of MSED and show the importance of multi-task and multi-modal clues for desire understanding. We hope this study provides a benchmark for human desire analysis. MSED will be publicly available for research.