HOMURA: Taming the Sand-Glass for Time-Constrained LLM Translation via Reinforcement Learning

• URL: http://arxiv.org/abs/2601.10187v1
• Date: Thu, 15 Jan 2026 08:45:54 GMT
• Title: HOMURA: Taming the Sand-Glass for Time-Constrained LLM Translation via Reinforcement Learning
• Authors: Ziang Cui, Mengran Yu, Tianjiao Li, Chenyu Shi, Yingxuan Shi, Lusheng Zhang, Hongwei Lin,
• Abstract summary: Large Language Models (LLMs) have achieved remarkable strides in multilingual translation but are hindered by a systemic cross-lingual verbosity bias.<n>Current prompt-engineering approaches struggle to resolve this conflict between semantic fidelity and rigid temporal feasibility.
• Score: 10.471350835897757
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large Language Models (LLMs) have achieved remarkable strides in multilingual translation but are hindered by a systemic cross-lingual verbosity bias, rendering them unsuitable for strict time-constrained tasks like subtitling and dubbing. Current prompt-engineering approaches struggle to resolve this conflict between semantic fidelity and rigid temporal feasibility. To bridge this gap, we first introduce Sand-Glass, a benchmark specifically designed to evaluate translation under syllable-level duration constraints. Furthermore, we propose HOMURA, a reinforcement learning framework that explicitly optimizes the trade-off between semantic preservation and temporal compliance. By employing a KL-regularized objective with a novel dynamic syllable-ratio reward, HOMURA effectively "tames" the output length. Experimental results demonstrate that our method significantly outperforms strong LLM baselines, achieving precise length control that respects linguistic density hierarchies without compromising semantic adequacy.

Related papers

• Redefining Machine Simultaneous Interpretation: From Incremental Translation to Human-Like Strategies [6.010207559477024]
  Simultaneous Machine Translation (SiMT) requires high-quality translations under strict real-time constraints.<n>We extend the action space of SiMT with four adaptive actions: Sentence_Cut, Drop, Partial_Summarization and Pronominalization.<n>We adapt these actions in a large language model (LLM) framework and construct training references through action-aware prompting.
  arXiv  Detail & Related papers  (2026-01-16T05:26:16Z)
• Beyond Fixed: Training-Free Variable-Length Denoising for Diffusion Large Language Models [74.15250326312179]
  Diffusion Large Language Models offer efficient parallel generation and capable global modeling.<n>The dominant application ofDLLMs is hindered by the need for a statically predefined generation length.<n>We introduce DAEDAL, a novel training-free denoising strategy that enables Dynamic Adaptive Length Expansion.
  arXiv  Detail & Related papers  (2025-08-01T17:56:07Z)
• Same Question, Different Words: A Latent Adversarial Framework for Prompt Robustness [9.208007322096535]
  We present Latent Adversarial Paraphrasing (LAP), a dual-loop adversarial framework.<n>LAP trains a learnable perturbation to serve as a "latent continuous paraphrase"<n>We conduct experiments to demonstrate the effectiveness of LAP across multiple LLM architectures.
  arXiv  Detail & Related papers  (2025-03-03T09:36:50Z)
• Multilingual LLMs Inherently Reward In-Language Time-Sensitive Semantic Alignment for Low-Resource Languages [19.863010475923414]
  The disparity in labeled resources between resource-rich languages and those considered low-resource remains a significant impediment for Large Language Models (LLMs)<n>Recent strides in cross-lingual in-context learning (X-ICL), mainly through semantically aligned examples retrieved from multilingual pre-trained transformers, have shown promise in mitigating this issue.<n>This study aims to bridge this gap by improving temporal reasoning capabilities in low-resource languages.
  arXiv  Detail & Related papers  (2024-12-11T04:16:39Z)
• Refining Translations with LLMs: A Constraint-Aware Iterative Prompting Approach [7.5069214839655345]
  Large language models (LLMs) have demonstrated remarkable proficiency in machine translation (MT) We propose a multi-step prompt chain that enhances translation faithfulness by prioritizing key terms crucial for semantic accuracy. Experiments using Llama and Qwen as base models on the FLORES-200 and WMT datasets demonstrate significant improvements over baselines.
  arXiv  Detail & Related papers  (2024-11-13T05:40:24Z)
• The Strong Pull of Prior Knowledge in Large Language Models and Its Impact on Emotion Recognition [74.04775677110179]
  In-context Learning (ICL) has emerged as a powerful paradigm for performing natural language tasks with Large Language Models (LLM) We show that LLMs have strong yet inconsistent priors in emotion recognition that ossify their predictions. Our results suggest that caution is needed when using ICL with larger LLMs for affect-centered tasks outside their pre-training domain.
  arXiv  Detail & Related papers  (2024-03-25T19:07:32Z)
• Towards Effective Disambiguation for Machine Translation with Large Language Models [65.80775710657672]
  We study the capabilities of large language models to translate "ambiguous sentences" Experiments show that our methods can match or outperform state-of-the-art systems such as DeepL and NLLB in four out of five language directions.
  arXiv  Detail & Related papers  (2023-09-20T22:22:52Z)
• Are Large Language Models Really Robust to Word-Level Perturbations? [68.60618778027694]
  We propose a novel rational evaluation approach that leverages pre-trained reward models as diagnostic tools. Longer conversations manifest the comprehensive grasp of language models in terms of their proficiency in understanding questions. Our results demonstrate that LLMs frequently exhibit vulnerability to word-level perturbations that are commonplace in daily language usage.
  arXiv  Detail & Related papers  (2023-09-20T09:23:46Z)
• Weakly Supervised Temporal Adjacent Network for Language Grounding [96.09453060585497]
  We introduce a novel weakly supervised temporal adjacent network (WSTAN) for temporal language grounding. WSTAN learns cross-modal semantic alignment by exploiting temporal adjacent network in a multiple instance learning (MIL) paradigm. An additional self-discriminating loss is devised on both the MIL branch and the complementary branch, aiming to enhance semantic discrimination by self-supervising.
  arXiv  Detail & Related papers  (2021-06-30T15:42:08Z)
• Reinforcement Learning for Weakly Supervised Temporal Grounding of Natural Language in Untrimmed Videos [134.78406021194985]
  We focus on the weakly supervised setting of this task that merely accesses to coarse video-level language description annotation without temporal boundary. We propose a emphBoundary Adaptive Refinement (BAR) framework that resorts to reinforcement learning to guide the process of progressively refining the temporal boundary.
  arXiv  Detail & Related papers  (2020-09-18T03:32:47Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.