← 研踪 PaperTrailarXiv:2611.06371 · v1
Training-Free Character-Length Control in Summarization with Diffusion Language Models
首次提交 2026-11-04 19:32
版本更新 2026-11-04 19:32
雷达观测 2026-11-05 11:10
摘要
Summarization with a fixed character length is important for tasks such as headline and article preview generation, where text must fit an exact space. Autoregressive language models (LMs) struggle with precise character-length control: training-based methods require task-specific fine-tuning, while zero-shot prompting strategies achieve only approximate compliance. We propose DELTA (Diffusion Expected Length Targeting Algorithm), a training-free inference-time method for masked diffusion LMs that estimates expected character length at each denoising step and dynamically inserts or removes masked positions to steer toward a target length. DELTA requires no fine-tuning and is designed for off-the-shelf masked diffusion LMs. Across three summarization benchmarks, it reduces character-length errors by an order of magnitude compared to baselines while maintaining competitive summary quality in both automatic and human evaluation.
作者备注：Accepted at EMNLP 2026 (Main Conference)
期刊信息：Proceedings of EMNLP 2026, pp. 2803-2813
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-04 19:32 · 最新观测版本
Training-Free Character-Length Control in Summarization with Diffusion Language Models
Summarization with a fixed character length is important for tasks such as headline and article preview generation, where text must fit an exact space. Autoregressive language models (LMs) struggle with precise character-length control: training-based methods require task-specific fine-tuning, while zero-shot prompting strategies achieve only approximate compliance. We propose DELTA (Diffusion Expected Length Targeting Algorithm), a training-free inference-time method for masked diffusion LMs that estimates expected character length at each denoising step and dynamically inserts or removes masked positions to steer toward a target length. DELTA requires no fine-tuning and is designed for off-the-shelf masked diffusion LMs. Across three summarization benchmarks, it reduces character-length errors by an order of magnitude compared to baselines while maintaining competitive summary quality in both automatic and human evaluation.
查看此版本解读 →查看此版本原文 ↗