通过回看机制修复扩散语言模型的生成错误，提升多标记并行生成质量。
BackPlay: Head-Only Look-Back Self-Correction for Diffusion Language Models
- 仅训练轻量修正头，不更新主干模型参数。
- 在多标记解码下显著提升数学推理与代码生成准确率。
- 适合追求高生成质量且资源受限的部署场景。
扩散语言模型（DLMs）并行生成多个标记，但激进的多标记解码会放大跨标记依赖错误，导致生成质量急剧下降。我们提出BackPlay，一种冻结主干的自纠正框架，仅在微调后的DLM上训练轻量修正头，不更新任何主干或适配器参数。由于修正头在与推理时相同的冻结生成器产生的错误上训练，其训练分布与实际部署模型的错误模式一致。我们进一步引入回看纠正机制，将早期更混乱的去噪状态预测注入后期更丰富的上下文，使修正头能利用后期上下文检测早期生成错误。推理时，BackPlay定期通过选择性重掩码和重生成，回溯先前生成的标记以限制错误累积。在数学推理和代码生成基准上，BackPlay在多标记解码下提升了底层DLM的速度-质量权衡。
原文摘要 · Abstract (English)
Diffusion Language Models (DLMs) decode multiple tokens in parallel, but aggressive multi-token decoding amplifies cross-token dependency errors and can sharply degrade generation quality. We propose BackPlay, a frozen-backbone self-correction framework that trains only a lightweight correction head on a finetuned DLM without updating any backbone or adapter parameters. Because the head is trained on errors produced by the same frozen generator used at inference time, its training distribution aligns with the error patterns of the deployed model. We further introduce Look-back Correction, a training mechanism that injects predictions from earlier, more corrupted denoising states into later, richer contexts, enabling the head to leverage later context to detect mistakes made in earlier generation steps. During inference, BackPlay periodically revisits previously generated tokens through selective remasking and regeneration to limit error accumulation. Across mathematical reasoning and code generation benchmarks, BackPlay improves the speed--quality trade-off of the underlying DLM under multi-token decoding.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。