In-Place Instruction Following in Diffusion Language Models

2026-09-07

AI Summary

研究解决了dLLMs在执行位置指令时对约束条件重视不足的问题，通过提出GRAFT框架，结合约束感知微调和偏好优化方法来提高模型性能。

Abstract

Diffusion Large Language Models (dLLMs) generate text via bidirectional iterative denoising, naturally supporting user-specified constraints anchored at arbitrary output positions, a paradigm known as In-place Prompting (IPP). We formalize this as the In-place Instruction Following (IIF) task and construct IIF-Bench, a hierarchical benchmark spanning literal, style, and discourse-function constraints, paired with a rubric-based local-global evaluation protocol. An inference-time attention-bias probe suggests that vanilla dLLMs often under-prioritize constraint spans during denoising. We then propose GRAFT, an IPP-oriented post-training framework combining constraint-aware SFT and preference optimization. On four representative dLLMs, GRAFT raises the average IIF score from 57.75 to 73.10 (+15.35 points), with absolute gains of 15.91 and 15.57 points on literal and discourse-function constraints, while preserving general generation ability.

Problem

• In-place Instruction Following
• Diffusion Language Models
• User-specified Constraints

Innovation

• In-place Instruction Following
• Diffusion Large Language Models
• GRAFT
• Constraint-aware SFT
• Preference Optimization