The Bitter Lesson of Diffusion Language Models for Agentic Workflows: A Comprehensive Reality Check

TL;DR

This paper critically evaluates diffusion-based large language models for agentic workflows, revealing significant limitations in their ability to perform reliable, long-horizon planning and precise tool use despite efficiency claims.

Contribution

The study provides a comprehensive assessment of dLLMs in agentic tasks, introduces DiffuAgent for multi-agent evaluation, and highlights the need for causal and logical reasoning integration.

Findings

dLLMs fail in long-horizon planning under temporal feedback

dLLMs struggle to maintain symbolic precision in tool-calling

dLLMs are effective in non-causal roles like memory summarization

Abstract

The pursuit of real-time agentic interaction has driven interest in Diffusion-based Large Language Models (dLLMs) as alternatives to auto-regressive backbones, promising to break the sequential latency bottleneck. However, does such efficiency gains translate into effective agentic behavior? In this work, we present a comprehensive evaluation of dLLMs (e.g., LLaDA, Dream) across two distinct agentic paradigms: Embodied Agents (requiring long-horizon planning) and Tool-Calling Agents (requiring precise formatting). Contrary to the efficiency hype, our results on Agentboard and BFCL reveal a "bitter lesson": current dLLMs fail to serve as reliable agentic backbones, frequently leading to systematically failure. (1) In Embodied settings, dLLMs suffer repeated attempts, failing to branch under temporal feedback. (2) In Tool-Calling settings, dLLMs fail to maintain symbolic precision (e.g. strict JSON schemas) under diffusion noise. To assess the potential of dLLMs in agentic workflows, we introduce DiffuAgent, a multi-agent evaluation framework that integrates dLLMs as plug-and-play cognitive cores. Our analysis shows that dLLMs are effective in non-causal roles (e.g., memory summarization and tool selection) but require the incorporation of causal, precise, and logically grounded reasoning mechanisms into the denoising process to be viable for agentic tasks.