Recently, Diffusion Large Language Models (dLLMs) have demonstrated unique efficiency advantages, enabled by their inherently parallel decoding mechanism and flexible generation paradigm. Meanwhile, despite the rapid advancement of Search Agents, their practical deployment is constrained by a fundamental limitation, termed as 1) Latency Challenge : the serial execution of multi-round reasoning, tool calling, and tool response waiting under the ReAct agent paradigm induces severe end-to-end latency. Intuitively, dLLMs can leverage their distinctive strengths to optimize the operational efficiency of agents under the ReAct agent paradigm. Practically, existing dLLM backbones face the 2) Agent Ability Challenge. That is, existing dLLMs exhibit remarkably weak reasoning and tool-calling capabilities, preventing these advantages from being effectively realized in practice. In this paper, we propose DLLM-Searcher, an optimization framework for dLLM-based Search Agents. To solve the Agent Ability Challenge, we design a two-stage post-training pipeline encompassing Agentic Supervised Fine-Tuning (Agentic SFT) and Agentic Variance-Reduced Preference Optimization (Agentic VRPO), which enhances the backbone dLLM's information seeking and reasoning capabilities. To mitigate the Latency Challenge, we leverage the flexible generation mechanism of dLLMs and propose a novel agent paradigm termed Parallel-Reasoning and Acting (P-ReAct). P-ReAct guides the model to prioritize decoding tool_call instructions, thereby allowing the model to keep thinking while waiting for the tool's return. Experimental results demonstrate that DLLM-Searcher achieves performance comparable to mainstream LLM-based search agents and P-ReAct delivers approximately 15% inference acceleration. Our code is available at https://github.com/bubble65/DLLM-Searcher

We report on our experience using common natural language processing (NLP) tools to analyze two vastly different data sets of free-form responses collected during a study of assignments in introductory computing courses. Our first data set consists of typically short comments left by hundreds of students on assignment surveys. Our second data set is comprised of semi-structured individual interviews of eight instructors of up to an hour long each. We collected the data across several years as part of our investigation of the use of pencil puzzles as a context for introductory computer science. In an earlier work, we manually analyzed a fraction of the student comments (all data collected until that point), using grounded theory. The results were illuminating, but the process was very time consuming, consisting of manual assignment of a small number of codes to each comment. In this work, we investigate the usability of common NLP tools to speed up the process for the entire data set of student comments. We also applied these tools to the instructor interviews. The NLP tools do not appear to be effective to create the code base, but, once the code base was determined, they performed the actual coding (assignment of codes to each student comment) promisingly well. For the long-form instructor interviews, the situation was much more challenging, due to the wide-ranging nature of semi-structured interviews, interleaving discussion topics, and elements of natural speech. We report on the lessons learned while automatically analyzing these complex data sets.