arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2606.26476cs.LGcs.AI
收藏

检索预热能量基推理:结构化推理任务上扩散即推断的五臂消融方法

Retrieval-Warmed Energy-Based Reasoning: A Five-Arm Ablation Methodology for Diffusion-as-Inference on Structured Reasoning Tasks

Department of Computer Science and Software Engineering, Auburn University(奥本大学计算机科学与软件工程系);Department of Information Management, National Central University(国立中央大学信息管理系)

Libo Sun, Po-Wei Harn, Zewei Zhang, Peixiong He, Xiao Qin

AI总结:

提出五臂消融方法(oracle、best-constant、per-query-random、shuffled、aligned)分离检索预热能量基推理中的三种混淆效应,在连通性-2任务上发现逐图对齐主导性能提升,而在数独任务上关键质量成为瓶颈。

摘要:

热启动扩散采样器加速迭代推理,但很少清楚管线的哪一部分带来了增益。我们研究\textbf{检索预热能量基推理(RW-EBR)}——一种IRED能量基扩散模型\cite{du2024ired},并增加了现代Hopfield轨迹记忆——贡献了一个\textbf{五臂消融方法}(oracle、best-constant、per-query-random、shuffled、aligned),用于分离三种混淆效应:类先验偏移、随机热启动和图对齐的值重用。该诊断分解改编自LLM-RAG评估\cite{ru2024ragchecker}。在\textbf{连通性-2}(Erdős–Rényi全对可达性)上,对齐与打乱oracle的摆动在固定1,000图验证集诊断上达到\textbf{+35个百分点}的平衡准确率,值分布和检索机制固定,仅破坏逐图对齐,而逐查询随机初始化低于冷启动——逐图对齐(而非偏移或随机性)占主导。然而,\emph{可部署的}冷预测管线在存储值质量处未通过接受门。同样的诊断逻辑,在关键质量筛查处停止,应用于\textbf{数独},使用任务特定的关键编码器,在\emph{不同}组件——当前设置下的关键质量——产生明确的阴性。该分解命名了每个任务上的第一个阻塞组件。该设置——通过迭代扩散采样器精化的图可达性,以失败模式的可解释性为视角——将工作置于结构和时空推理中。

英文摘要:

Warm-started diffusion samplers accelerate iterative inference, but it is rarely clear which part of the pipeline carries the gain. We study \textbf{retrieval-warmed energy-based reasoning (RW-EBR)} -- an IRED energy-based diffusion model \cite{du2024ired} augmented with a Modern Hopfield trajectory memory -- and contribute a \textbf{five-arm ablation methodology} (oracle, best-constant, per-query-random, shuffled, aligned) that separates three confounded effects: class-prior bias shift, stochastic warm-starting, and graph-aligned value reuse. The diagnostic decomposition is adapted from LLM-RAG evaluation \cite{ru2024ragchecker}. On \textbf{connectivity-2} (Erdős--Rényi all-pairs reachability), the aligned-vs-shuffled-oracle swing reaches \textbf{$+35$\,pp} balanced accuracy on a fixed 1{,}000-graph validation-set diagnostic, with value distribution and retrieval mechanics fixed, only per-graph alignment destroyed, while per-query random initialisation falls below cold -- per-graph alignment, not bias shift or stochasticity, dominates. Yet the \emph{deployable} cold-prediction pipeline misses the acceptance gate at stored-value quality. The same diagnostic logic, stopped at the key-quality screen, applied to \textbf{Sudoku} with a task-specific key encoder produces a clean negative at a \emph{different} component -- key quality, under the current setup. The decomposition names the first blocking component on each task. The setting -- graph reachability refined by an iterative diffusion sampler, with explainability of failure modes as the lens -- places the work within structured and spatio-temporal reasoning.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑