Search papers, labs, and topics across Lattice.
This paper introduces TIDE, a template-guided iterative framework designed to proactively discover multiple hidden problems within user contexts, rather than relying solely on explicit user requests. By employing iterative discovery and thought templates, TIDE enhances the identification and resolution of coexisting issues in personal workspaces and software repositories. The results demonstrate significant improvements in task coverage and problem resolution compared to traditional single-shot and parallel multi-agent approaches across four model backbones.
TIDE uncovers hidden problems in user contexts, achieving substantial gains in task coverage and resolution that traditional methods miss.
Agents are widely deployed as assistants over documents, tools, and code. However, they typically act only on explicit user requests, which surface only the problems the user has noticed, while many other important problems coexist, hidden in plain sight, within the broader user context, with their total number unknown in advance. We frame this as the task of discovering multiple hidden problems from context, in which coexisting problems should be uncovered, grounded in supporting evidence, and paired with concrete actions. To this end, we introduce TIDE, a template-guided iterative framework with two complementary mechanisms. Specifically, motivated by the observation that single-pass prediction anchors on the most salient cases and yields generic claims, we propose iterative discovery, which surfaces a small batch of candidates per round while conditioning on what has already been found, so subsequent rounds extend coverage; and thought templates, reusable schemas distilled from previously solved cases that specify what contextual signals to attend to and how to connect them, anchoring each prediction in a recognizable problem class. We validate TIDE on two realistic settings, personal workspaces and software repositories, across four model backbones, showing substantial gains over single-shot and parallel multi-agent baselines on task coverage, identification, and resolution.