Nov 2026· ACM Symposium on User Interface Software and Technology· pp. 1-20· 0 citations· 67 references
Computer Science
TL;DR
This work presents Great Grave, a system that transforms LLM chat logs into Research Workflow Graphs, a system that segments conversations into intent-driven Context Units, groups them into decision-tracked Work Units, and extracts typed relations to form a navigable graph.
Abstract
Researchers use LLM chat for sustained intellectual work across long conversations—formulating problems, designing methods, and iterating on ideas. The resulting chat logs function as documents co-authored by the researcher and the model. Yet their linear structure turns them into a rich but practically unusable repository, a “great grave” where valuable decisions and rationale are buried in ever-growing scrolls of text. We present Great Grave, a system that transforms LLM chat logs into Research Workflow Graphs. Our approach segments conversations into intent-driven Context Units, groups them into decision-tracked Work Units, and extracts typed relations to form a navigable graph. An overlay interface provides a List View for hierarchical retrieval and a Graph View for relational exploration. A within-subjects study (N = 12) showed that the overlay significantly improved performance on research comprehension tasks, with the largest gains in tasks requiring cross-temporal comparison and evolutionary reconstruction. While techniques such as context windows and retrieval-augmented generation (RAG) help models leverage history, Great Grave enables users to do the same.
Conversational Large Language Models (LLMs) are widely employed in software engineering for code generation, debugging, and architectural design. However, as developers engage in multi-session development tasks, technical knowledge becomes siloed in isolated, linear chat histories. Reusing context across sessions leads...
This survey presents a structured, design-oriented analysis of RAG-driven conversational systems through a principled framework that decomposes architectures along critical dimensions, including document segmentation and chunking strategies, embedding and indexing mechanisms, retriever and re-ranking models, knowledge...
V. Ghaywat, Abhyuday Singh, A. Shahade et al.· Discover Artificial Intellig...· 0 citations
Long-context LLMs can now ingest entire online discussion threads, but understanding their social discourse requires more than reading a long document: models must track parent-reply relations, turning points, scoped subtrees, cross-branch contrasts, and participant trajectories. To test this structure-aware social rea...
Xin-Yi Liu, R. Khaziev, Dilek Hakkani-Tur et al.· Proceedings of the 2026 Conf...· 0 citations
Aligning user intent in complex, evolving dialogues with Large Language Models (LLMs) is a significant challenge. We attribute this to a cognitive divergence between LLM’s linear reasoning based on context and statistical semantic representations, and the exploratory, non-linear, and personalized nature of human intent...
Jin-Pei Ye, Yao-Rui Fei, Yin-Yu Lu et al.· ACM Symposium on User Interf...· 0 citations
CoWriter is introduced, a proactive agentic scheme that adapts expert human (co-)authoring methodologies from cognitive science to flexibly steer the entire writing workflow while actively encouraging the user to engage.
Shmuel Amar, Benjamin Lieber, Ori Shapira et al.· 0 citations
DocuDebug, an interactive editor that reframes document verification as a software engineering task, is presented, demonstrating how structured visual diagnostics can shift AI-assisted writing from opaque global generation to systematic, user-controlled co-creation.
Hitoshi Matsuyama, Keita Higuchi· ACM Symposium on User Interf...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.