Representation of Heterogeneous Temporal Signals

Determine how heterogeneous temporal data should be represented and prioritized for autonomous refactoring decisions, develop architectures for real-time integration of multi-speed data streams, and reliably extract implicit design rationale from historical software artifacts.

Background

Autonomous refactoring must combine signals that occur at different frequencies, including static metrics, commit-level information, production behavior, and customer feedback. Historical artifacts may encode constraints and design rationale that are absent from source code, but extracting and prioritizing this information across spatial and temporal levels remains unresolved.

References

More specifically, the open questions include: How should heterogeneous temporal data be represented and prioritized for refactoring decisions? What architectures can support real-time integration of multi-speed data streams? How can implicit design rationale be reliably extracted from historical artifacts such as commit histories, bug reports, code reviews, and operational logs?

— Continuous Autonomous Refactoring: A Research Roadmap for AI-Driven Code Quality Maintenance  (2609.01236 - Sun et al., 1 Sep 2026) in Section 3.3, Multi-timescale Integration of Heterogeneous Signals

Fewshot prompting therefore fixed the structure of the ADDs (length and style) but not their content, leaving the extraction of concise, well-grounded ADDs from commits an open problem.

— Can LLMs Extract Architectural Design Decisions from Source Code Commits? - A Preliminary Exploratory Study  (2609.03721 - Karan et al., 3 Sep 2026) in Section 4, Results, subsection “RQ₂: Impact of fewshot prompting on ADD alignment,” paragraph “Manual inspection”