Temporal QA
QA: Question Answering
Overview
Temporal question answering studies questions that require locating events in time, comparing intervals, tracking changes, or identifying what was true at a specified date. Approaches include temporal knowledge graphs, interval-algebra rules and automata, retrieval over versioned documents and structured memories, and timestamped evidence selection for video and audio. Benchmarks span knowledge-graph tasks such as TimeQA and MultiTQ, evolving legal and policy documents, time-series analysis, and long-video understanding. A central challenge is grounding answers in the right temporal evidence: language models often rely on stale knowledge or miss version boundaries, while retrieval and structured reasoning improve reliability but can add cost and still struggle with complex evidence.
Latest papers
54 papers5 papers in the last four weeks, up 25% on the four weeks before. 0.0% of all new papers.