Temporal Graph RAG needs to distinguish when a fact was true in the world from when the system recorded or believed it. These are valid time and transaction time. Keeping both lets a system answer different historical questions; freshness ranking is a separate mechanism and cannot determine whether a fact applies at the time a user asked about.
What are valid time and transaction time?
Valid time is the period when a fact applies in the modeled reality. For example, it can describe when a person actually held a role or when two organizations had a business relationship.
Transaction time is when the database records or treats a fact as current. It captures the system’s own history of assertions, which can differ from the real-world timeline because information may arrive late or be corrected later.
A model that tracks both dimensions is called bitemporal. The distinction matters because “What was true then?” and “What did the database know then?” are not the same question. A single timestamp cannot always answer both.
Free tools Windows power users keep installed
One-click scans. No signup required.
#1 Best Overall
How do the two timelines work?
Consider a graph edge stating that Company A supplied Company B. The edge can have a valid-time interval for when the supply relationship existed and a transaction-time interval for when the database held that assertion. In the temporal property graph model described by Rost and colleagues, vertices and edges carry intervals; the model uses closed-open intervals, meaning the start is included and the end is excluded. Adjacent periods can therefore meet at a boundary without overlapping. Rost et al., “A formal model for temporal property graphs,” VLDB Journal.
Ordinary change
If the relationship is true from January through June and then ends, its valid-time interval ends in June. A query about May should find it; a query about July should not treat it as currently valid. The transaction history can record when the database made that change.
Late-arriving information
Suppose the system learns in April that the supply relationship actually began in January. Its valid time begins in January, while its transaction time begins when the system records the information in April. A query asking what was true in February may include the relationship; a query asking what the database knew in February should not.
Corrections
If the system later discovers that an earlier assertion was wrong, that correction does not necessarily mean reality changed. A bitemporal history can preserve what the database believed before the correction while representing the corrected assertion and its applicable valid time. The exact mechanics—such as whether records are superseded or revised—depend on the database model.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Rank #3
How does this help Graph RAG answer historical questions?
A graph containing only its latest state may answer questions about its current edges but lack enough information to reconstruct an earlier world state or the system’s earlier belief. A bitemporal graph can let retrieval constrain both timelines: the time the fact applied and the time the database held that assertion.
That distinction supports questions such as:
- “What was true on March 1?” Retrieve assertions whose valid-time intervals include March 1.
- “What did we believe on March 1?” Retrieve assertions whose transaction-time intervals include March 1.
- “What did we believe on March 1 about what was true in January?” Constrain transaction time to March 1 and valid time to the relevant period in January.
Temporal fields alone do not guarantee correct answers. Retrieval has to apply the requested time constraint, and the generated answer should preserve enough provenance to identify the supporting assertion and interval. A 2026 research preprint describes one approach using typed temporal operators and trace-grounded answer verification; that is an example design, not a universal requirement for Graph RAG. Xiaofei Zhang, “Temporal Graph Memory Systems for Reliable Graph RAG,” arXiv, July 11, 2026.
Freshness is different from validity
Validity filtering asks whether a fact applies at the time specified in the question. Freshness or recency ranking helps prioritize newer evidence, for example when multiple document versions are otherwise relevant. An expired relationship should not be returned as currently true simply because its source is recent; likewise, an older fact may be the right evidence for a historical question.
One temporal RAG project illustrates a design that separates validity classification from document type and handles expiry and time decay. Its implementation is a project-specific example, not an established standard. Temporal-RAG project README.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteBest Value
What should you check when choosing a temporal graph approach?
“Supports time” can mean different things across databases and frameworks. Evaluate the data model, query language, and retrieval pipeline separately rather than assuming one feature guarantees the others.
- Time dimensions: Does the system support valid time, transaction time, or both?
- What is timed: Can intervals attach to vertices, edges, properties, documents, or only selected records?
- Interval rules: Are boundaries inclusive or exclusive? How are open-ended intervals represented?
- History preservation: Can the system answer the audit or replay questions you need after late arrivals and corrections?
- Query expressiveness: Can queries express both “valid at time T” and “known as of transaction time T”?
- RAG integration: How do temporal filters interact with vector retrieval, graph traversal, ranking, and evidence provenance?
- Evaluation fit: Do benchmarks test your application’s update, correction, and historical-query patterns?
Research on temporal property graphs notes variation in supported time dimensions, graph changes, and whether history is represented as snapshots or time properties. Product documentation is also a reminder that a temporal model and access to temporal behavior through a query language are separate matters.
Check database behavior, not just its terminology
XTDB’s version 1 documentation states that, when a write has no explicit valid-time value, valid time and transaction time take the same value. It also notes a limitation on using valid time in Datalog queries unless a temporal component is present in the documents. This is version-specific documentation, not a claim about current XTDB behavior; check the current documentation before relying on it. XTDB version 1 Clojure API documentation.
What does the recent TGMS benchmark show?
A July 11, 2026 arXiv preprint on Temporal Graph Memory Systems (TGMS) reports results on its own development benchmark. These figures describe that paper’s setup; they are not an industry-wide comparison or independently replicated performance ranking. TGMS preprint.
| Reported result | Qualification |
|---|---|
| 0.409 exact match | TGMS with a 14B open-source model, on the paper’s development benchmark. |
| 0.045–0.182 exact match | Range reported for Vector-RAG, static-graph RAG, and text-to-Cypher baselines in the same setup. |
| 0.67 exact match | TGMS result on the paper’s correction probes; the three 14B baselines were reported at zero. |
| All 500 injected count and entity errors detected | The paper reports no false positives on its clean answers for this verifier evaluation. |
These results illustrate why temporal retrieval and correction handling are worth testing, but they do not establish how another system will perform. The useful comparison is against the historical queries, data changes, and failure cases your application actually needs to handle.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
