RAFT: A Stateful Retrieval-Augmented Framework for Troubleshooting Agents

Source facts

RAFT models each closed support case as a directed chain of timeline entries and retrieves at the entry level, returning the parent-case trajectory anchored at a matching state. It evaluates against a synthetic Microsoft Learn benchmark and real Apache Jira issues with duplicate labels, and reports gains over vanilla RAG and GraphRAG baselines at every case-progress stage.

Analysis

The work treats support histories as state transitions rather than static documents. That design is relevant to enterprise agents whose answer depends on where a case is in its workflow, not only on semantic similarity to a whole document.

Uncertainty/Doubt

• This is an arXiv preprint; the page states EMNLP 2026 Industry Track acceptance, but the evaluation and transfer claim were not independently verified here.