Conversational search systems often rewrite a user's query to include context from earlier turns, making the request explicit for retrieval. However, as a new arXiv preprint (2610.15034) points out, this added context can introduce an information need the user did not actually express. That raises a question about how query rewriting interacts with the user's original intent.
The paper, titled "Faithful Query Resolution for Conversational Search," appears to address this problem directly. The abstract suggests the authors are concerned with ensuring that rewritten queries remain faithful to what the user meant, rather than expanding the request in unintended directions. The full details of their proposed resolution method are not visible in the abstract, but the framing is clear: conversational search must balance context enrichment with fidelity to user intent.
This is a timely issue as conversational interfaces become more common. If a system silently adds unstated assumptions, it may retrieve results that miss the user's actual goal. The paper's focus on "faithful" resolution indicates a move toward more disciplined query rewriting, though the specifics await the full text.