A developer is building a question-answering system. They have successfully implemented a retrieval component that finds relevant text passages for a user's query. The developer proposes two different approaches for the final generation step:

1. Approach A: Ignore the original query and simply ask a language model to summarize the retrieved text passages.
2. Approach B: Combine the original query with the retrieved text passages and ask the language model to answer the query based on the provided text.

Critique both approaches. Which approach is more likely to produce a helpful and accurate answer for the user, and why? Justify your reasoning by explaining the role of each input component (the original query and the retrieved text).