Abstract
The NLG pipeline of Reiter and Dale has long served as the foundational framework for data-to-text system design and evaluation. However its relationship to modern generative architec- tures remains underexplored. In this conceptual analysis, we argue that Retrieval-Augmented Generation (RAG) constitutes a collapsed and partially reconstructed instantiation of the classical NLG pipeline, using it to identify failure modes of RAG around context faithfulness and retrieval non-determinism.
- Anthology ID:
- 2026.retroeval-main.5
- Volume:
- Proceedings of the 1st Symposium on Natural Language Generation Evaluations
- Month:
- June
- Year:
- 2026
- Address:
- Aberdeen, United Kingdom
- Editors:
- Saad Mahamood, David M. Howcroft, Kees van Deemter, Simone Balloccu, Adarsa Sivaprasad, Barkavi Sundararajan, Alberto Bugarín Diz, Jose María Alonso-Moral
- Venue:
- RetroEval
- SIG:
- SIGGEN
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 33–38
- Language:
- URL:
- https://aclanthology.org/2026.retroeval-main.5/
- DOI:
- Bibkey:
- Cite (ACL):
- Adarsa Sivaprasad, Barkavi Sundararajan, and David M. Howcroft. 2026. RAG as a collapsed NLG pipeline. In Proceedings of the 1st Symposium on Natural Language Generation Evaluations, pages 33–38, Aberdeen, United Kingdom. Association for Computational Linguistics.
- Cite (Informal):
- RAG as a collapsed NLG pipeline (Sivaprasad et al., RetroEval 2026)
- Copy Citation:
- PDF:
- https://aclanthology.org/2026.retroeval-main.5.pdf