mlsanityData to 14 Oct 2026

EverydayGPT: Confidence-Gated Routing for Efficient and Safe Hybrid GPT-RAG Conversational QA

arXiv 2606.11212 ↗First posted 24 Apr 2026cs.CLcited by 0

On the map: Retrieval-augmented generation → Conversational and speech RAG

Abstract (arXiv)

Standard Retrieval-Augmented Generation (RAG) pipelines route every query through retrieval and generation unconditionally, incurring unnecessary computation and propagating low-quality context to the generator. We introduce EverydayGPT, a lightweight conversational QA system built around a Confidence-Gated Routing (CGR) mechanism that formalises the routing decision as a joint policy over retrieval distance and extraction adequacy. The backbone is a 205M-parameter GPT trained from scratch on 10B tokens of FineWeb-Edu. CGR avoids invoking the costly GPT pathway (~5.9s) for 85 percent of queries by resolving them via fast RAG extraction (~45 ms), yielding over 120x latency reduction on the majority of queries while maintaining answer quality. On a 500-question in-domain benchmark, the system achieves F1 = 0.226 +/- 0.004 compared to 0.171 for GPT-only and 0.210 for unconditional RAG. Gains over strong baselines are modest but consistent, while efficiency improvements are substantial (6.3x mean latency reduction). A structured grounding audit finds no unsupported claims in the sampled set, with explicit scope limitations. We position this work as a study of routing strategies under resource constraints rather than a claim of state-of-the-art performance.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Conversational and speech RAG, one of 15 directions in Retrieval-augmented generation. Builds on 14 of its 17 references that are in the map. First posted on arXiv 24 Apr 2026.

Builds on

14 of 17 references.

References · 14

  1. 2017Attention Is All You Needcited by 60,361
  2. 2020Language Models are Few-Shot Learnerscited by 32,419
  3. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,297
  4. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,170
  5. 2017On Calibration of Modern Neural Networkscited by 4,960
  6. 2016SQuAD: 100,000+ Questions for Machine Comprehension of Textcited by 4,148
  7. 2020Dense Passage Retrieval for Open-Domain Question Answeringcited by 3,824
  8. 2019The Curious Case of Neural Text Degenerationcited by 2,916
  9. 2020REALM: Retrieval-Augmented Language Model Pre-Trainingcited by 2,141
  10. 2022Survey of Hallucination in Natural Language Generationcited by 2,075
  11. 2018Know What You Don't Know: Unanswerable Questions for SQuADcited by 1,554
  12. 2020Leveraging Passage Retrieval with Generative Models for Open Domain Question Answeringcited by 1,210
2 more
  1. 2024The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scalecited by 1,160
  2. 2021Retrieval Augmentation Reduces Hallucination in Conversationcited by 692

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024ChatQA: Surpassing GPT-4 on Conversational QA and RAG
  2. 2026Comprehensive Comparison of RAG Methods Across Multi-Domain Conversational QA
  3. 2026The Reasoning Bottleneck in Graph-RAG: Structured Prompting and Context Compression for Multi-Hop QA
  4. 2026Dissecting Agentic RAG: A Component Ablation for Multi-Hop QA with a Local 7B Model
  5. 2024RAGONITE: Iterative Retrieval on Induced Databases and Verbalized RDF for Conversational QA over KGs with RAG
  6. 2024Boosting Conversational Question Answering with Fine-Grained Retrieval-Augmentation and Self-Check
  7. 2025ComposeRAG: A Modular and Composable RAG for Corpus-Grounded Multi-Hop Question Answering
  8. 2024Learning When to Retrieve, What to Rewrite, and How to Respond in Conversational QA