# Sandwich Reasoning: An Answer-Reasoning-Answer Approach for Low-Latency Query Correction

arXiv: https://arxiv.org/abs/2601.03672 · HTML: https://mlsanity.com/paper/2601.03672/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 23 references that are in the map.
- First posted on arXiv 7 Jan 2026.

## Builds on (most cited first, 12 of 12)

- [BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension](https://mlsanity.com/paper/1910.13461/) (2019)
- [mT5: A massively multilingual pre-trained text-to-text transformer](https://mlsanity.com/paper/2010.11934/) (2020)
- [LIMO: Less is More for Reasoning](https://mlsanity.com/paper/2502.03387/) (2025)
- [Chain-of-Thought Reasoning Without Prompting](https://mlsanity.com/paper/2402.10200/) (2024)
- [FlashRAG: A Modular Toolkit for Efficient Retrieval-Augmented Generation Research](https://mlsanity.com/paper/2405.13576/) (2024)
- [Can ChatGPT Defend its Belief in Truth? Evaluating LLM Reasoning via Debate](https://mlsanity.com/paper/2305.13160/) (2023)
- [On the (In)Effectiveness of Large Language Models for Chinese Text Correction](https://mlsanity.com/paper/2307.09007/) (2023)
- [Multi-CPR: A Multi Domain Chinese Dataset for Passage Retrieval](https://mlsanity.com/paper/2203.03367/) (2022)
- [Prompting open-source and commercial language models for grammatical error correction of English learner text](https://mlsanity.com/paper/2401.07702/) (2024)
- [GrammarGPT: Exploring Open-Source LLMs for Native Chinese Grammatical Error Correction with Supervised Fine-Tuning](https://mlsanity.com/paper/2307.13923/) (2023)
- [QE-RAG: A Robust Retrieval-Augmented Generation Benchmark for Query Entry Errors](https://mlsanity.com/paper/2504.04062/) (2025)
- [TaoSR1: The Thinking Model for E-commerce Relevance Search](https://mlsanity.com/paper/2508.12365/) (2025)

## Abstract (arXiv)

Query correction is a critical entry point in modern search pipelines, demanding high accuracy strictly within real-time latency constraints. Chain-of-Thought (CoT) reasoning improves accuracy but incurs prohibitive latency for real-time query correction. A potential solution is to output an answer before reasoning to reduce latency; however, under autoregressive decoding, the early answer is independent of subsequent reasoning, preventing the model from leveraging its reasoning capability to improve accuracy. To address this issue, we propose Sandwich Reasoning (SandwichR), a novel approach that explicitly aligns a fast initial answer with post-hoc reasoning, enabling low-latency query correction without sacrificing reasoning-aware accuracy. SandwichR follows an Answer-Reasoning-Answer paradigm, producing an initial correction, an explicit reasoning process, and a final refined correction. To align the initial answer with post-reasoning insights, we design a consistency-aware reinforcement learning (RL) strategy: a dedicated consistency reward enforces alignment between the initial and final corrections, while margin-based rejection sampling prioritizes borderline samples where reasoning drives the most impactful corrective gains. Additionally, we construct a high-quality query correction dataset, addressing the lack of specialized benchmarks for complex query correction. Experimental results demonstrate that SandwichR achieves SOTA accuracy comparable to standard CoT while delivering a 40-70% latency reduction, resolving the latency-accuracy trade-off in online search.