FineRef: Fine-Grained Error Reflection and Correction for Long-Form Generation with Citations

TL;DR

FineRef introduces a novel two-stage training framework enabling LLMs to self-identify and correct citation errors, significantly improving citation accuracy and answer quality in long-form generation with citations.

Contribution

The paper presents FineRef, a new error reflection and correction framework with a two-stage training strategy, enhancing citation fidelity and robustness in LLM-generated long-form answers.

Findings

FineRef outperforms GPT-4 by up to 18% in Citation F1.

It surpasses state-of-the-art models in answer accuracy.

Demonstrates strong generalization in noisy and domain transfer scenarios.

Abstract

Generating with citations is crucial for trustworthy Large Language Models (LLMs), yet even advanced LLMs often produce mismatched or irrelevant citations. Existing methods over-optimize citation fidelity while overlooking relevance to the user query, which degrades answer quality and robustness in real-world settings with noisy or irrelevant retrieved content. Moreover, the prevailing single-pass paradigm struggles to deliver optimal answers in long-form generation that requiring multiple citations. To address these limitations, we propose FineRef, a framework based on Fine-grained error Reflection, which explicitly teaches the model to self-identify and correct two key citation errors, mismatch and irrelevance, on a per-citation basis. FineRef follows a two-stage training strategy. The first stage instills an "attempt-reflect-correct" behavioral pattern via supervised fine-tuning, using fine-grained and controllable reflection data constructed by specialized lightweight models. An online self-reflective bootstrapping strategy is designed to improve generalization by iteratively enriching training data with verified, self-improving examples. To further enhance the self-reflection and correction capability, the second stage applies process-level reinforcement learning with a multi-dimensional reward scheme that promotes reflection accuracy, answer quality, and correction gain. Experiments on the ALCE benchmark demonstrate that FineRef significantly improves both citation performance and answer accuracy. Our 7B model outperforms GPT-4 by up to 18% in Citation F1 and 4% in EM Recall, while also surpassing the state-of-the-art model across key evaluation metrics. FineRef also exhibits strong generalization and robustness in domain transfer settings and noisy retrieval scenarios.