ARK: Answer-Centric Retriever Tuning via KG-augmented Curriculum Learning

TL;DR

ARK redefines retriever optimization for retrieval-augmented generation by prioritizing answer sufficiency over generic relevance. It constructs a compact knowledge graph from long contexts and uses personalized PageRank to extract query-specific subgraphs, which fuel augmented queries for a three-stage, curriculum-based contrastive fine-tuning. The retriever learns to prefer truly answer-informative chunks while resisting misleading but related content, improving F1 scores by about 14.5% on 10 long-context benchmarks without changing the RAG architecture. Results show strong cross-domain generalization and transferability to different generators, with notable gains on reasoning-heavy tasks. The approach offers a practical, scalable path to truly answer-centric retrieval in long-context settings.

Abstract

Retrieval-Augmented Generation (RAG) has emerged as a powerful framework for knowledge-intensive tasks, yet its effectiveness in long-context scenarios is often bottlenecked by the retriever's inability to distinguish sparse yet crucial evidence. Standard retrievers, optimized for query-document similarity, frequently fail to align with the downstream goal of generating a precise answer. To bridge this gap, we propose a novel fine-tuning framework that optimizes the retriever for Answer Alignment. Specifically, we first identify high-quality positive chunks by evaluating their sufficiency to generate the correct answer. We then employ a curriculum-based contrastive learning scheme to fine-tune the retriever. This curriculum leverages LLM-constructed Knowledge Graphs (KGs) to generate augmented queries, which in turn mine progressively challenging hard negatives. This process trains the retriever to distinguish the answer-sufficient positive chunks from these nuanced distractors, enhancing its generalization. Extensive experiments on 10 datasets from the Ultradomain and LongBench benchmarks demonstrate that our fine-tuned retriever achieves state-of-the-art performance, improving 14.5% over the base model without substantial architectural modifications and maintaining strong efficiency for long-context RAG. Our work presents a robust and effective methodology for building truly answer-centric retrievers.

Figures (3)

• Figure 1: Our RAG Retriever Finetuning Framework ARK, which consists of two major stages: A (Query Construction): From long documents and their corresponding QA pairs, we extract a query-based subgraph using an LLM-generated KG. The subgraph is reformulated with knowledge injection to produce enriched queries. B (Contrastive Finetuning): Using both the original query and injected variants, we identify positive chunks (via alignment scoring) and hard negatives (that match injected queries but differ semantically from ground truth).
• Figure 2: Query Construction Phase. The pipeline begins with KG Construction, where we extract entities, relations, and covariates from long documents to construct an LLM-generated KG. Given a corresponding QA pair, relevant entities are extracted and used to construct PPR-based subgraphs from the KG, with varying maximum sizes to control difficulty. Finally, Augmented Queries are formulated with LLM conditioned on these candidate subgraphs.
• Figure 3: Contrastive Finetuning Phase. Our fine-tuning pipeline comprises two sequential components: Ranking Alignment, in which for each sample, we combine three alignment scores to select the Top-$M$ chunks as positive chunks; followed by Curriculum-based Contrastive Learning, which progressively refines the retriever through (i) in-batch negative sampling, (ii) hard negatives $\mathcal{T}^-_{\texttt{hard}_L}$ mined via query set $\mathcal{T}^-_{\texttt{hard}_L}$, and (iii) more challenging negatives $\mathcal{T}^-_{\texttt{hard}_S}$ obtained from $\mathcal{Q}^{\texttt{aug}}_{S}$.