Search-on-Graph-R1 (\sogrone{}), which internalizes this navigation into a compact 8B model through supervised fine-tuning (SFT) followed by reinforcement learning (RL), which surpasses every frozen frontier-LLM system in their comparison and posts the strongest results on CWQ of any system the authors compare against.
Abstract
Knowledge graph question answering (KGQA) requires navigating from topic entities to an answer several relations away. Recent methods prompt a frontier LLM to explore the graph through a retrieval tool, but their reliance on frontier-scale inference makes them costly to deploy. We present Search-on-Graph-R1 (\sogrone{}), which internalizes this navigation into a compact 8B model through supervised fine-tuning (SFT) followed by reinforcement learning (RL). Our central idea is to scaffold a frontier teacher with each question's gold SPARQL query, so the teacher traverses a known answer-bearing path with a live \texttt{Search} tool rather than having to discover the path itself. Since every call executes against a live Freebase server, the resulting trajectories are grounded in the knowledge graph by construction. On WebQSP, CWQ, and GrailQA, \sogrone{} at 8B surpasses every frozen frontier-LLM system in our comparison and posts the strongest results on CWQ of any system we compare against. It does so using no auxiliary module at inference and no LLM judge during training. Isolating each training stage shows that SFT and RL contribute complementary gains, our approach transfers across model families, and RL learns to reach answers in fewer \texttt{Search} calls than its SFT initialization.
Retrieval-Augmented Generation (RAG) methods that integrate external knowledge sources have demonstrated significant effectiveness in addressing the knowledge scarcity and hallucination issues of Large Language Models (LLMs). Compared to traditional approaches relying on documents as knowledge sources, Knowledge Graphs (KGs) enable GraphRAG to emerge as a superior solution through their capability to capture complex inter-dependencies. However, existing GraphRAG methods still exhibit limitations: insufficient attention to optimizing KG construction, introduction of excessive irrelevant information during recall optimization in retriever, and inadequate consideration of the generator's data preferences during retrieval refinement. In this paper, we propose a novel framework KD-GAG that efficiently constructs KGs by distilling teacher LLMs' semantics and reasoning into a smaller student LLM. Furthermore, we propose a preference-based subgraph pruning method to optimize the retrieval process. This method automatically learns subgraph pruning strategies through answer quality feedback and aligns the retriever's selection mechanism with the generator's knowledge preferences. Extensive experiments on four QA benchmarks, including HotpotQA, 2WikiMQA, NQ, and MedHop, demonstrate that KD-GAG achieves strong and consistent performance across multi-hop QA, open-domain QA, and biomedical reasoning settings.
Long Zhao, Yin Xu, Yanyan Wang et al.· Neural Networks· 0 citations
Harness-G, a graph-structured retrieval framework that reformulates free-form query generation as finite action selection, and introduces Structured Non-myopic Credit (SNC), which uses a frozen answer scorer to compare the selected action with its alternatives and assigns downstream gains to the earlier actions that enabled them.
Flat retrieval-augmented generation treats a corpus as a bag of chunks, discarding document hierarchy and cross document structure. We introduce SearchWiki, a harness framework that synthesizes a corpus into a hierarchical, typed, navigable wiki and trains an agent, WikiResearcher-9B, to retrieve information through multi-turn tool use. The wiki organizes knowledge into three layers - document overviews, cross- document topic pages, and page-level source records; enabling progressive refinement of retrieval when initial lookup misses. We optimize the agent's navigation policy with on-policy reinforcement learning with a multi-component reward function balancing answer correctness, retrieval quality and trajectory efficiency. Evaluation on ViDoRe-V3 (8 domains), FinanceBench, and memory benchmarks (LoCoMo, LongMemEval, PersonaMem-v2) shows that WikiResearcher- 9B which is our RL-tuned Qwen 9B model, significantly outperforms same-size untrained baselines and exceeds or matches larger external models. SearchWiki paired with WikiResearcher-9B demonstrates that learned navigation over structured corpora is a superior alternative to flat retrieval.
Guransh Singh, Vishwajeet Kumar, Arkadeep Acharya et al.· 0 citations
KGCaRe is proposed, a hybrid approach that combines neural retrieval with symbolic reasoning over LLM-generated KGs that consistently outperforms existing baselines, including Vanilla LLM, Code Prompt, Text Prompt, Think-on-Graph, Vanilla RAG, and HybridContextQA.
Ghanshyam Verma, Sima Sarkar, Devishree Pillai et al.· 0 citations
G-ReAct is a reasoning framework for deep search that organizes reasoning as state evolution over a fixed-topology query graph, transforming exploratory search driven by textual history into graph-guided reasoning under explicit constraints.
Shaoxiong Yang, Mengyuan Zhang, Shao-Jun Lin et al.· 0 citations
Automated fact-checking remains a challenge for Large Language Models (LLMs) due to"query brittleness"in traditional retrieval systems. We propose DeLIVeR (Decomposed Learning for Information-grounded Veracity Recognition), a framework that treats evidence retrieval as a reinforced strategic exploration task. DeLIVeR utilizes a Planner LLM to decompose complex claims into targeted question sets, which are used to traverse structured Knowledge Graphs (KGs) for high-precision evidence. We optimize the Planner's policy using Group Relative Policy Optimization (GRPO) with a reward system prioritizing structural diversity and verdict accuracy. Our evaluation on LIAR, FEVER, and PolitiFact shows that DeLIVeR significantly outperforms state-of-the-art baselines. Using Qwen2.5-7B, our framework achieved peak F1-scores of 83.73, 84.57, and 79.70 respectively, representing a 10-15% improvement over HippoRAG2. By shifting to a reinforced question-planning strategy, DeLIVeR effectively bridges multi-hop reasoning gaps and provides an auditable, transparent path for verifiable misinformation detection.
C. Nguyen, Thomas Hoang, H. Duong et al.· arXiv.org· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
