Autonomous large language model (LLM) agents have shown strong potential for open-ended information-seeking tasks. However, existing agent frameworks primarily rely on sequential execution, making them vulnerable to error accumulation in long-horizon reasoning. Recent plan-and-execute approaches partially alleviate this issue, but their fixed planning strategies still struggle to recover from intermediate failures. We propose Logic Tree, a non-linear agentic reasoning framework formulated as a Directed Acyclic Graph (DAG). Instead of following a fixed plan with sequential execution, Logic Tree models complex task solving as a dynamically evolving and recoverable graph search process. The framework decomposes queries into parallel execution branches and incorporates a two-level self-correction mechanism. Local backtracking performs targeted rollback to repair failed reasoning states, while global replanning revises high-level execution strategies and preserves verified intermediate conclusions through a knowledge sanitation process. Experiments on BrowseComp, GAIA, and XBench show that Logic Tree consistently outperforms both sequential and plan-and-execute baselines. Further ablation studies demonstrate that combining DAG-based parallel execution with dynamic error recovery substantially improves robustness in long-horizon research tasks.

Large Language Models (LLMs) achieve excellent performance in natural language reasoning tasks through pre-training on vast unstructured text, enabling them to understand the logic in natural language and generate logic-consistent responses. However, the representational differences between unstructured and structured knowledge make LLMs inherently struggle to maintain logic consistency, leading to Logic Drift challenges in structured knowledge reasoning tasks such as Knowledge Graph Question Answering (KGQA). Existing methods address this limitation by designing complex workflows embedded in prompts to guide LLM reasoning. Nevertheless, these approaches only provide input-level guidance and fail to fundamentally address the Logic Drift in LLM outputs. Additionally, their inflexible reasoning workflows cannot adapt to different tasks and knowledge graphs. To enhance LLMs’ logic consistency in structured knowledge reasoning, we specifically target the logits output from the autoregressive generation process. We propose the Logits-to-Logic framework, which incorporates logits strengthening and logits filtering as core modules to correct logical defects in LLM outputs. Extensive experiments show that our approach significantly improves LLMs’ logic consistency in structured knowledge reasoning and achieves state-of-the-art performance on multiple KGQA benchmarks.

Large Language Models (LLMs) exhibit strong reasoning capabilities in complex tasks. However, they still struggle with hallucinations and factual errors in knowledge-intensive scenarios like knowledge graph question answering (KGQA). We attribute this to the semantic gap between structured knowledge graphs (KGs) and unstructured queries, caused by inherent differences in their focuses and structures. Existing methods usually employ resource-intensive, non-scalable workflows reasoning on vanilla KGs, but overlook this gap. To address this challenge, we propose a flexible framework, Enrich-on-Graph (EoG), which leverages LLMs’ prior knowledge to enrich KGs, bridge the semantic gap between graphs and queries. EoG enables efficient evidence extraction from KGs for precise and robust reasoning, while ensuring low computational costs, scalability, and adaptability across different methods. Furthermore, we propose three graph quality evaluation metrics to analyze query-graph alignment in KGQA task, supported by theoretical validation of our optimization objectives. Extensive experiments on two KGQA benchmark datasets indicate that EoG can effectively generate high-quality KGs and achieve the state-of-the-art performance.