用网络结构提升大模型复杂推理能力，效果优于传统线性或树状方法。
Reasoning Topology Matters: Network-of-Thought for Complex Reasoning Tasks
- 将推理建模为带类型节点和边的有向图，由启发式策略控制搜索过程。
- 在多跳问答任务中，网络结构比树状结构准确率高3个百分点（91.0% vs 88.0%）。
- 适合需要整合多方证据、反复修正的复杂推理场景，尤其适合72B级开源模型使用。
现有提示范式仅支持有限的推理拓扑：链式思维（CoT）生成线性轨迹，树式思维（ToT）进行分支搜索。但复杂推理常需合并中间结果、回溯假设、整合多源证据。本文提出网络式思维（NoT），将推理建模为带类型节点与边的有向图，由基于启发式的控制器策略引导。在四个基准（GSM8K、Game of 24、HotpotQA、ProofWriter）和三个模型（GPT-4o-mini、Llama-3.3-70B-Instruct、Qwen2.5-72B-Instruct）上，我们考察了网络拓扑在何时优于链式或树式结构，评估了大模型自动生成启发式策略的效果，以及不同拓扑下的计算-精度权衡。结果显示，对于顺序任务，CoT在GPT-4o-mini上仍表现良好（GSM8K达89.5%）；而对多跳推理，NoT超越ToT（HotpotQA达91.0% vs 88.0%）。使用72B开源模型时，NoT在GSM8K上取得最高准确率（91.5%），而Qwen2.5-72B在多跳问答中实现最佳结果（HotpotQA达91.7%）。自生成启发式策略优于固定与随机策略，在逻辑推理任务中，仅基于不确定性的权重设置在ProofWriter上达到57.0%。此外，评估方式显著影响排名：字符串匹配会低估所有方法，尤其对NoT影响最大，三模型在HotpotQA上差距达14–18个百分点。
原文摘要 · Abstract (English)
Existing prompting paradigms structure LLM reasoning in limited topologies: Chain-of-Thought (CoT) produces linear traces, while Tree-of-Thought (ToT) performs branching search. Yet complex reasoning often requires merging intermediate results, revisiting hypotheses, and integrating evidence from multiple sources. We propose Network-of-Thought (NoT), a framework that models reasoning as a directed graph with typed nodes and edges, guided by a heuristic-based controller policy. Across four benchmarks (GSM8K, Game of 24, HotpotQA, ProofWriter) and three models (GPT-4o-mini, Llama-3.3-70B-Instruct, Qwen2.5-72B-Instruct), we investigate when network topology outperforms chain or tree structures, whether LLM-generated heuristics can guide graph-based reasoning search, and the computation-accuracy tradeoff across topologies, evaluating each method on accuracy, topology simplicity, and token efficiency. Our results show that CoT remains effective for sequential tasks with GPT-4o-mini (89.5\% on GSM8K), while NoT surpasses ToT on multi-hop reasoning (91.0\% vs.\ 88.0\% on HotpotQA with LLM-as-Judge). With 72B open-source models, NoT achieves the highest accuracy on GSM8K (91.5\%), and Qwen2.5-72B achieves the best multi-hop QA result overall (91.7\% on HotpotQA). Self-generated controller heuristics outperform fixed and random strategies on logical reasoning, with uncertainty-only weighting achieving 57.0\% on ProofWriter. We also find that evaluation methodology significantly impacts method rankings: string-match underestimates all methods on open-ended QA, with the largest gap for NoT, a pattern consistent across all three models (14--18 percentage point gap on HotpotQA).
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。