用简短推理链验证事实，小模型也能达到高准确率且结果可解释。
ThinknCheck: Grounded Claim Verification with Compact, Reasoning-Driven, and Interpretable Models
- 先生成结构化推理再判断真假，训练数据来自增强版事实数据集。
- 10亿参数模型在测试集上准确率达78.1%，比大模型少7倍参数仍更优。
- 适合需要可解释性与低资源部署的智能验证场景，如新闻核查。
我们提出ThinknCheck，一个10亿参数的可信声明验证模型，先生成简短结构化推理，再输出二分类结论。构建了包含24.1k样本的推理增强训练集LLMAggreFact-Think，基于4比特量化后的Gemma3模型进行微调。在LLMAggreFact数据集上，该模型取得78.1%的平衡准确率（BAcc），超越70亿参数的MiniCheck-7B（77.4%），且参数量仅为后者的1/7；若去除推理步骤，准确率降至57.5%。在SciFact数据集上，准确率达到64.7%，较MiniCheck-7B提升14.7个百分点。相比之下，直接使用基础Gemma3-1B进行零样本思维链会降低准确率，而仅用格式+准确率奖励的偏好优化也表现不佳。为深入分析，我们引入GSMClaims数据集及专用版本ThinknCheck-Science，后者在多项基准上均有提升，包括在GSMClaims上达到61.0%准确率。总体表明，显式、监督式推理使小型验证器在保持高效的同时具备竞争力与可解释性。
原文摘要 · Abstract (English)
We present ThinknCheck, a 1B-parameter verifier for grounded claim verification that first produces a short, structured rationale and then a binary verdict. We construct LLMAggreFact-Think, a 24.1k reasoning-augmented training set derived from LLMAggreFact, and fine-tune a 4-bit Gemma3 model to follow this format. On LLMAggreFact, ThinknCheck attains 78.1 balanced accuracy (BAcc), surpassing MiniCheck-7B (77.4) with 7x fewer parameters; removing the reasoning step reduces BAcc to 57.5. On SciFact, ThinknCheck reaches 64.7 BAcc, a +14.7 absolute gain over MiniCheck-7B. By contrast, zero-shot chain-of-thought on the base Gemma3-1B harms accuracy relative to direct answers, and preference optimization with a simple format+accuracy reward underperforms supervised reasoning. To probe the latter, we introduce GSMClaims and a domain-specialized variant, ThinknCheck-Science, which improves across benchmarks, including 61.0\% accuracy on GSMClaims. Overall, explicit, supervised reasoning enables compact verifiers that are competitive while remaining resource-efficient and interpretable.
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。