Despite the remarkable reasoning capabilities of large language models, they still struggle with one-shot relation extraction without predefined relation labels. We identify two pitfalls: models are often misled by irrelevant tokens instead of relation-conveying semantics, and they often fail to align with the abstraction level human annotators expect. We introduce a novel framework that closes this gap with two components: (1) COGRE, a cognitively-inspired reasoning framework that structures RE into a series of processes mimicking human text-processing; and (2) HIT@DICT, a reinforcement learning intermediate reward strategy that encourages reasoning to align with relational labels by rewarding relation-relevant phrases in reasoning. The reward is derived on a credit dictionary automatically extracted from correct predictions. Our experiments show that our framework improves both accuracy and explanation quality by addressing these two pitfalls. For example, COGRE with Qwen2.5-14B-Instruct on One-shot NYT29 achieves 24.65% F1, surpassing prior reasoning-based designs. Optimizing this approach with RL using HIT@DICT further improves performance by +23.46% points. Finally, human evaluation shows that our best model generates relational phrases closely aligned with gold labels, increasing human explanation quality ratings by 54% (relative).

翻译：尽管大语言模型具有显著的推理能力，但在无预定义关系标签的一次性关系抽取任务中仍存在困难。我们识别出两个缺陷：模型常被无关词元而非传递关系的语义误导，且往往无法与人类标注者预期的抽象层级对齐。我们提出一种新框架，通过两个组件弥合这一差距：（1）COGRE，一种受认知启发的推理框架，将关系抽取建模为模仿人类文本处理的一系列过程；（2）HIT@DICT，一种强化学习中间奖励策略，通过奖励推理中与关系相关的短语，促使推理过程与关系标签对齐。该奖励基于从正确预测中自动提取的信用词典衍生。实验表明，我们的框架通过解决这两个缺陷，同时提升了准确性与解释质量。例如，基于Qwen2.5-14B-Instruct的COGRE在一次性NYT29数据集上达到24.65%的F1值，超越先前基于推理的设计。使用HIT@DICT强化学习优化该方法后，性能进一步提升了23.46个百分点。最终，人工评估显示，我们最优模型生成的关系短语与黄金标签高度一致，使人类解释质量评分相对提升54%。