Decider: A Dual-System Rule-Controllable Decoding Framework for Language Generation

Chen Xu, Tian Lan, Yu Ji, Changlong Yu, Wei Wang, Jun Gao, Qunxi Dong, Kun Qian, Piji Li, Wei Bi, Bin Hu

Abstract

Constrained decoding approaches aim to control the meaning or style of text generated by a Pre-trained Language Model (PLM) for various task-specific objectives at inference time. However, these methods often guide plausible continuations by greedily and explicitly selecting targets, which, while fulfilling the task requirements, may overlook the natural patterns of human language generation. In this work, we propose a novel decoding framework, Decider, which enables us to program high-level rules on how we might effectively complete tasks to control a PLM. Differing from previous works, our framework transforms the encouragement of concrete target words into the encouragement of all words that satisfy the high-level rules. Specifically, Decider is a dual system in which a PLM is equipped and controlled by a First-Order Logic (FOL) reasoner to express and evaluate the rules, along with a decision function that merges the outputs from both systems to guide the generation. Experiments on CommonGen and PersonaChat demonstrate that Decider can effectively follow given rules to guide a PLM in achieving generation tasks in a more human-like manner.
BibTeX
@article{xu2025decider,
  author = {Chen Xu and Tian Lan and Yu Ji and Changlong Yu and Wei Wang and Jun Gao and Qunxi Dong and Kun Qian and Piji Li and Wei Bi and Bin Hu},
  title = {{Decider: A Dual-System Rule-Controllable Decoding Framework for Language Generation}},
  journal = {IEEE Transactions on Knowledge and Data Engineering},
  year = {2025},
  url = {https://openreview.net/forum?id=5M5CJUJjR6},
}
Figure