利用不同角色推理的差异性提升大模型推理鲁棒性。
CHOIR: Collaborative Harmonization fOr Inference Robustness
- 通过协作解码融合多个角色视角的推理结果。
- 在五个群体上平均提升19.2%，最高达26.4%。
- 无需训练，适用于不同模型和任务，适合提升推理可靠性。
角色设定的大语言模型可采用多样角色，实现个性化与情境感知推理。然而，即使在人物设定中出现微小的性别代词变化等人口统计学扰动，也可能改变推理路径，导致正确答案集出现分歧。我们不将这些差异视为需消除的偏见，而是探索其作为提升推理鲁棒性的建设性资源。提出CHOIR（Collaborative Harmonization for Inference Robustness）——一种测试时框架，将多个角色条件化推理信号协同调和为统一预测。CHOIR在反事实角色间执行协作解码，动态平衡推理路径的一致性与差异性。在多个推理基准上的实验表明，CHOIR在不同人口统计群体、模型架构、规模与任务下均能持续提升性能，且无需额外训练。个体群体性能提升最高达26.4%，五个人口群体平均提升19.2%。即使基础角色设定不佳，该方法仍有效。通过将角色差异重构为建设性信号，CHOIR提供了一种可扩展、通用的可靠大模型推理方案。
原文摘要 · Abstract (English)
Persona-assigned Large Language Models (LLMs) can adopt diverse roles, enabling personalized and context-aware reasoning. However, even minor demographic perturbations in personas, such as simple pronoun changes, can alter reasoning trajectories, leading to divergent sets of correct answers. Instead of treating these variations as biases to be mitigated, we explore their potential as a constructive resource to improve reasoning robustness. We propose CHOIR (Collaborative Harmonization fOr Inference Robustness), a test-time framework that harmonizes multiple persona-conditioned reasoning signals into a unified prediction. CHOIR orchestrates a collaborative decoding process among counterfactual personas, dynamically balancing agreement and divergence in their reasoning paths. Experiments on various reasoning benchmarks demonstrate that CHOIR consistently enhances performance across demographics, model architectures, scales, and tasks - without additional training. Improvements reach up to 26.4% for individual demographic groups and 19.2% on average across five demographics. It remains effective even when base personas are suboptimal. By reframing persona variation as a constructive signal, CHOIR provides a scalable and generalizable approach to more reliable LLM reasoning.
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。