让AI学会在信息不足时提问、条件回答或放弃，提升真实场景推理能力。
Ask, Condition or Abstain: Reinforcement Learning for Missing-Premise Reasoning
- 设计新框架让模型在缺前提时选择问、假设或放弃
- 在274个真实问题上测试，显著提升不确定情境下的推理表现
- 适合研究AI可靠性与人类交互的学者，尤其关注模型边界判断
仅基于答案的强化学习训练模型解决完整问题，但现实查询常缺少关键前提。此时，最优回应不一定是拒绝：模型应能主动提问、基于未知量作答，或在无法提供有效条件回答时选择放弃。本文提出「询问-条件-放弃强化学习」（ACA-RL），通过推理图引导的增强数据流程，将完整问题转化为带局部缺失标注的缺前提训练实例，并采用五类可观察响应行为的结构化奖励进行训练。我们还构建了「缺前提基准测试」（MPB），一个由274个经人工验证的问题组成的基准集，涵盖数学、逻辑和现实世界应用题。在Qwen3与Llama系列模型上，ACA-RL在MPB上表现持续优于基线，同时保持对完整问题的良好性能。配合开源代码、数据集与训练材料，本工作推动自然语言处理评估进入新阶段：不仅衡量模型能否解答完整问题，更检验其识别任务未定性与应对不确定性能力。
原文摘要 · Abstract (English)
Answer-only reinforcement learning (RL) trains reasoning models to solve fully specified problems, but many realistic queries omit a premise needed for a unique answer. In this setting, the useful response is not always refusal: the model should ask for the missing premise, condition its answer on the unknown quantity, or abstain when no informative conditional response is available. We present \emph{Ask-Condition-Abstain Reinforcement Learning} (ACA-RL), a data-augmented RL framework for this setting. Its reasoning-graph-guided pipeline converts well-posed problems into missing-premise training instances with localized gap annotations; ACA-RL then trains on these instances with a structured reward over five observable response behaviors. We also introduce the \emph{Missing-Premise Benchmark} (MPB), a 274-instance human-verified benchmark spanning mathematical, logical, and real-world word problems. Across Qwen3 and Llama models, ACA-RL consistently improves on MPB while preserving competitive performance on well-posed reasoning tasks. Together with the released code, MPB, and training data, this work supports a new mission for NLP evaluation: measuring whether models can recognize when a task is underdetermined and handle uncertainty, not only whether they can answer fully specified questions.
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。