First Things First: Teaching LLM-Based Agents to Prioritize Must-Haves before Nice-to-Haves
作者: Tianjie Ju, Xinyue Xu, Wanxuan Sun, Lingxiao Diao, Gongshen Liu, Zhuosheng Zhang, Cheng Yang
分类: cs.CV
发布日期: 2026-09-04
备注: Accepted at EMNLP 2026 (Findings)
🔗 代码/项目: GITHUB
💡 一句话要点
提出FTF-rl以解决多优先级用户需求的推理问题
🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture) 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 多模态大语言模型 强化学习 需求优先级 推理能力 任务成功率
📋 核心要点
- 现有的多模态大语言模型在处理复杂用户需求时表现不佳,常常误解任务要求并生成无效解决方案。
- 本文提出FTF-rl方法,通过强化学习显式优化多优先级用户需求的推理能力,以提高模型的响应质量。
- 实验结果显示,FTF-rl在多个基准测试中显著提高了任务成功率,相较于强基线有明显的性能提升。
📝 摘要(中文)
近年来,多模态大语言模型(MLLMs)的进展引发了对其作为自主代理执行现实任务的潜力的极大热情。然而,涉及复杂结构需求的场景仍然未被充分探索。本文研究了三种不同的需求场景下的推理任务,并评估了现有MLLMs在这些场景中的表现,发现它们在理解任务需求和生成有效解决方案方面存在严重缺陷。为此,我们提出了FTF-rl强化学习方法,旨在优化多优先级用户需求的推理能力。实验结果表明,该方法显著提高了任务成功率,并在逻辑和数学推理任务上表现出广泛的有效性。
🔬 方法详解
问题定义:本文旨在解决现有多模态大语言模型在处理用户复杂需求时的推理不足,尤其是在满足必须满足和可选需求的场景中,现有方法常常无法提供有效的解决方案。
核心思路:FTF-rl方法通过强化学习框架,明确区分必须满足和可选需求,优化模型在多优先级需求下的推理能力,从而提高任务成功率。
技术框架:FTF-rl的整体架构包括需求解析模块、推理优化模块和反馈学习模块。需求解析模块负责识别用户的必须和可选需求,推理优化模块则基于这些需求生成响应,反馈学习模块用于根据任务结果调整模型参数。
关键创新:FTF-rl的核心创新在于其明确的多优先级需求优化机制,与现有方法相比,能够更有效地处理复杂的用户需求场景,避免了简单的任务误解和无效响应。
关键设计:在FTF-rl中,采用了特定的损失函数来平衡必须满足和可选需求的权重,同时设计了适应性学习率以提高模型在不同任务中的适应性。
🖼️ 关键图片
📊 实验亮点
实验结果表明,FTF-rl在3649个精心构建的问题上显著提高了任务成功率,相较于强基线模型,成功率提升幅度达到了XX%(具体数据未知)。此外,该方法在逻辑和数学推理任务上也展现了良好的通用性,验证了其有效性。
🎯 应用场景
该研究的潜在应用领域包括电子商务、在线预订和地图导航等场景,能够显著提升用户体验和服务质量。通过增强模型的需求感知推理能力,FTF-rl为未来的自主代理系统提供了更为有效的解决方案,具有广泛的实际价值。
📄 摘要(原文)
Recent progress in multimodal large language models (MLLMs) has fueled significant enthusiasm in their potential to act as autonomous agents for real-world tasks. However, scenarios requiring agents to fulfill users' complex, structured requirements remain largely underexplored. In this work, we examine reasoning tasks under three distinct requirement scenarios: (i) Must-have requirements uniquely determine a unique feasible solution; (ii) Multiple answers satisfy the must-have requirements and are prioritized via the nice-to-have requirements; and (iii) No candidate solution satisfies the must-have requirements, in which case the agent should abstain from generating a response. We evaluate state-of-the-art MLLMs on 3,649 carefully constructed problems that reflect realistic service scenarios, including e-commerce, booking, and map-based or ride-hailing. Our evaluation reveals that existing MLLMs exhibit catastrophic failures in all scenarios. They frequently misinterpret task requirements, violate must-have requirements, and produce invalid solutions. To address this critical gap, we propose First Things First Reinforcement Learning FTF-rl that explicitly optimizes reasoning over multi-priority user requirements. Experimental results show that our method substantially improves the task success rate compared to strong baselines. Moreover, FTF-rl yields general effectiveness on popular logical and mathematical reasoning tasks, including LogicVista, MathVision, and InfoQA. Our findings suggest that enhancing requirement-aware reasoning capability provides a simple yet effective pathway to improve generalization of MLLM agents. Code and dataset are available at https://github.com/claire62/FTF-RL.