First Things First: Teaching LLM-Based Agents to Prioritize Must-Haves before Nice-to-Haves

• URL: http://arxiv.org/abs/2609.05224v1
• Date: Fri, 04 Sep 2026 14:54:30 GMT
• Title: First Things First: Teaching LLM-Based Agents to Prioritize Must-Haves before Nice-to-Haves
• Abstract summary: We evaluate state-of-the-art multimodal large language models (MLLMs) on 3,649 carefully constructed problems.<n>Existing MLLMs exhibit catastrophic failures in all scenarios.<n>We propose First Things First Reinforcement Learning FTF-rl that explicitly optimize reasoning over multi-priority user requirements.
• Score: 29.390999668963122
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: 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.

Related papers

• A2RBench: An Automatic Paradigm for Formally Verifiable Abstract Reasoning Benchmark Generation [59.98516959731531]
  Abstract reasoning ability reflects intelligence and capacity of LLMs to extract and apply abstract rules.<n>Existing benchmarks either rely on expensive manual annotation, limiting their scale, or risk measuring memorization rather than genuine reasoning.<n>We introduce an automated pipeline named A2RBench, encompassing generation, expansion, evaluation, and analysis.
  arXiv  Detail & Related papers  (2026-05-17T06:14:20Z)
• REAgent: Requirement-Driven LLM Agents for Software Issue Resolution [35.30650311049838]
  Issue resolution aims to automatically generate patches from given issue descriptions.<n>We propose REAgent, a requirement-driven agent framework to guide patch generation.
  arXiv  Detail & Related papers  (2026-04-08T09:22:30Z)
• EIFBENCH: Extremely Complex Instruction Following Benchmark for Large Language Models [64.70546873396624]
  We present the Extremely Complex Instruction Following Benchmark (EIFBENCH) for evaluating large language models (LLMs)<n>EIFBENCH includes multi-task scenarios that enable comprehensive assessment across diverse task types concurrently.<n>We also propose the Segment Policy Optimization (SegPO) algorithm to enhance the LLM's ability to accurately fulfill multi-task workflow.
  arXiv  Detail & Related papers  (2025-06-10T02:39:55Z)
• AGENTIF: Benchmarking Instruction Following of Large Language Models in Agentic Scenarios [51.46347732659174]
  Large Language Models (LLMs) have demonstrated advanced capabilities in real-world agentic applications.<n>AgentIF is the first benchmark for systematically evaluating LLM instruction following ability in agentic scenarios.
  arXiv  Detail & Related papers  (2025-05-22T17:31:10Z)
• Why Do Multi-Agent LLM Systems Fail? [87.90075668488434]
  We introduce MAST-Data, a comprehensive dataset of 1600+ annotated traces collected across 7 popular MAS frameworks.<n>We build the first Multi-Agent System Failure taxonomy (MAST)<n>We leverage MAST and MAST-Data to analyze failure patterns across models (GPT4, Claude 3, Qwen2.5, CodeLlama) and tasks (coding, math, general agent)
  arXiv  Detail & Related papers  (2025-03-17T19:04:38Z)
• Can Long-Context Language Models Subsume Retrieval, RAG, SQL, and More? [54.667202878390526]
  Long-context language models (LCLMs) have the potential to revolutionize our approach to tasks traditionally reliant on external tools like retrieval systems or databases. We introduce LOFT, a benchmark of real-world tasks requiring context up to millions of tokens designed to evaluate LCLMs' performance on in-context retrieval and reasoning. Our findings reveal LCLMs' surprising ability to rival state-of-the-art retrieval and RAG systems, despite never having been explicitly trained for these tasks.
  arXiv  Detail & Related papers  (2024-06-19T00:28:58Z)
• Elicitron: An LLM Agent-Based Simulation Framework for Design Requirements Elicitation [38.98478510165569]
  This paper introduces a novel framework that leverages Large Language Models (LLMs) to automate and enhance the requirements elicitation process. LLMs are used to generate a vast array of simulated users (LLM agents), enabling the exploration of a much broader range of user needs.
  arXiv  Detail & Related papers  (2024-04-04T17:36:29Z)
• Understanding the Weakness of Large Language Model Agents within a Complex Android Environment [21.278266207772756]
  Large language models (LLMs) have empowered intelligent agents to execute intricate tasks within domain-specific software such as browsers and games. LLMs face three primary challenges when applied to general-purpose software systems like operating systems. These challenges motivate AndroidArena, an environment and benchmark designed to evaluate LLM agents on a modern operating system.
  arXiv  Detail & Related papers  (2024-02-09T18:19:25Z)
• Classification, Challenges, and Automated Approaches to Handle Non-Functional Requirements in ML-Enabled Systems: A Systematic Literature Review [10.09767622002672]
  We propose a systematic literature review targeting two key aspects: the classification of the non-functional requirements investigated so far, and the challenges to be faced when developing models in ML-enabled systems. We report that current research identified 30 different non-functional requirements, which can be grouped into six main classes. We also compiled a catalog of more than 23 software engineering challenges, based on which further research should consider the nonfunctional requirements of machine learning-enabled systems.
  arXiv  Detail & Related papers  (2023-11-29T09:45:41Z)
• TOP-Training: Target-Oriented Pretraining for Medical Extractive Question Answering [53.92585020805746]
  We study extractive question-answering in the medical domain (Medical-EQA)<n>This problem has two main challenges: (i) domain specificity, and (ii) extraction-based answering style.<n>We propose TOP-Training, a target-oriented pre-training paradigm.
  arXiv  Detail & Related papers  (2023-10-25T20:48:16Z)
• SatLM: Satisfiability-Aided Language Models Using Declarative Prompting [68.40726892904286]
  We propose a new satisfiability-aided language modeling (SatLM) approach for improving the reasoning capabilities of large language models (LLMs) We use an LLM to generate a declarative task specification rather than an imperative program and leverage an off-the-shelf automated theorem prover to derive the final answer. We evaluate SATLM on 8 different datasets and show that it consistently outperforms program-aided LMs in the imperative paradigm.
  arXiv  Detail & Related papers  (2023-05-16T17:55:51Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.