mlsanityData to 7 Oct 2026

METAL: A Multi-Agent Framework for Chart Generation with Test-Time Scaling

arXiv 2502.17651 ↗First posted 24 Feb 2025cs.CV cs.AI cs.CLcited by 33

On the map: LLM reasoning → Multimodal reasoning

also in 1 more
Abstract (arXiv)

Chart generation aims to generate code to produce charts satisfying the desired visual properties, e.g., texts, layout, color, and type. It has great potential to empower the automatic professional report generation in financial analysis, research presentation, education, and healthcare. In this work, we build a vision-language model (VLM) based multi-agent framework for effective automatic chart generation. Generating high-quality charts requires both strong visual design skills and precise coding capabilities that embed the desired visual properties into code. Such a complex multi-modal reasoning process is difficult for direct prompting of VLMs. To resolve these challenges, we propose METAL, a multi-agent framework that decomposes the task of chart generation into the iterative collaboration among specialized agents. METAL achieves 5.2% improvement over the current best result in the chart generation task. The METAL framework exhibits the phenomenon of test-time scaling: its performance increases monotonically as the logarithmic computational budget grows from 512 to 8192 tokens. In addition, we find that separating different modalities during the critique process of METAL boosts the self-correction capability of VLMs in the multimodal context.

Key facts. Cited by 33 papers in the map (data through 7 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 12 of its 56 references that are in the map. First posted on arXiv 24 Feb 2025.

Builds on

12 of 56 references.

Closest ancestors · 2

Same direction, newest first

  1. 2024ChartMimic: Evaluating LMM's Cross-Modal Reasoning Capability via Chart-to-Code Generation14 Jun
  2. 2023ChartBench: A Benchmark for Complex Visual Reasoning in Charts26 Dec

Other references · 10

  1. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,928
  2. 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,739
  3. 2025s1: Simple test-time scalingcited by 1,311
  4. 2024Inference Scaling Laws: An Empirical Analysis of Compute-Optimal Inference for Problem-Solving with Language Modelscited by 282
  5. 2024TradingAgents: Multi-Agents LLM Financial Trading Frameworkcited by 182
  6. 2023ChartLlama: A Multimodal LLM for Chart Understanding and Generationcited by 158
  7. 2024MatPlotAgent: Method and Evaluation for LLM-Based Agentic Scientific Data Visualizationcited by 98
  8. 2024Agents' Room: Narrative Generation through Multi-step Collaborationcited by 57
  9. 2024Scaling LLM Inference with Optimized Sample Compute Allocationcited by 16
  10. 2023Hint-enhanced In-Context Learning wakes Large Language Models up for knowledge-intensive taskscited by 6

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 12; 2026: 21.

33 papers.

Sort:

By area:

  1. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 381
  2. 2025A Survey on Test-Time Scaling in Large Language Models: What, How, Where, and How Well?cited by 215
  3. 2025Large Language Model Agent: A Survey on Methodology, Applications and Challengescited by 194
  4. 2025Where LLM Agents Fail and How They can Learn From Failurescited by 118
  5. 2025From AI for Science to Agentic Science: A Survey on Autonomous Scientific Discoverycited by 74
  6. 2025MDocAgent: A Multi-Modal Multi-Agent Framework for Document Understandingcited by 66
  7. 2025Visual Document Understanding and Reasoning: A Multi-Agent Collaboration Framework with Agent-Wise Adaptive Test-Time Scalingcited by 30
  8. 2025ChartReasoner: Code-Driven Modality Bridging for Long-Chain Reasoning in Chart Question Answeringcited by 11
  9. 2026Dual Latent Memory for Visual Multi-agent Systemcited by 9
  10. 2025Improved Iterative Refinement for Chart-to-Code Generation via Structured Instructioncited by 8
  11. 2026AgentDebugX: An Open-Source Toolkit for Failure Observability, Attribution, and Recovery in LLM Agentscited by 7
  12. 2026Scientific Graphics Program Synthesis via Dual Self-Consistency Reinforcement Learningcited by 6
21 more
  1. 2025EchoFoley: Event-Centric Hierarchical Control for Video Grounded Creative Sound Generationcited by 5
  2. 2026Deep-Reporter: Deep Research for Grounded Multimodal Long-Form Generationcited by 4
  3. 2025OpusAnimation: Code-Based Dynamic Chart Generationcited by 3
  4. 2026PEARL: Self-Evolving Assistant for Time Management with Reinforcement Learningcited by 3
  5. 2026MARTI-MARS2: Scaling Multi-Agent Self-Search via Reinforcement Learning for Code Generationcited by 3
  6. 2026Vision-Guided Iterative Refinement for Frontend Code Generationcited by 3
  7. 2026Multi-Turn Agentic Scientific Literature Search via Workflow Inductioncited by 2
  8. 2026NL2Dashboard: A Lightweight and Controllable Framework for Generating Dashboards with LLMscited by 1
  9. 2026Robotic Policy Adaptation via Weight-Space Meta-Learningcited by 1
  10. 2026BioInsight: Multi-Agent Orchestration for Interactive Biomedical Knowledge Discoverycited by 1
  11. 2025A2R2: Advancing Img2LaTeX Conversion via Visual Reasoning with Attention-Guided Refinementcited by 0
  12. 2026Agentic Mixed-Source Multi-Modal Misinformation Detection with Adaptive Test-Time Scalingcited by 0
  13. 2026Beyond Pixels: Introspective and Interactive Grounding for Visualization Agentscited by 0
  14. 2026A Multi-Agent Framework for Feature-Constrained Difficulty Control in Reading Comprehension Item Generationcited by 0
  15. 2026Test-Time Scaling in Multimodal Foundation Models: A Comprehensive Survey of Generation and Reasoningcited by 0
  16. 2026Imperfect Visual Verification for Code Edition : A Case Study on TikZcited by 0
  17. 2026Animation2Code: Evaluating Temporal Visual Reasoning in Video-to-Code Generationcited by 0
  18. 2026Long-Horizon Embodied Decision-Making via Multimodal Memory Compressioncited by 0
  19. 2026ChartAnno: Benchmarking Multimodal Large Language Models for Chart Annotation Generationcited by 0
  20. 2026FigMirror: Ground It, Code It, Plot Itcited by 0
  21. 2026ChatDev 2.0: A No-Code Multi-Agent Platform for Developing Everythingcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Does It Run and Is That Enough? Revisiting Text-to-Chart Generation with a Multi-Agent Approach
  2. 2026Chart Specification: Structural Representations for Incentivizing VLM Reasoning in Chart-to-Code Generation
  3. 2025Socratic Chart: Cooperating Multiple Agents for Robust SVG Chart Understanding
  4. 2026CharTool: Tool-Integrated Visual Reasoning for Chart Understanding
  5. 2026Beyond Description: A Multimodal Agent Framework for Insightful Chart Summarization
  6. 2025ChartAgent: A Chart Understanding Framework with Tool Integrated Reasoning
  7. 2026RealChart2Code: Advancing Chart-to-Code Generation with Real Data and Multi-Task Evaluation
  8. 2026Hierarchical Visual Agent: Managing Contexts in Joint Image-Text Space for Advanced Chart Reasoning