mlsanityData to 14 Oct 2026

General Modular Harness for LLM Agents in Multi-Turn Gaming Environments

arXiv 2507.11633 ↗First posted 15 Jul 2025cs.AIcited by 11

On the map: LLM agents and tool use → Reinforcement learning for LLM agents

Abstract (arXiv)

We introduce a modular harness design for LLM agents that composes of perception, memory, and reasoning components, enabling a single LLM or VLM backbone to tackle a wide spectrum of multi turn gaming environments without domain-specific engineering. Using classic and modern game suites as low-barrier, high-diversity testbeds, our framework provides a unified workflow for analyzing how each module affects performance across dynamic interactive settings. Extensive experiments demonstrate that the harness lifts gameplay performance consistently over un-harnessed baselines and reveals distinct contribution patterns, for example, memory dominates in long-horizon puzzles while perception is critical in vision noisy arcades. These findings highlight the effectiveness of our modular harness design in advancing general-purpose agent, given the familiarity and ubiquity of games in everyday human experience.

Key facts. Cited by 11 papers in the map (data through 14 Oct 2026). Placed in Reinforcement learning for LLM agents, one of 16 directions in LLM agents and tool use. Builds on 36 of its 70 references that are in the map. First posted on arXiv 15 Jul 2025.

Builds on

36 of 70 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025RAGEN: Understanding Self-Evolution in LLM Agents via Multi-Turn Reinforcement Learning24 Apr
  2. 2025SWEET-RL: Training Multi-Turn LLM Agents on Collaborative Reasoning Tasks19 Mar
  3. 2024BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games20 Nov
  4. 2024GameBench: Evaluating Strategic Reasoning Abilities of LLM Agents7 Jun
  5. 2023Voyager: An Open-Ended Embodied Agent with Large Language Models25 May

Other references · 31

  1. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  2. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,879
  3. 2023MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGIcited by 2,080
  4. 2023WebArena: A Realistic Web Environment for Building Autonomous Agentscited by 2,006
  5. 2023AgentBench: Evaluating LLMs as Agentscited by 1,267
  6. 2023Graph of Thoughts: Solving Elaborate Problems with Large Language Modelscited by 886
  7. 2023DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelinescited by 832
  8. 2024BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructionscited by 546
  9. 2025Humanity's Last Examcited by 491
  10. 2024WebVoyager: Building an End-to-End Web Agent with Large Multimodal Modelscited by 438
  11. 2024Agent Workflow Memorycited by 383
  12. 2024LiveBench: A Challenging, Contamination-Limited LLM Benchmarkcited by 262
19 more
  1. 2023Auto-GPT for Online Decision Making: Benchmarks and Additional Opinionscited by 255
  2. 2024Agent S: An Open Agentic Framework that Uses Computers Like a Humancited by 187
  3. 2025MultiChallenge: A Realistic Multi-Turn Conversation Evaluation Benchmark Challenging to Frontier LLMscited by 180
  4. 2025Can MLLMs Reason in Multimodality? EMMA: An Enhanced MultiModal ReAsoning Benchmarkcited by 134
  5. 2024Cradle: Empowering Foundation Agents Towards General Computer Controlcited by 117
  6. 2023LMRL Gym: Benchmarks for Multi-Turn Reinforcement Learning with Language Modelscited by 88
  7. 2025PHYBench: Holistic Evaluation of Physical Perception and Reasoning in Large Language Modelscited by 61
  8. 2025lmgame-Bench: How Good are LLMs at Playing Games?cited by 55
  9. 2024Gymnasium: A Standard Interface for Reinforcement Learning Environmentscited by 46
  10. 2024ScribeAgent: Towards Specialized Web Agents Using Production-Scale Workflow Datacited by 39
  11. 2023GameEval: Evaluating LLMs on Conversational Gamescited by 37
  12. 2024Code Comprehension then Auditing for Unsupervised LLM Evaluationcited by 32
  13. 2024LMAct: A Benchmark for In-Context Imitation Learning with Long Multimodal Demonstrationscited by 32
  14. 2024GameArena: Evaluating LLM Reasoning through Live Computer Gamescited by 32
  15. 2025EnigmaEval: A Benchmark of Long Multimodal Reasoning Challengescited by 19
  16. 2019SIMBA: Scalable Inversion in Optical Tomography using Deep Denoising Priorscited by 16
  17. 2025TextGames: Learning to Self-Play Text-Based Puzzle Games via Language Model Reasoningcited by 8
  18. 2025Factorio Learning Environmentcited by 7
  19. 2023Improving the Variance of Differentially Private Randomized Experiments through Clusteringcited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 11.

11 papers.

  1. 2026Natural-Language Agent Harnessescited by 61
  2. 2026SafeHarness: Lifecycle-Integrated Security Architecture for LLM-based Agent Deploymentcited by 15
  3. 2026Harness-1: Reinforcement Learning for Search Agents with State-Externalizing Harnessescited by 10
  4. 2026Towards Generalist Game Players: An Investigation of Foundation Models in the Game Multiversecited by 3
  5. 2026Environment-Grounded Automated Prompt Optimization for LLM Game Agentscited by 3
  6. 2026From Agent Loops to Deterministic Graphs: Execution Lineage for Reproducible AI-Native Workcited by 2
  7. 2026HarnessRisk: A Lifecycle-Oriented Benchmark for Agent Harness Safetycited by 2
  8. 2026Think-Before-Speak: From Internal Evaluation to Public Expression in Multi-Agent Social Simulationcited by 1
  9. 2026Intermediate Artifacts as First-Class Citizens: A Data Model for Durable Intermediate Artifacts in Agentic Systemscited by 0
  10. 2026SRHarness: A Harness for Agentic Symbolic Regressioncited by 0
  11. 2026PlaySuite: A Large-Scale Benchmark for Interactive Visual Intelligencecited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026PTCG-Bench: Can LLM Agents Master Pok'emon Trading Card Game?
  2. 2026Sparks of Cooperative Reasoning: LLMs as Strategic Hanabi Agents
  3. 2026VISTA: A Visual Harness for Reasoning in an Interactive World
  4. 2025Orak: A Foundational Benchmark for Training and Evaluating LLM Agents on Diverse Video Games
  5. 2026Compiled Agency: Frontier General-Purpose Coding Agents Build Winning Game Players from Bare Interaction - from Flappy Bird to StarCraft II and Civilization
  6. 2026GameReplica: A Benchmark for Black-Box Visual Game Replication by Vision-Language Agents
  7. 2024A Survey on Large Language Model-Based Game Agents
  8. 2025Leveraging LLM Agents for Automated Video Game Testing