mlsanityData to 14 Oct 2026

GlyphBench: A Playground for Language-Model Reinforcement Learning

arXiv 2609.34214 ↗First posted 28 Sep 2026cs.AIcited by 0

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

also in 1 more
Abstract (arXiv)

We introduce GlyphBench, an environment suite for reinforcement learning (RL) post-training of language-model agents, with over 360 tasks spanning diverse games. GlyphBench renders spatial observations as two-dimensional Unicode grids and connects training, evaluation, and trajectory replay through a unified interface designed to support efficient and reproducible research. We use GlyphBench to study how observation interfaces, reasoning effort, and agent harnesses affect performance, and how RL configurations shape learning dynamics. Our results show that glyph observations outperform native text and pixels in our Craftax experiments, with further gains on several BALROG environments. RL on 100 GlyphBench tasks improves Qwen3.5-4B on held-out Reasoning Gym problems, reaching 63.48% accuracy and outperforming the base model, a math-trained baseline, and a code-trained baseline. These experiments provide empirical evidence that reasoning gains from gameplay can yield stronger transfer than math or code. Together, these results highlight GlyphBench's value as a testbed for systematic research on how language-model agents learn, interact, and generalize.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Reinforcement learning for LLM agents, one of 16 directions in LLM agents and tool use. Builds on 34 of its 43 references that are in the map. First posted on arXiv 28 Sep 2026.

Builds on

34 of 43 references.

Closest ancestors · 5

Same direction, newest first

  1. 2025Training Long-Context, Multi-Turn Software Engineering Agents with Reinforcement Learning5 Aug
  2. 2025RAGEN: Understanding Self-Evolution in LLM Agents via Multi-Turn Reinforcement Learning24 Apr
  3. 2024BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games20 Nov
  4. 2023Voyager: An Open-Ended Embodied Agent with Large Language Models25 May
  5. 2022MineDojo: Building Open-Ended Embodied Agents with Internet-Scale Knowledge17 Jun

Other references · 29

  1. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,695
  2. 2021Evaluating Large Language Models Trained on Codecited by 8,320
  3. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
  4. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  5. 2023SWE-bench: Can Language Models Resolve Real-World GitHub Issues?cited by 3,373
  6. 2019The StarCraft Multi-Agent Challengecited by 597
  7. 2019Leveraging Procedural Generation to Benchmark Reinforcement Learningcited by 483
  8. 2023Minigrid & Miniworld: Modular & Customizable Reinforcement Learning Environments for Goal-Oriented Taskscited by 327
  9. 2018TextWorld: A Learning Environment for Text-based Gamescited by 310
  10. 2025MathArena: Evaluating LLMs on Uncontaminated Math Competitionscited by 297
  11. 2025DeepMath-103K: A Large-Scale, Challenging, Decontaminated, and Verifiable Mathematical Dataset for Advancing Reasoningcited by 280
  12. 2025SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?cited by 279
17 more
  1. 2019MineRL: A Large-Scale Dataset of Minecraft Demonstrationscited by 215
  2. 2021Benchmarking the Spectrum of Agent Capabilitiescited by 200
  3. 2020The NetHack Learning Environmentcited by 198
  4. 2023Grounding Large Language Models in Interactive Environments with Online Reinforcement Learningcited by 195
  5. 2023Guiding Pretraining in Reinforcement Learning with Large Language Modelscited by 187
  6. 2016ViZDoom: A Doom-based AI Research Platform for Visual Reinforcement Learningcited by 179
  7. 2019Interactive Fiction Games: A Colossal Adventurecited by 168
  8. 2019Behaviour Suite for Reinforcement Learningcited by 132
  9. 2025REASONING GYM: Reasoning Environments for Reinforcement Learning with Verifiable Rewardscited by 130
  10. 2021MiniHack the Planet: A Sandbox for Open-Ended Reinforcement Learning Researchcited by 97
  11. 2024Craftax: A Lightning-Fast Benchmark for Open-Ended Reinforcement Learningcited by 88
  12. 2025lmgame-Bench: How Good are LLMs at Playing Games?cited by 55
  13. 2019Obstacle Tower: A Generalization Challenge in Vision, Control, and Planningcited by 54
  14. 2025Play to Generalize: Learning to Reason Through Game Playcited by 16
  15. 2023SPRING: Studying the Paper and Reasoning to Play Gamescited by 11
  16. 2026CODE-SHARP: Continuous Open-ended Discovery and Evolution of Skills as Hierarchical Reward Programscited by 2
  17. 2026Agentick: A Unified Benchmark for General Sequential Decision-Making Agentscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025On the Interplay of Pre-Training, Mid-Training, and RL on Reasoning Language Models
  2. 2025Tool-R1: Sample-Efficient Reinforcement Learning for Agentic Tool Use
  3. 2025ImagineBench: Evaluating Reinforcement Learning with Large Language Model Rollouts
  4. 2025PARL: Prompt-based Agents for Reinforcement Learning
  5. 2025Representation-Based Exploration for Language Models: From Test-Time to Post-Training
  6. 2025GVGAI-LLM: Evaluating Large Language Model Agents with Infinite Games
  7. 2025Reinforcement Learning Meets Large Language Models: A Survey of Advancements and Applications Across the LLM Lifecycle
  8. 2025Beyond Accuracy: Dissecting Mathematical Reasoning for LLMs Under Reinforcement Learning