mlsanityData to 7 Oct 2026

XSkill: Continual Learning from Experience and Skills in Multimodal Agents

arXiv 2603.12056 ↗First posted 12 Mar 2026cs.AI cs.CLcited by 71

Not on the map. Why?

Cited by 47 in LLM agents and tool use, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

Multimodal agents can now tackle complex reasoning tasks with diverse tools, yet they still suffer from inefficient tool use and inflexible orchestration in open-ended settings. A central challenge is enabling such agents to continually improve without parameter updates by learning from past trajectories. We identify two complementary forms of reusable knowledge essential for this goal: experiences, providing concise action-level guidance for tool selection and decision making, and skills, providing structured task-level guidance for planning and tool use. To this end, we propose XSkill, a dual-stream framework for continual learning from experience and skills in multimodal agents. XSkill grounds both knowledge extraction and retrieval in visual observations. During accumulation, XSkill distills and consolidates experiences and skills from multi-path rollouts via visually grounded summarization and cross-rollout critique. During inference, it retrieves and adapts this knowledge to the current visual context and feeds usage history back into accumulation to form a continual learning loop. Evaluated on five benchmarks across diverse domains with four backbone models, XSkill consistently and substantially outperforms both tool-only and learning-based baselines. Further analysis reveals that the two knowledge streams play complementary roles in influencing the reasoning behaviors of agents and show superior zero-shot generalization.

Key facts. Cited by 71 papers in the map (data through 7 Oct 2026). Builds on 34 of its 41 references that are in the map. First posted on arXiv 12 Mar 2026.

Builds on

34 of 41 references.

References · 34

  1. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,655
  2. 2025Qwen3-VL Technical Reportcited by 2,534
  3. 2023ExpeL: LLM Agents Are Experiential Learnerscited by 814
  4. 2025DeepEyes: Incentivizing "Thinking with Images" via Reinforcement Learningcited by 398
  5. 2024Agent Workflow Memorycited by 342
  6. 2025Memory in the Age of AI Agentscited by 285
  7. 2025Pixel Reasoner: Incentivizing Pixel-Space Reasoning with Curiosity-Driven Reinforcement Learningcited by 267
  8. 2025ReasoningBank: Scaling Agent Self-Evolving with Reasoning Memorycited by 253
  9. 2025Thinking with Images for Multimodal Reasoning: Foundations, Methods, and Future Frontierscited by 193
  10. 2023Synapse: Trajectory-as-Exemplar Prompting with Memory for Computer Controlcited by 180
  11. 2025SkillWeaver: Web Agents can Self-Improve by Discovering and Honing Skillscited by 174
  12. 2025Dynamic Cheatsheet: Test-Time Learning with Adaptive Memorycited by 146
22 more
  1. 2025EvolveR: Self-Evolving LLM Agents through an Experience-Driven Lifecyclecited by 129
  2. 2025DeepEyesV2: Toward Agentic Multimodal Modelcited by 119
  3. 2025WebWatcher: Breaking New Frontier of Vision-Language Deep Research Agentcited by 114
  4. 2025Inducing Programmatic Skills for Agentic Taskscited by 103
  5. 2025Agent KB: Leveraging Cross-Domain Experience for Agentic Problem Solvingcited by 90
  6. 2025Perception, Reason, Think, and Plan: A Survey on Large Multimodal Reasoning Modelscited by 89
  7. 2025PyVision: Agentic Vision with Dynamic Toolingcited by 71
  8. 2025Remember Me, Refine Me: A Dynamic Procedural Memory Framework for Experience-Driven Agent Evolutioncited by 65
  9. 2025FLEX: Continuous Agent Evolution via Forward Learning from Experiencecited by 51
  10. 2026REDSearcher: A Scalable and Cost-Efficient Framework for Long-Horizon Search Agentscited by 43
  11. 2025MM-BrowseComp: A Comprehensive Benchmark for Multimodal Browsing Agentscited by 39
  12. 2025MMSearch-Plus: Benchmarking Provenance-Aware Search for Multimodal Browsing Agentscited by 35
  13. 2025TIR-Bench: A Comprehensive Benchmark for Agentic Thinking-with-Images Reasoningcited by 34
  14. 2025MemVerse: Multimodal Memory for Lifelong Learning Agentscited by 31
  15. 2025Beyond Seeing: Evaluating Multimodal LLMs on Tool-Enabled Image Perception, Transformation, and Reasoningcited by 28
  16. 2026AgentVista: Evaluating Multimodal Agents in Ultra-Challenging Realistic Visual Scenarioscited by 22
  17. 2025Agent-X: Evaluating Deep Multimodal Reasoning in Vision-Centric Agentic Taskscited by 17
  18. 2025Thinking with Programming Vision: Towards a Unified View for Thinking with Imagescited by 13
  19. 2025EchoTrail-GUI: Building Actionable Memory for GUI Agents via Critic-Guided Self-Explorationcited by 13
  20. 2025Review of Case-Based Reasoning for LLM Agents: Theoretical Foundations, Architectural Components, and Cognitive Integrationcited by 12
  21. 2025Octopus: Agentic Multimodal Reasoning with Six-Capability Orchestrationcited by 8
  22. 2025Transductive Visual Programming: Evolving Tool Libraries from Experience for Spatial Reasoningcited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 71.

71 papers.

Sort:

By area:

  1. 2026Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skillscited by 161
  2. 2026SkillRet: A Large-Scale Benchmark for Skill Retrieval in LLM Agentscited by 37
  3. 2026Agent-World: Scaling Real-World Environment Synthesis for Evolving General Agent Intelligencecited by 35
  4. 2026ClawsBench: Evaluating Capability and Safety of LLM Productivity Agents in Simulated Workspacescited by 31
  5. 2026Skill1: Unified Evolution of Skill-Augmented Agents via Reinforcement Learningcited by 28
  6. 2026SkillRevise: Improving LLM-Authored Agent Skills via Trace-Conditioned Skill Revisioncited by 21
  7. 2026A Comprehensive Survey on Agent Skills: Taxonomy, Techniques, and Applicationscited by 17
  8. 2026GEMS: Agent-Native Multimodal Generation with Memory and Skillscited by 16
  9. 2026GUI Agents for Continual Game Generationcited by 16
  10. 2026SkillDAG: Self-Evolving Typed Skill Graphs for LLM Skill Selection at Scalecited by 15
  11. 2026Co-Evolving LLM Decision and Skill Bank Agents for Long-Horizon Taskscited by 13
  12. 2026MMSkills: Towards Multimodal Skills for General Visual Agentscited by 13
38 more
  1. 2026Skill-R1: Agent Skill Evolution via Reinforcement Learningcited by 12
  2. 2026EmbodiSkill: Skill-Aware Reflection for Self-Evolving Embodied Agentscited by 12
  3. 2026Evo-Harness: Context-to-Harness Skill Compilation for Self-Evolving Agentscited by 12
  4. 2026SCOPE: Structured Decomposition and Conditional Skill Orchestration for Complex Image Generationcited by 6
  5. 2026SkillFlow: Flow-Driven Recursive Skill Evolution for Agentic Orchestrationcited by 6
  6. 2026Agent Skill Evaluation and Evolution: Frameworks and Benchmarkscited by 6
  7. 2026Dynamic Agent Skills: A Lifecycle Survey and Taxonomy of Evolving Skill Librariescited by 6
  8. 2026AtlasVA: Self-Evolving Visual Skill Memory for Teacher-Free VLM Agentscited by 5
  9. 2026VideoWeaver: Evaluating and Evolving Skills for Agentic Long Video Generationcited by 5
  10. 2026SkillCAT: Contrastive, Assessment-Augmented and Topology-AwareSkill Self-Evolution for LLM Agentscited by 5
  11. 2026COMFYCLAW: Self-Evolving Skill Harnesses for Image Generation Workflowscited by 5
  12. 2026SkillSmith: Co-Evolving Skills and Tools for Self-Improving Agent Systemscited by 4
  13. 2026GeoBrowse: A Geolocation Benchmark for Agentic Tool Use with Expert-Annotated Reasoning Tracescited by 3
  14. 2026PhotoCraft: Agentic Reasoning with Hierarchical Self-Evolving Memory for Deep Image Searchcited by 3
  15. 2026VISUALSKILL: Multimodal Skills for Computer-Use Agentscited by 3
  16. 2026Rethinking Self-Evolving Agent Skills: Feedback Dynamics over Multiple Roundscited by 3
  17. 2026Ace-Skill: Bootstrapping Multimodal Agents with Prioritized and Clustered Evolutioncited by 2
  18. 2026OptSkills: Learning Generalizable Optimization Skills from Problem Archetypes via Cluster-Based Distillationcited by 2
  19. 2026Online Skill Learning for Web Agents via State-Grounded Dynamic Retrievalcited by 2
  20. 2026RedAct: Redacting Agent Capability Traces for Procedural Skill Protectioncited by 2
  21. 2026OpenClaw-Skill: Collective Skill Tree Search for Agentic Large Language Modelscited by 2
  22. 2026Metis: Bridging Text and Code Memory for Self-Evolving Agentscited by 2
  23. 2026SPyCE: Skill-Policy Co-evolution for Multimodal Agentscited by 2
  24. 2026AgentStream: How Well Do Self-Evolving LLM Agents Perform Under Streaming Tasks?cited by 2
  25. 2026SkillReason: Reasoning-Enhanced Agent Skill Retrieval for Implicit User Requestscited by 2
  26. 2026On the Fragility of Self-Improving Agents: Variance, Task Order, and Underspecificationcited by 2
  27. 2026Skill-CMIB: Multimodal Agent Skill for Consistent Action via Conditional Multimodal Information Bottleneckcited by 1
  28. 2026Workflow Closure Is Not Scientific Closure in Auto-Research Systemscited by 1
  29. 2026Dynamo: Dynamic Skill-Tool Evolution for Vision-Language Agentscited by 1
  30. 2026Self-Evolving Code-with-Image Reasoningcited by 1
  31. 2026Routing Before Looking: Query-Adaptive Evidence Acquisition for Long-form Video Understandingcited by 1
  32. 2026SwarmBench: Can Large Language Models Act as Agent Swarm Orchestrators?cited by 1
  33. 2026PRACTICE: From Experience to Expertise in Self-Evolving Embodied Agentscited by 1
  34. 2026SkillGLoW: Procedural-Family Skill Consolidation for Self-Improving Agents on Long-Horizon Task Streamscited by 1
  35. 2026From Interaction Traces to Persistent Skills: Online Evolution for Computer-Use Agentscited by 1
  36. 2026Reflect, Revise, Reuse: Training-Free Skill Evolution for GUI Agentscited by 1
  37. 2026ORACLE: Anticipating Scams from Partial Trajectories in Streaming App Usagecited by 0
  38. 2026You Live More Than Once: Towards Hierarchical Skill Meta-Evolvingcited by 0

50 most cited of 71 All 71 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026MuSEAgent: A Multimodal Reasoning Agent with Stateful Experiences
  2. 2025Agentic Learner with Grow-and-Refine Multimodal Semantic Memory
  3. 2026Agent Skills Should Go Beyond Text: The Case for Visual Skills
  4. 2026RLHarness: Co-evolving Procedural Skills with Reinforcement Learning for Long-horizon Multimodal Reasoning
  5. 2026VC-Tooler: Learning Compositional and Adaptive Visual Tool Use
  6. 2024Optimus-1: Hybrid Multimodal Memory Empowered Agents Excel in Long-Horizon Tasks
  7. 2025Agent-ScanKit: Unraveling Memory and Reasoning of Multimodal Agents via Sensitivity Perturbations
  8. 2026Anchoring What Matters: A Dual-Level Learning Framework for Visually-Grounded Multimodal Reasoning