• Include only a primary research source, official organization post, technical report, release note, or maintained project page.
• Tag each item as Research or Industry, then with its material type (Paper, Survey, Technical Report, Official Blog, Release, or Project). State its publication date and a neutral, one-sentence explanation of RSI relevance.
• Link research to an official proceedings page, DOI, or arXiv record; link industry items to the organization's original publication, never to a repost or social-media summary.
• An update is not automatically an in-scope taxonomy paper. It enters the L1-L5 catalog only after the same paper-level relevance review.

Table-derived extensions

The following additions are explicitly discussed in Table 3 of the survey, which organizes representative L1 systems by AI development pipeline level.

1. Phi-4-reasoning Technical Report
   [Paper] · arXiv:2504.21318
   
   Survey-table rationale: Training-data selection through LLM evaluation and boundary filtering; selection criteria remain human-defined. Updated object(s): 8.1 Training / Experience Data.

2. NVIDIA Nemotron-4 340B Technical Report
   [Paper] · arXiv:2406.11704
   
   Survey-table rationale: Synthetic supervision is generated and reward-filtered under human-defined quality dimensions. Updated object(s): 8.1 Training / Experience Data.

3. EDIT: Evidence-Diagnosed Intervention Training for Rule-Faithful LLM Grading
   [Paper] · arXiv:2606.06350
   
   Survey-table rationale: Diagnosed-step revisions refine a training signal while diagnostic criteria remain human-defined. Updated object(s): 6.1 Training Objective.

4. AIPC: Agent-Based Automation for AI Model Deployment with Qualcomm AI Runtime
   [Paper] · arXiv:2604.14661
   
   Survey-table rationale: Skill-guided deployment adaptation follows a human-defined deployment procedure. Updated object(s): 3.6 Harness Implementation / Scaffold Code.

5. $V_1$: Unifying Generation and Self-Verification for Parallel Reasoners
   [Paper] · arXiv:2603.04304
   
   Taxonomy: L1. Updated object(s): 5.1 Model Weights.

6. A Co-Evolving Agentic AI System for Medical Imaging Analysis
   [Paper] · arXiv:2509.20279
   
   Taxonomy: L1. Updated object(s): 5.2 Adapter; 2.1 Experience Memory; 3.3 Harness / Workflow.

7. A Context Engineering Framework for Improving Enterprise AI Agents based on Digital-Twin MDP
   [Paper] · arXiv:2603.22083
   
   Taxonomy: L1. Updated object(s): 1.5 Context; 3.1 Workflow.

8. A Dynamic Self-Evolving Extraction System
   [Paper] · arXiv:2603.06915
   
   Taxonomy: L1. Updated object(s): 2.2 Knowledge Memory; 1.5 Context.

9. A Model Can Help Itself: Reward-Free Self-Training for LLM Reasoning
   [Paper] · arXiv:2510.18814
   
   Taxonomy: L1. Updated object(s): 5.1 Model Weights.

10. A Self-Improvable Polymer Discovery Framework Based on Conditional Generative Model
    [Paper] · arXiv:2312.04013
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

11. ACE: Self-Evolving LLM Coding Framework via Adversarial Unit Test Generation and Preference Optimization
    [Paper] · arXiv:2605.16299
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

12. AceSearcher: Bootstrapping Reasoning and Search for LLMs via Reinforced Self-Play
    [Paper] · arXiv:2509.24193
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

13. Active-GRPO: Adaptive Imitation and Self-Improving Reasoning for Molecular Optimization
    [Paper] · arXiv:2607.00531
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

14. AdaExplore: Failure-Driven Adaptation and Diversity-Preserving Search for Efficient Kernel Generation
    [Paper] · arXiv:2604.16625
    
    Taxonomy: L1. Updated object(s): 1.1 Instruction; 2.2 Knowledge Memory.

15. Agent Hospital: A Simulacrum of Hospital with Evolvable Medical Agents
    [Paper] · arXiv:2405.02957
    
    Taxonomy: L1. Updated object(s): 2.1 Experience Memory; 2.3 Procedural Memory.

16. Agent0-VL: Exploring Self-Evolving Agent for Tool-Integrated Vision-Language Reasoning
    [Paper] · arXiv:2511.19900
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

17. AgentGym: Evolving Large Language Model-based Agents across Diverse Environments
    [Paper] · arXiv:2406.04151
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

18. Agentic-Ideation: Sample Efficient Agentic Trajectories Synthesis for Scientific Ideation Agents
    [Paper] · arXiv:2606.31229
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

19. AlgoSkill: Learning to Design Algorithms by Scheduling Human-Like Skills
    [Paper] · arXiv:2606.29999
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

20. AMR-SD: Asymmetric Meta-Reflective Self-Distillation for Token-Level Credit Assignment
    [Paper] · arXiv:2605.18529
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

21. Anyprefer: An Agentic Framework for Preference Data Synthesis
    [Paper] · arXiv:2504.19276
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 1.1 Instruction.

22. APR: Penalizing Structural Redundancy in Large Reasoning Models via Anchor-based Process Rewards
    [Paper] · arXiv:2602.00760
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

23. AutoMalDesc: Large-Scale Script Analysis for Cyber Threat Research
    [Paper] · arXiv:2511.13333
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

24. AutoResearchClaw: Self-Reinforcing Autonomous Research with Human-AI Collaboration
    [Paper] · arXiv:2605.20025
    
    Taxonomy: L1. Updated object(s): 2.1 Experience Memory.

25. Better, Faster: Harnessing Self-Improvement in Large Reasoning Models
    [Paper] · arXiv:2605.24998
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

26. Beyond Imitation: Self-Improving Robot Policies via Off-Policy Q-Planning
    [Paper] · arXiv:2608.21204
    
    Taxonomy: L1. Updated object(s): 5.2 Adapter.

27. Beyond Self-Play: Hierarchical Reasoning for Continuous Motion in Closed-Loop Traffic Simulation
    [Paper] · arXiv:2605.09153
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

28. Binary Decompilation LLM with Feedback-Driven Multi-Turn Refinement
    [Paper] · arXiv:2606.16162
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

29. BiWM: Advancing Open-Source Interactive Video World Models with Bidirectional Autoregression
    [Paper] · arXiv:2606.10135
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

30. Breaking the Self-Confirming Loop: Diagnosing and Mitigating Systemic Reward Bias in Self-Rewarding RL
    [Paper] · arXiv:2510.08977
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

31. Bridging What the Model Thinks and How It Speaks: Expressive Speech Generation via Self-Aware Intent-Realization Alignment
    [Paper] · arXiv:2604.11424
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

32. Can Large Reasoning Models Self-Train?
    [Paper] · arXiv:2505.21444
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

33. Can Recommender Systems Teach Themselves? A Recursive Self-Improving Framework with Fidelity Control
    [Paper] · arXiv:2602.15659
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

34. CAST: Non-Privileged Clipped Asymmetric Self-Teaching with Advantage Flipping for GRPO
    [Paper] · arXiv:2606.00172
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

35. CATPO: Critique-Augmented Tree Policy Optimization
    [Paper] · arXiv:2606.08346
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.4 Curriculum.

36. CEPO: RLVR Self-Distillation using Contrastive Evidence Policy Optimization
    [Paper] · arXiv:2605.19436
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

37. CIIR@LiveRAG 2025: Optimizing Multi-Agent Retrieval Augmented Generation through Self-Training
    [Paper] · arXiv:2506.10844
    
    Taxonomy: L1. Updated object(s): 5.2 Adapter.

38. Cleansing the Artificial Mind: A Self-Reflective Detoxification Framework for Large Language Models
    [Paper] · arXiv:2601.11776
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

39. Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models
    [Paper] · arXiv:2508.00410
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

40. CoAct: Co-Active LLM Preference Learning with Human-AI Synergy
    [Paper] · arXiv:2604.17501
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

41. Consistent Paths Lead to Truth: Self-Rewarding Reinforcement Learning for LLM Reasoning
    [Paper] · arXiv:2506.08745
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

42. Constitutional AI: Harmlessness from AI Feedback
    [Paper] · arXiv:2212.08073
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 7.2 Reward / Fitness.

43. Continual Self-Improvement with Lightweight Experiential Latent Memories
    [Paper] · arXiv:2606.17803
    
    Taxonomy: L1. Updated object(s): 5.2 Adapter; 2.1 Experience Memory.

44. Continuous Self-Improvement of Large Language Models by Test-time Training with Verifier-Driven Sample Selection
    [Paper] · arXiv:2505.19475
    
    Taxonomy: L1. Updated object(s): 5.2 Adapter.

45. CoVerRL: Breaking the Consensus Trap in Label-Free Reasoning via Generator-Verifier Co-Evolution
    [Paper] · arXiv:2603.17775
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

46. CREAM: Consistency Regularized Self-Rewarding Language Models
    [Paper] · arXiv:2410.12735
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

47. CritiCal: Can Critique Help LLM Uncertainty or Confidence Calibration?
    [Paper] · arXiv:2510.24505
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

48. Deciphering Scientific Reasoning Steps from Outcome Data for Molecule Optimization
    [Paper] · arXiv:2603.20262
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

49. Decouple to Generalize: Context-First Self-Evolving Learning for Data-Scarce Vision-Language Reasoning
    [Paper] · arXiv:2512.06835
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

50. DeepInnovator: Triggering the Innovative Capabilities of LLMs
    [Paper] · arXiv:2602.18920
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

51. DeepSeekMath-V2: Towards Self-Verifiable Mathematical Reasoning
    [Paper] · arXiv:2511.22570
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 7.3 Verifier.

52. DemoPSD: Disagreement-Modulated Policy Self-Distillation
    [Paper] · arXiv:2607.02502
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

53. Depression Diagnosis Dialogue Simulation: Self-improving Psychiatrist with Tertiary Memory
    [Paper] · arXiv:2409.15084
    
    Taxonomy: L1. Updated object(s): 2.1 Experience Memory; 2.3 Procedural Memory.

54. DermoGPT: Open Weights and Open Data for Morphology-Grounded Dermatological Reasoning MLLMs
    [Paper] · arXiv:2601.01868
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

55. Differentiable Mixture-of-Agents Incentivizes Swarm Intelligence of Large Language Models
    [Paper] · arXiv:2605.15706
    
    Taxonomy: L1. Updated object(s): 3.3 Harness / Workflow; 5.2 Adapter.

56. Distilling LLM Feedback for Lean Theorem Proving
    [Paper] · arXiv:2605.30861
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

57. Double-Checker: Enhancing Reasoning of Slow-Thinking LLMs via Self-Critical Fine-Tuning
    [Paper] · arXiv:2506.21285
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

58. DRIFT: Difficulty Routing Self-DIstillation with Rhythm-Gated Exploration and Success BuFfer Training
    [Paper] · arXiv:2606.30345
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

59. Dual-Modality Multi-Stage Adversarial Safety Training: Robustifying Multimodal Web Agents Against Cross-Modal Attacks
    [Paper] · arXiv:2603.04364
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

60. DuPO: Enabling Reliable LLM Self-Verification via Dual Preference Optimization
    [Paper] · arXiv:2508.14460
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

61. Dynamic Alignment for Collective Agency: Toward a Scalable Self-Improving Framework for Open-Ended LLM Alignment
    [Paper] · arXiv:2512.05464
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

62. EARS: Explanatory Abstention for Reliable Sub-Agent Modeling in Large-scale Multi-Agent Systems
    [Paper] · arXiv:2606.18668
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

63. Easy Samples Are All You Need: Self-Evolving LLMs via Data-Efficient Reinforcement Learning
    [Paper] · arXiv:2604.18639
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

64. Emulating Clinician Cognition via Self-Evolving Deep Clinical Research
    [Paper] · arXiv:2603.10677
    
    Taxonomy: L1. Updated object(s): 2.1 Experience Memory.

65. Endless Terminals: Scaling RL Environments for Terminal Agents
    [Paper] · arXiv:2601.16443
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.3 Environment.

66. Enhancing Mathematical Reasoning in LLMs with Background Operators
    [Paper] · arXiv:2412.04110
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

67. EpiCaR: Knowing What You Don't Know Matters for Better Reasoning in LLMs
    [Paper] · arXiv:2601.06786
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

68. Evo-MedAgent: Beyond One-Shot Diagnosis with Agents That Remember, Reflect, and Improve
    [Paper] · arXiv:2604.14475
    
    Taxonomy: L1. Updated object(s): 2.1 Experience Memory; 2.3 Procedural Memory.

69. EvoDefense: Co-Evolving Black-Box Defense with Large Language Models
    [Paper] · arXiv:2605.31140
    
    Taxonomy: L1. Updated object(s): 5.2 Adapter; 2.1 Experience Memory.

70. EVOLVE-VLA: Test-Time Training from Environment Feedback for Vision-Language-Action Models
    [Paper] · arXiv:2512.14666
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

71. ExpeL: LLM Agents Are Experiential Learners
    [Paper] · arXiv:2308.10144
    
    Taxonomy: L1. Updated object(s): 2.1 Experience Memory; 2.2 Knowledge Memory.

72. Exploring and Exploiting the Inherent Efficiency within Large Reasoning Models for Self-Guided Efficiency Enhancement
    [Paper] · arXiv:2506.15647
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

73. ExpRL: Exploratory RL for LLM Mid-Training
    [Paper] · arXiv:2606.17024
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

74. FiRe: Fine-grained Multimodal Reasoning for Enhanced Image Generation
    [Paper] · arXiv:2604.13491
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

75. First SFT, Second RL, Third UPT: Continual Improving Multi-Modal LLM Reasoning via Unsupervised Post-Training
    [Paper] · arXiv:2505.22453
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

76. Forward-Free Diffusion Language Models with BPTT-Free Looped Refinement
    [Paper] · arXiv:2606.08357
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

77. From Memorization to Creativity: LLM as a Designer of Novel Neural Architectures
    [Paper] · arXiv:2601.02997
    
    Taxonomy: L1. Updated object(s): 5.2 Adapter; 8.1 Experience Data.

78. Generating on Generated: An Approach Towards Self-Evolving Diffusion Models
    [Paper] · arXiv:2502.09963
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

79. GRAM-R^2: Self-Training Generative Foundation Reward Models for Reward Reasoning
    [Paper] · arXiv:2509.02492
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

80. GSEM: Graph-based Self-Evolving Memory for Experience Augmented Clinical Reasoning
    [Paper] · arXiv:2603.22096
    
    Taxonomy: L1. Updated object(s): 2.1 Experience Memory; 2.4 Memory & Knowledge.

81. Heterogeneous Self-Play for Realistic Highway Traffic Simulation
    [Paper] · arXiv:2604.16406
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

82. How to Train Your Long-Context Visual Document Model
    [Paper] · arXiv:2602.15257
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 7.1 Judge.

83. HS-STaR: Hierarchical Sampling for Self-Taught Reasoners via Difficulty Estimation and Budget Reallocation
    [Paper] · arXiv:2505.19866
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.4 Curriculum.

84. ICRL: Learning to Internalize Self-Critique with Reinforcement Learning
    [Paper] · arXiv:2605.15224
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

85. Improving LLM Code Reasoning via Semantic Equivalence Self-Play with Formal Verification
    [Paper] · arXiv:2604.17010
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 5.2 Adapter.

86. Improving Reasoning in Vision-Language Models via Perception Verified Self-Training
    [Paper] · arXiv:2606.22158
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

87. IRIS: Interpolative Rényi Iterative Self-play for Large Language Model Fine-Tuning
    [Paper] · arXiv:2604.20933
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

88. Iterative Compositional Data Generation for Robot Control
    [Paper] · arXiv:2512.10891
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

89. JailDAM: Jailbreak Detection with Adaptive Memory for Vision-Language Model
    [Paper] · arXiv:2504.03770
    
    Taxonomy: L1. Updated object(s): 2.2 Knowledge Memory; 2.5 Memory & Knowledge.

90. KBSpec: LLM-driven Formal Specification Generation with Evolving Domain Knowledge Base
    [Paper] · arXiv:2606.21339
    
    Taxonomy: L1. Updated object(s): 2.2 Knowledge Memory; 2.3 Procedural Memory.

91. Keep Policy Gradient in Charge: Sibling-Guided Credit Distillation for Long-Horizon Tool-Use Agents
    [Paper] · arXiv:2606.12634
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

92. Know When to Stop: Segment-Level Credit Assignment for Reducing Overthinking
    [Paper] · arXiv:2607.00482
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

93. KVerus: Scalable and Resilient Formal Verification Proof Generation for Rust Code
    [Paper] · arXiv:2605.03822
    
    Taxonomy: L1. Updated object(s): 9.1 Program Code; 2.2 Knowledge Memory.

94. Latent Preference Modeling for Cross-Session Personalized Tool Calling
    [Paper] · arXiv:2604.17886
    
    Taxonomy: L1. Updated object(s): 2.2 Knowledge Memory.

95. Learn from Your Mistakes: Tree-like Self-Play for Secure Code LLMs
    [Paper] · arXiv:2606.03489
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

96. Learn Your Reference Model for Real Good Alignment
    [Paper] · arXiv:2404.09656
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

97. Learning from Language Feedback via Variational Policy Distillation
    [Paper] · arXiv:2605.15113
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights.

98. Learning from Supervision with Semantic and Episodic Memory: A Reflective Approach to Agent Adaptation
    [Paper] · arXiv:2510.19897
    
    Taxonomy: L1. Updated object(s): 2.1 Experience Memory; 2.2 Knowledge Memory.

99. Learning from Your Own Mistakes: Constructing Learnable Micro-Reflective Trajectories for Self-Distillation
    [Paper] · arXiv:2606.18844
    
    Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

100. Learning Steerable Clarification Policies with Collaborative Self-play
     [Paper] · arXiv:2512.04068
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

101. Learning to Better Search with Language Models via Guided Reinforced Self-Training
     [Paper] · arXiv:2410.02992
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

102. Learning to Plan & Reason for Evaluation with Thinking-LLM-as-a-Judge
     [Paper] · arXiv:2501.18099
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

103. Learning to Plan Before Answering: Self-Teaching LLMs to Learn Abstract Plans for Problem Solving
     [Paper] · arXiv:2505.00031
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

104. Learning to Self-Verify Makes Language Models Better Reasoners
     [Paper] · arXiv:2602.07594
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

105. Learning to Solve and Verify: A Self-Play Framework for Code and Test Generation
     [Paper] · arXiv:2502.14948
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

106. LLM-Personalize: Aligning LLM Planners with Human Preferences via Reinforced Self-Training for Housekeeping Robots
     [Paper] · arXiv:2404.14285
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

107. Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judge
     [Paper] · arXiv:2407.19594
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

108. MiGrATe: Mixed-Policy GRPO for Adaptation at Test-Time
     [Paper] · arXiv:2508.08641
     
     Taxonomy: L1. Updated object(s): 5.2 Adapter.

109. MobileIPL: Enhancing Mobile Agents Thinking Process via Iterative Preference Learning
     [Paper] · arXiv:2505.12299
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

110. Monte Carlo Tree Search with Reasoning Path Refinement for Small Language Models in Conversational Text-to-NoSQL
     [Paper] · arXiv:2602.12574
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

111. Motus2: A Self-Evolving General World Model for Dexterous Manipulation
     [Paper] · arXiv:2608.30237
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.4 Data & Environment; 7.2 Reward / Fitness.

112. Not All Synthetic Data Is Yours to Learn From
     [Paper] · arXiv:2605.31126
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

113. Object-Centric Residual RL for Zero-Shot Sim-to-Real VLA Enhancement
     [Paper] · arXiv:2606.18953
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 5.2 Adapter.

114. Observations and Remedies for Large Language Model Bias in Self-Consuming Performative Loop
     [Paper] · arXiv:2601.05184
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

115. OmniOPSD: Rationale-Privileged On-Policy Self-Distillation for Affective Computing
     [Paper] · arXiv:2606.15920
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

116. OWL: Unsupervised 3D Object Detection by Occupancy Guided Warm-up and Large Model Priors Reasoning
     [Paper] · arXiv:2512.05698
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

117. PaperSearchQA: Learning to Search and Reason over Scientific Papers with RLVR
     [Paper] · arXiv:2601.18207
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

118. PARNESS: A Paper Harness for End-to-End Automated Scientific Research with Dynamic Workflows, Full-Text Indexing, and Cross-Run Knowledge Accumulation
     [Paper] · arXiv:2605.05258
     
     Taxonomy: L1. Updated object(s): 2.2 Knowledge Memory; 2.1 Experience Memory.

119. Paying More Attention to Visual Tokens in Self-Evolving Large Multimodal Models
     [Paper] · arXiv:2606.27373
     
     Taxonomy: L1. Updated object(s): 5.2 Adapter; 6.1 Objective.

120. PBSD: Privileged Bayesian Self-Distillation for Long-Horizon Credit Assignment
     [Paper] · arXiv:2606.09348
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

121. PEARL: Training Socratic Tutors with Pedagogically Aligned Reinforcement Learning
     [Paper] · arXiv:2605.29582
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

122. Peer-Predictive Self-Training for Language Model Reasoning
     [Paper] · arXiv:2604.13356
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

123. Perception, Verdict, and Evolution: Hindsight-Driven Self-Refining Forensics Agent for AI-Generated Image Detection
     [Paper] · arXiv:2606.26552
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

124. PLLM: Pseudo-Labeling Large Language Models for CAD Program Synthesis
     [Paper] · arXiv:2602.12561
     
     Taxonomy: L1. Updated object(s): 5.2 Adapter.

125. Policy-Guided Stepwise Model Routing for Cost-Effective Reasoning
     [Paper] · arXiv:2605.06116
     
     Taxonomy: L1. Updated object(s): 3.3 Harness / Workflow.

126. Post-Training Probability Manifold Correction via Structured SVD Pruning and Self-Referential Distillation
     [Paper] · arXiv:2602.00372
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

127. Poster: ClawdGo: Endogenous Security Awareness Training for Autonomous AI Agents
     [Paper] · arXiv:2604.24020
     
     Taxonomy: L1. Updated object(s): 2.3 Procedural Memory; 2.5 Memory & Knowledge; 8.2 Task Generator.

128. PRefLexOR: Preference-based Recursive Language Modeling for Exploratory Optimization of Reasoning and Agentic Thinking
     [Paper] · arXiv:2410.12375
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

129. PRISMA: Preference-Reinforced Self-Training Approach for Interpretable Emotionally Intelligent Negotiation Dialogues
     [Paper] · arXiv:2604.18354
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

130. Probing Visual Language Priors in VLMs
     [Paper] · arXiv:2501.00569
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

131. Procedural Memory Distillation: Online Reflection for Self-Improving Language Models
     [Paper] · arXiv:2607.01480
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

132. Process-based Self-Rewarding Language Models
     [Paper] · arXiv:2503.03746
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

133. ProcessThinker: Enhancing Multi-modal Large Language Models Reasoning via Rollout-based Process Reward
     [Paper] · arXiv:2606.11209
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

134. PsychAgent: An Experience-Driven Lifelong Learning Agent for Self-Evolving Psychological Counselor
     [Paper] · arXiv:2604.00931
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 4.4 Skill Library.

135. Pythagoras-Prover: Advancing Efficient Formal Proving via Augmented Lean Formalisation
     [Paper] · arXiv:2606.12594
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

136. Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking
     [Paper] · arXiv:2403.09629
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

137. Reasoning and Learning a Perceptual Metric for Self-Training of Reflective Objects in Bin-Picking with a Low-cost Camera
     [Paper] · arXiv:2503.20207
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

138. Reasoning-Enhanced Self-Training for Long-Form Personalized Text Generation
     [Paper] · arXiv:2501.04167
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

139. RebuttalAgent: Strategic Persuasion in Academic Rebuttal via Theory of Mind
     [Paper] · arXiv:2601.15715
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

140. Red-Bandit: Test-Time Adaptation for LLM Red-Teaming via Bandit-Guided LoRA Experts
     [Paper] · arXiv:2510.07239
     
     Taxonomy: L1. Updated object(s): 5.2 Adapter; 3.3 Harness / Workflow.

141. ReGuide: From Test-Time Guidance to Self-Improving Diffusion Policies
     [Paper] · arXiv:2606.28939
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

142. Reinforced Self-Training (ReST) for Language Modeling
     [Paper] · arXiv:2308.08998
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

143. Reinforcement Learning for Self-Improving Agent with Skill Library
     [Paper] · arXiv:2512.17102
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 4.4 Skill Library.

144. Reinforcement Learning from Rich Feedback with Distributional DAgger
     [Paper] · arXiv:2606.05152
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

145. ReNIO: Reweighting Negative Trajectory Importance for LLM On-Policy Distillation
     [Paper] · arXiv:2606.23104
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

146. ReST-MCTS: LLM Self-Training via Process Reward Guided Tree Search*
     [Paper] · arXiv:2406.03816
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 7.2 Reward / Fitness.

147. ReSteer: Quantifying and Refining the Steerability of Multitask Robot Policies
     [Paper] · arXiv:2603.17300
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

148. Rethinking Reward Supervision: Rubric-Conditioned Self-Distillation
     [Paper] · arXiv:2606.19327
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

149. RetroAgent: From Solving to Evolving via Retrospective Dual Intrinsic Feedback
     [Paper] · arXiv:2603.08561
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 2.1 Experience Memory.

150. Right in the Right Way: LM Training with Verifiable Rewards and Human Demonstrations
     [Paper] · arXiv:2607.01181
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective; 5.2 Adapter.

151. RLSR: Reinforcement Learning from Self Reward
     [Paper] · arXiv:2505.08827
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

152. Robot Self-Improvement via Human-Video Dynamics Models
     [Paper] · arXiv:2606.21406
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

153. ROSD: Reflective On-Policy Self-Distillation for Language Model Reasoning across Domains
     [Paper] · arXiv:2605.28014
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

154. Route, Retrieve, Reflect, Repair: Self-Improving Agentic Framework for Visual Detection and Linguistic Reasoning in Medical Imaging
     [Paper] · arXiv:2601.08192
     
     Taxonomy: L1. Updated object(s): 2.1 Experience Memory; 1.4 Prompt & Context.

155. S-SPPO: Semantic-Calibrated Self-Play Preference Optimization
     [Paper] · arXiv:2606.01561
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

156. SAEExplainer: Interpreting SAE Features with Activation-Guided Preference Optimization
     [Paper] · arXiv:2606.08496
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

157. Safe Autoregressive Image Generation with Iterative Self-Improving Codebooks
     [Paper] · arXiv:2606.27147
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

158. SafeCtrl-RL: Inference-Time Adaptive Behaviour Control for LLM Dialogue via RL-Driven Prompt Optimisation
     [Paper] · arXiv:2605.25984
     
     Taxonomy: L1. Updated object(s): 5.2 Adapter; 3.4 Harness / Workflow.

159. SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment
     [Paper] · arXiv:2606.02530
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

160. SD-Search: On-Policy Hindsight Self-Distillation for Search-Augmented Reasoning
     [Paper] · arXiv:2605.18299
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

161. Search-E1: Self-Distillation Drives Self-Evolution in Search-Augmented Reasoning
     [Paper] · arXiv:2605.22511
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

162. See, Think, Learn: A Self-Taught Multimodal Reasoner
     [Paper] · arXiv:2512.02456
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

163. Self-Adapting Improvement Loops for Robotic Learning
     [Paper] · arXiv:2506.06658
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

164. Self-Alignment with Instruction Backtranslation
     [Paper] · arXiv:2308.06259
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

165. Self-Consistency of the Internal Reward Models Improves Self-Rewarding Language Models
     [Paper] · arXiv:2502.08922
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

166. SELF-EMO: Emotional Self-Evolution from Recognition to Consistent Expression
     [Paper] · arXiv:2604.18003
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

167. Self-Evolution for Multi-Turn Tool-Calling Agents via Divergence-Point Preference Learning
     [Paper] · arXiv:2606.23112
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

168. Self-evolving LLM agents with in-distribution Optimization
     [Paper] · arXiv:2606.07367
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 5.2 Adapter.

169. Self-Improving Embodied Foundation Models
     [Paper] · arXiv:2509.15155
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

170. Self-Improving Tabular Language Models via Iterative Reward-Guided Post-Training
     [Paper] · arXiv:2604.18966
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

171. Self-Improving VLM Judges Without Human Annotations
     [Paper] · arXiv:2512.05145
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

172. Self-Play Enhancement via Advantage-Weighted Refinement in Online Federated LLM Fine-Tuning with Real-Time Feedback
     [Paper] · arXiv:2605.07977
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

173. Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models (SPIN)
     [Paper] · arXiv:2401.01335
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

174. Self-playing Adversarial Language Game Enhances LLM Reasoning (SPAG)
     [Paper] · arXiv:2404.10642
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

175. Self-Policy Distillation via Capability-Selective Subspace Projection
     [Paper] · arXiv:2605.22675
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

176. Self-Reasoning Language Models: Unfold Hidden Reasoning Chains with Few Reasoning Catalyst
     [Paper] · arXiv:2505.14116
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

177. Self-rewarding correction for mathematical reasoning
     [Paper] · arXiv:2502.19613
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

178. Self-Rewarding Language Models
     [Paper] · arXiv:2401.10020
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

179. Self-Rewarding Large Vision-Language Models for Optimizing Prompts in Text-to-Image Generation
     [Paper] · arXiv:2505.16763
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

180. Self-Rewarding PPO: Aligning Large Language Models with Demonstrations Only
     [Paper] · arXiv:2510.21090
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

181. Self-Rewarding Rubric-Based Reinforcement Learning for Open-Ended Reasoning
     [Paper] · arXiv:2509.25534
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

182. Self-Taught Evaluators
     [Paper] · arXiv:2408.02666
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

183. Self-Trained Verification for Training- and Test-Time Self-Improvement
     [Paper] · arXiv:2605.30290
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

184. Self-Training Large Language Models for Improved Visual Program Synthesis With Visual Reinforcement
     [Paper] · arXiv:2404.04627
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

185. Semi-supervised Instruction Tuning for Large Language Models on Text-Attributed Graphs
     [Paper] · arXiv:2601.12807
     
     Taxonomy: L1. Updated object(s): 5.2 Adapter.

186. SeRL: Self-Play Reinforcement Learning for Large Language Models with Limited Data
     [Paper] · arXiv:2505.20347
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

187. Skill-Conditioned Gated Self-Distillation for LLM Reasoning
     [Paper] · arXiv:2605.28791
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 2.3 Procedural Memory; 6.1 Objective.

188. Skill-Guided Continuation Distillation for GUI Agents
     [Paper] · arXiv:2606.18890
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

189. SmartSnap: Proactive Evidence Seeking for Self-Verifying Agents
     [Paper] · arXiv:2512.22322
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

190. SocraticPO: Policy Optimization via Interactive Guidance
     [Paper] · arXiv:2606.09887
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

191. Soft-NBCE: Entropy-Weighted Chunk Fusion for Long-Context
     [Paper] · arXiv:2606.01101
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

192. SPPD: Self-training with Process Preference Learning Using Dynamic Value Margin
     [Paper] · arXiv:2502.13516
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

193. SRPO: Self-Referential Policy Optimization for Vision-Language-Action Models
     [Paper] · arXiv:2511.15605
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 7.2 Reward / Fitness.

194. STaR: Self-Taught Reasoner—Bootstrapping Reasoning With Reasoning
     [Paper] · arXiv:2203.14465
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

195. STEP: Enhancing Video-LLMs' Compositional Reasoning by Spatio-Temporal Graph-guided Self-Training
     [Paper] · arXiv:2412.00161
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

196. StepPRM-RTL: Stepwise Process-Reward Guided LLM Fine-Tuning for Enhanced RTL Synthesis
     [Paper] · arXiv:2606.04246
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 7.2 Reward / Fitness.

197. Stratagem: Learning Transferable Reasoning via Trajectory-Modulated Game Self-Play
     [Paper] · arXiv:2604.17696
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

198. Structured Reasoning for Large Language Models
     [Paper] · arXiv:2601.07180
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

199. Synergizing Discriminative Exemplars and Self-Refined Experience for MLLM-based In-Context Learning in Medical Diagnosis
     [Paper] · arXiv:2603.27737
     
     Taxonomy: L1. Updated object(s): 1.4 Prompt & Context; 2.3 Procedural Memory.

200. Synthetic Computers at Scale for Long-Horizon Productivity Simulation
     [Paper] · arXiv:2604.28181
     
     Taxonomy: L1. Updated object(s): 4.4 Skill Library; 8.3 Environment.

201. Teach-to-Reason: Competition-Guided Reasoning with a Self-Improving Teacher
     [Paper] · arXiv:2606.25407
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

202. Teaching Large Reasoning Models Effective Reflection
     [Paper] · arXiv:2601.12720
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

203. Team-Based Self-Play With Dual Adaptive Weighting for Fine-Tuning LLMs
     [Paper] · arXiv:2605.09922
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

204. Temporal Self-Rewarding Language Models: Decoupling Chosen-Rejected via Past-Future
     [Paper] · arXiv:2508.06026
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 7.1 Judge.

205. The Attacker in the Mirror: Breaking Self-Consistency in Safety via Anchored Bipolicy Self-Play
     [Paper] · arXiv:2605.08427
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

206. Think Longer to Explore Deeper: Learn to Explore In-Context via Length-Incentivized Reinforcement Learning
     [Paper] · arXiv:2602.11748
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

207. Thinking Before Running! Efficient Code Generation with Thorough Exploration and Optimal Refinement
     [Paper] · arXiv:2502.17442
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

208. Training-Free Test-Time Contrastive Learning for Large Language Models
     [Paper] · arXiv:2604.13552
     
     Taxonomy: L1. Updated object(s): 2.3 Procedural Memory; 1.5 Context.

209. Transformers Provably Learn Chain-of-Thought Reasoning with Length Generalization
     [Paper] · arXiv:2511.07378
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

210. Transitivity Meets Cyclicity: Explicit Preference Decomposition for Dynamic Large Language Model Alignment
     [Paper] · arXiv:2605.17342
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 7.2 Reward / Fitness.

211. TRON: Targeted Rule-Verifiable Online Environments for Visual Reasoning RL
     [Paper] · arXiv:2606.01599
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

212. Unsupervised Process Reward Models
     [Paper] · arXiv:2605.10158
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 7.1 Judge.

213. Utilizing and Calibrating Hindsight Process Rewards via Reinforcement with Mutual Information Self-Evaluation
     [Paper] · arXiv:2604.11611
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

214. VeriGate: Verifier-Gated Step-Level Supervision for GRPO
     [Paper] · arXiv:2605.30451
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 6.1 Objective.

215. VibeThinker-3B: Exploring the Frontier of Verifiable Reasoning in Small Language Models
     [Paper] · arXiv:2606.16140
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

216. VideoSAVi: Self-Aligned Video Language Models without Human Supervision
     [Paper] · arXiv:2412.00624
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights.

217. When Good Verifiers Go Bad: Self-Improving VLMs Can Regress on New Tasks
     [Paper] · arXiv:2606.14629
     
     Taxonomy: L1. Updated object(s): 5.2 Adapter.

218. World Model Self-Distillation: Training World Models to Solve General Tasks
     [Paper] · arXiv:2606.12072
     
     Taxonomy: L1. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

219. ZAS-SQL: Distilling Rules from Failures for Zero-Shot Text-to-SQL
     [Paper] · arXiv:2606.08245
     
     Taxonomy: L1. Updated object(s): 1.3 Prompt & Context; 3.4 Harness / Workflow.

Table-derived extensions

The following additions are explicitly listed in Table 4, which groups representative L2 systems by the object of improvement-strategy search.

1. Multimodal Prompt Optimization: Why Not Leverage Multiple Modalities for MLLMs
   [Paper] · arXiv:2510.09201
   
   Survey-table rationale: Evaluation-derived semantic feedback guides multimodal prompt candidates under a fixed task metric. Updated object(s): 1.1 Instruction; 1.2 Task Prompt / Template.

2. Agentic Neural Architecture Search
   [Paper] · arXiv:2607.07984
   
   Survey-table rationale: An LLM constructs a task-specific architecture search space, but the validation metric remains fixed. Updated object(s): 5.3 Architecture.

3. Auto Research with Specialist Agents Develops Effective and Non-Trivial Training Recipes
   [Paper] · arXiv:2605.05724
   
   Survey-table rationale: Specialist agents search over training recipes with a shared lineage under an external evaluator. Updated object(s): 6.5 Search / Meta-optimization Procedure.

4. AutoKernel: Autonomous GPU Kernel Optimization via Iterative Agent-Driven Search
   [Paper] · arXiv:2603.21331
   
   Survey-table rationale: A profile-rewrite-benchmark loop searches kernel implementations under fixed correctness and speed criteria. Updated object(s): 9.1 Program / Solution Code.

5. Adapting AlphaEvolve to Optimize Fully Homomorphic Encryption on TPUs
   [Paper] · arXiv:2605.14718
   
   Taxonomy: L2. Updated object(s): 9.1 Program Code.

6. AI-PROPELLER: Warehouse-Scale Interprocedural Code Layout Optimization with AlphaEvolve
   [Paper] · arXiv:2606.00131
   
   Taxonomy: L2. Updated object(s): 9.2 Algorithm.

7. Magellan: Autonomous Discovery of Novel Compiler Optimization Heuristics with AlphaEvolve
   [Paper] · arXiv:2601.21096
   
   Taxonomy: L2. Updated object(s): 9.1 Program Code.

8. Scientific Algorithm Discovery by Augmenting AlphaEvolve with Deep Research
   [Paper] · arXiv:2510.06056
   
   Taxonomy: L2. Updated object(s): 9.2 Algorithm; 2.1 Experience Memory.

9. A Self-Improving Coding Agent
   [Paper] · arXiv:2504.15228
   
   Taxonomy: L2. Updated object(s): 3.6 Scaffold Code; 4.3 Tool Code; 1.1 Instruction.

10. Adaptive Self-improvement LLM Agentic System for ML Library Development
    [Paper] · arXiv:2502.02534
    
    Taxonomy: L2. Updated object(s): 1.4 Prompt & Context; 1.5 Context.

11. Advanced For-Loop for QML algorithm search
    [Paper] · arXiv:2506.18260
    
    Taxonomy: L2. Updated object(s): 9.2 Algorithm; 9.1 Program Code.

12. AFlow: Automating Agentic Workflow Generation
    [Paper] · arXiv:2410.10762
    
    Taxonomy: L2. Updated object(s): 3.6 Scaffold Code; 1.2 Task Prompt.

13. AgentDisCo: Towards Disentanglement and Collaboration in Open-ended Deep Research Agents
    [Paper] · arXiv:2605.11732
    
    Taxonomy: L2. Updated object(s): 1.1 Instruction; 3.6 Scaffold Code.

14. Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models
    [Paper] · arXiv:2510.04618
    
    Taxonomy: L2. Updated object(s): 1.5 Context; 2.4 Memory & Knowledge.

15. Agentic Discovery of Neural Architectures: AIRA-Compose and AIRA-Design
    [Paper] · arXiv:2605.15871
    
    Taxonomy: L2. Updated object(s): 9.2 Algorithm; 9.3 External Artifact.

16. Agentic Harness Engineering: Observability-Driven Automatic Evolution of Coding-Agent Harnesses
    [Paper] · arXiv:2604.25850
    
    Taxonomy: L2. Updated object(s): 3.6 Scaffold Code; 4.3 Tool Code.

17. Agentic MIP Research: Accelerated Constraint Handler Generation
    [Paper] · arXiv:2605.09186
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 9.2 Algorithm.

18. AgenticRL: Self-Refining Agentic Reinforcement Learning for Vision-Conditioned UAV Navigation
    [Paper] · arXiv:2606.03963
    
    Taxonomy: L2. Updated object(s): 6.1 Objective; 5.1 Model Weights.

19. Ask, Don't Judge: Binary Questions for Interpretable LLM Evaluation and Self-Improvement
    [Paper] · arXiv:2606.27226
    
    Taxonomy: L2. Updated object(s): 1.1 Instruction.

20. ASPIRE: Agentic Skills Discovery for Robotics
    [Paper] · arXiv:2607.00272
    
    Taxonomy: L2. Updated object(s): 4.4 Skill Library; 9.1 Program Code.

21. Auto-Configuring Scientific Simulators with Lightweight Coding-Agent Adapters
    [Paper] · arXiv:2606.09774
    
    Taxonomy: L2. Updated object(s): 2.3 Procedural Memory; 1.1 Instruction; 4.4 Skill Library.

22. Automated Design of Agentic Systems (ADAS)
    [Paper] · arXiv:2408.08435
    
    Taxonomy: L2. Updated object(s): 3.6 Scaffold Code.

23. Automated Heuristic Design for Unit Commitment Using Large Language Models
    [Paper] · arXiv:2506.12495
    
    Taxonomy: L2. Updated object(s): 9.2 Algorithm; 9.1 Program Code.

24. AutoMegaKernel: A Statically-Checked Agent Harness for Self-Retargeting Megakernel Synthesis
    [Paper] · arXiv:2606.09682
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code.

25. AutoPass: Evidence-Guided LLM Agents for Compiler Performance Tuning
    [Paper] · arXiv:2606.20373
    
    Taxonomy: L2. Updated object(s): 9.2 Algorithm.

26. AutoPrompt: Eliciting Knowledge from Language Models with Automatically Generated Prompts
    [Paper] · arXiv:2010.15980
    
    Taxonomy: L2. Updated object(s): 1.2 Task Prompt.

27. AutoPyVerifier: Learning Compact Executable Verifiers for Large Language Model Outputs
    [Paper] · arXiv:2604.22937
    
    Taxonomy: L2. Updated object(s): 7.3 Verifier; 7.1 Judge.

28. AutoVQA-G: Self-Improving Agentic Framework for Automated Visual Question Answering and Grounding Annotation
    [Paper] · arXiv:2604.17488
    
    Taxonomy: L2. Updated object(s): 1.2 Task Prompt; 1.1 Instruction.

29. Beyond Algorithm Evolution: An LLM-Driven Framework for the Co-Evolution of Swarm Intelligence Optimization Algorithms and Prompts
    [Paper] · arXiv:2512.09209
    
    Taxonomy: L2. Updated object(s): 9.2 Algorithm; 1.1 Instruction.

30. Beyond Scalar Rewards: Dense Feedback for LLM Policy Synthesis in Sequential Social Dilemmas
    [Paper] · arXiv:2603.19453
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code.

31. Beyond Static Tools: Test-Time Tool Evolution for Scientific Reasoning
    [Paper] · arXiv:2601.07641
    
    Taxonomy: L2. Updated object(s): 4.1 Tool Set; 4.2 Tool Interface.

32. BT-APE: A Computationally Light Backtracking Approach to Automatic Prompt Engineering for Requirements Classification
    [Paper] · arXiv:2607.00427
    
    Taxonomy: L2. Updated object(s): 1.2 Task Prompt.

33. CausalEvolve: Towards Open-Ended Discovery with Causal Scratchpad
    [Paper] · arXiv:2603.14575
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 2.1 Experience Memory.

34. CHIA: An open-source framework for principled, agentic AI-driven hardware/software co-design research
    [Paper] · arXiv:2606.27350
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code.

35. Co-evolving Agent Architectures and Interpretable Reasoning for Automated Optimization
    [Paper] · arXiv:2604.17708
    
    Taxonomy: L2. Updated object(s): 3.1 Workflow; 2.2 Knowledge Memory.

36. CodeEvolve: an open source evolutionary coding agent for algorithmic discovery and optimization
    [Paper] · arXiv:2510.14150
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 1.2 Task Prompt.

37. CoEvoSkills: Self-Evolving Agent Skills via Co-Evolutionary Verification
    [Paper] · arXiv:2604.01687
    
    Taxonomy: L2. Updated object(s): 4.4 Skill Library.

38. Combee: Scaling Prompt Learning for Self-Improving Language Model Agents
    [Paper] · arXiv:2604.04247
    
    Taxonomy: L2. Updated object(s): 1.1 Instruction; 1.5 Context.

39. Combining Large Language Models and Gradient-Free Optimization for Automatic Control Policy Synthesis
    [Paper] · arXiv:2510.00373
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code.

40. Competing with AI Scientists: Agent-Driven Approach to Astrophysics Research
    [Paper] · arXiv:2604.09621
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code.

41. Continually self-improving AI
    [Paper] · arXiv:2603.18073
    
    Taxonomy: L2. Updated object(s): 6.5 Meta-optimization; 6.1 Objective; 5.1 Model Weights.

42. CORAL: Towards Autonomous Multi-Agent Evolution for Open-Ended Discovery
    [Paper] · arXiv:2604.01658
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 4.4 Skill Library.

43. DataEvolver: Automatic Data Preparation for Large Language Models through Multi-Level Self-Evolving
    [Paper] · arXiv:2606.07001
    
    Taxonomy: L2. Updated object(s): 8.1 Experience Data; 9.1 Program Code.

44. Defining Cost Function of Steganography with Large Language Models
    [Paper] · arXiv:2512.09769
    
    Taxonomy: L2. Updated object(s): 9.2 Algorithm.

45. DeltaEvolve: Accelerating Scientific Discovery through Momentum-Driven Evolution
    [Paper] · arXiv:2602.02919
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code.

46. Discovering Multiagent Learning Algorithms with Large Language Models
    [Paper] · arXiv:2602.16928
    
    Taxonomy: L2. Updated object(s): 9.2 Algorithm.

47. DrugSAGE: Self-evolving Agent Experience for Efficient State-of-the-Art Drug Discovery
    [Paper] · arXiv:2605.15461
    
    Taxonomy: L2. Updated object(s): 4.4 Skill Library; 2.2 Knowledge Memory.

48. Each Judge Its Own Yardstick: Discovering Per-VLM Taxonomies for Physical Video Evaluation
    [Paper] · arXiv:2606.22918
    
    Taxonomy: L2. Updated object(s): 7.1 Judge.

49. EEVEE: Towards Test-time Prompt Learning in the Real World for Self-Improving Agents
    [Paper] · arXiv:2606.11182
    
    Taxonomy: L2. Updated object(s): 1.1 Instruction; 3.3 Harness / Workflow.

50. Effective Harness Engineering for Algorithm Discovery with Coding Agents
    [Paper] · arXiv:2605.15221
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 9.2 Algorithm.

51. ELMES+: Automated Construction of Fine-Grained Evaluation Rubrics for Large Language Models in Long-Tail Educational Scenarios
    [Paper] · arXiv:2606.06546
    
    Taxonomy: L2. Updated object(s): 7.1 Judge; 8.2 Task Generator.

52. EmbodiSkill: Skill-Aware Reflection for Self-Evolving Embodied Agents
    [Paper] · arXiv:2605.10332
    
    Taxonomy: L2. Updated object(s): 4.4 Skill Library.

53. Enhancing Operational Safety via Agentic Dialogue Hazard Identification Analysis
    [Paper] · arXiv:2606.03812
    
    Taxonomy: L2. Updated object(s): 3.5 Multi-agent; 1.1 Instruction.

54. Environment-Grounded Automated Prompt Optimization for LLM Game Agents
    [Paper] · arXiv:2606.17838
    
    Taxonomy: L2. Updated object(s): 1.1 Instruction; 1.2 Task Prompt.

55. EpiAgent: An Agent-Centric System for Ancient Inscription Restoration
    [Paper] · arXiv:2604.09367
    
    Taxonomy: L2. Updated object(s): 2.1 Experience Memory; 2.3 Procedural Memory.

56. Eureka: Human-Level Reward Design via Coding Large Language Models
    [Paper] · arXiv:2310.12931
    
    Taxonomy: L2. Updated object(s): 7.2 Reward / Fitness.

57. Eureka: Intelligent Feature Engineering for Enterprise AI Cloud Resource Demand Prediction
    [Paper] · arXiv:2605.25297
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 5.1 Model Weights.

58. EvoClinician: A Self-Evolving Agent for Multi-Turn Medical Diagnosis via Test-Time Evolutionary Learning
    [Paper] · arXiv:2601.22964
    
    Taxonomy: L2. Updated object(s): 1.1 Instruction; 2.1 Experience Memory.

59. EvoLattice: Persistent Internal-Population Evolution through Multi-Alternative Quality-Diversity Graph Representations for LLM-Guided Program Discovery
    [Paper] · arXiv:2512.13857
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 9.2 Algorithm.

60. Evolution of Heuristics: Towards Efficient Automatic Algorithm Design Using Large Language Model
    [Paper] · arXiv:2401.02051
    
    Taxonomy: L2. Updated object(s): 9.2 Algorithm; 9.1 Program Code.

61. EvolveSignal: A Large Language Model Powered Coding Agent for Discovering Traffic Signal Control Strategies
    [Paper] · arXiv:2509.03335
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code.

62. Evolving Deception: When Agents Evolve, Deception Wins
    [Paper] · arXiv:2603.05872
    
    Taxonomy: L2. Updated object(s): 1.1 Instruction.

63. Evolving Medical Imaging Agents via Experience-Driven Self-Skill Discovery
    [Paper] · arXiv:2603.05860
    
    Taxonomy: L2. Updated object(s): 4.4 Skill Library; 5.1 Model Weights; 2.1 Experience Memory.

64. EVOM: Agentic Meta-Evolution of Actor-Critic Architectures for Reinforcement Learning
    [Paper] · arXiv:2606.26327
    
    Taxonomy: L2. Updated object(s): 5.3 Architecture.

65. EvoPrompt: Connecting LLMs with Evolutionary Algorithms Yields Powerful Prompt Optimizers
    [Paper] · arXiv:2309.08532
    
    Taxonomy: L2. Updated object(s): 1.2 Task Prompt.

66. EvoScientist: Towards Multi-Agent Evolving AI Scientists for End-to-End Scientific Discovery
    [Paper] · arXiv:2603.08127
    
    Taxonomy: L2. Updated object(s): 2.2 Knowledge Memory; 1.5 Context.

67. EvoTest: Evolutionary Test-Time Learning for Self-Improving Agentic Systems
    [Paper] · arXiv:2510.13220
    
    Taxonomy: L2. Updated object(s): 1.1 Instruction; 2.1 Experience Memory; 5.4 Model.

68. Experience Graphs: The Data Foundation for Self-Improving Agents
    [Paper] · arXiv:2606.29823
    
    Taxonomy: L2. Updated object(s): 2.1 Experience Memory; 2.4 Memory & Knowledge.

69. Experience Makes Skillful: Enabling Generalizable Medical Agent Reasoning via Self-Evolving Skill Memory
    [Paper] · arXiv:2606.09365
    
    Taxonomy: L2. Updated object(s): 2.3 Procedural Memory; 2.5 Memory & Knowledge.

70. ExpGraph: Model-Agnostic Experience Learning with Graph-Structured Memory for LLM Agents
    [Paper] · arXiv:2605.30712
    
    Taxonomy: L2. Updated object(s): 2.1 Experience Memory; 2.4 Memory & Knowledge; 5.2 Adapter.

71. FederatedSkill: Federated Learning for Agentic Skill Evolution
    [Paper] · arXiv:2606.03143
    
    Taxonomy: L2. Updated object(s): 4.4 Skill Library; 2.2 Knowledge Memory.

72. Feedback Over Form: Why Execution Feedback Matters More Than Pipeline Topology in 1-3B Code Generation
    [Paper] · arXiv:2604.21950
    
    Taxonomy: L2. Updated object(s): 3.1 Workflow.

73. Feedback-to-Rubrics: Can We Learn Expert Criteria from Inline Comments?
    [Paper] · arXiv:2605.29857
    
    Taxonomy: L2. Updated object(s): 7.1 Judge; 1.1 Instruction.

74. From AI Assistant to AI Scientist: Autonomous Discovery of LLM-RL Algorithms with LLM Agents
    [Paper] · arXiv:2603.23951
    
    Taxonomy: L2. Updated object(s): 6.1 Objective; 6.2 Trainer / Optimization.

75. From Failed Trajectories to Reliable LLM Agents: Diagnosing and Repairing Harness Flaws
    [Paper] · arXiv:2606.06324
    
    Taxonomy: L2. Updated object(s): 3.6 Scaffold Code; 1.2 Task Prompt; 4.2 Tool Interface.

76. From Self-Evolving Synthetic Data to Verifiable-Reward RL: Post-Training Multi-turn Interactive Tool-Using Agents
    [Paper] · arXiv:2601.22607
    
    Taxonomy: L2. Updated object(s): 8.1 Experience Data; 5.1 Model Weights.

77. From Understanding to Excelling: Template-Free Algorithm Design through Structural-Functional Co-Evolution
    [Paper] · arXiv:2503.10721
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 9.2 Algorithm; 6.5 Meta-optimization.

78. From User Interface to Agent Interface: Efficiency Optimization of UI Representations for LLM Agents
    [Paper] · arXiv:2512.13438
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code.

79. FuzzAgent: Multi-Agent System for Evolutionary Library Fuzzing
    [Paper] · arXiv:2605.14431
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code.

80. GeoEvolve: Automating Geospatial Model Discovery via Multi-Agent Large Language Models
    [Paper] · arXiv:2509.21593
    
    Taxonomy: L2. Updated object(s): 9.2 Algorithm.

81. GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning
    [Paper] · arXiv:2507.19457
    
    Taxonomy: L2. Updated object(s): 1.1 Instruction.

82. GigaEvo: An Open Source Optimization Framework Powered By LLMs And Evolution Algorithms
    [Paper] · arXiv:2511.17592
    
    Taxonomy: L2. Updated object(s): 9.2 Algorithm; 9.1 Program Code; 1.1 Instruction.

83. Hardening Agent Benchmarks with Adversarial Hacker-Fixer Loops
    [Paper] · arXiv:2606.08960
    
    Taxonomy: L2. Updated object(s): 7.3 Verifier; 8.3 Environment.

84. Healthcare Mechanisms from Policy-as-Code Search under Strategic Provider Response
    [Paper] · arXiv:2605.30680
    
    Taxonomy: L2. Updated object(s): 9.2 Algorithm.

85. Heuresis: Search Strategies for Autonomous AI Research Agents Across Quality, Diversity and Novelty
    [Paper] · arXiv:2606.25198
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 9.2 Algorithm.

86. Hierarchical Experimentalist Agents
    [Paper] · arXiv:2606.29315
    
    Taxonomy: L2. Updated object(s): 4.4 Skill Library.

87. HSEvo: Elevating Automatic Heuristic Design with Diversity-Driven Harmony Search and Genetic Algorithm Using LLMs
    [Paper] · arXiv:2412.14995
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 6.5 Meta-optimization.

88. HypoForge: A Self-Improving Multi-Agent Framework for Automated Hypothesis Generation and Testing via Scientific Skill Learning
    [Paper] · arXiv:2608.25770
    
    Taxonomy: L2. Updated object(s): 4.4 Skill Library.

89. ImprovEvolve: Basin-Hopping Meets LLM-Guided Evolutionary Search
    [Paper] · arXiv:2602.10233
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 9.2 Algorithm.

90. InvEvolve: Evolving White-Box Inventory Policies via Large Language Models with Performance Guarantees
    [Paper] · arXiv:2605.00369
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 5.1 Model Weights.

91. ISM: Self-Improving Strategy Memory for Continual Mathematical Reasoning
    [Paper] · arXiv:2606.31191
    
    Taxonomy: L2. Updated object(s): 2.3 Procedural Memory; 2.4 Memory & Knowledge.

92. IterInject: Indirect Prompt Injection Against LLM Agents via Feedback-Guided Iterative Optimization
    [Paper] · arXiv:2605.24659
    
    Taxonomy: L2. Updated object(s): 1.2 Task Prompt.

93. Kernel Foundry: A Diagnosis-driven Evolutionary Kernel Optimizer with Multi-Experts
    [Paper] · arXiv:2605.30359
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 2.3 Procedural Memory.

94. Large Language Models Are Human-Level Prompt Engineers (Automatic Prompt Engineer, APE)
    [Paper] · arXiv:2211.01910
    
    Taxonomy: L2. Updated object(s): 1.1 Instruction.

95. Large Language Models as Optimizers
    [Paper] · arXiv:2309.03409
    
    Taxonomy: L2. Updated object(s): 1.2 Task Prompt.

96. Learning from Failure: Inference-Time Self-Improvement for Computer-Use Agents
    [Paper] · arXiv:2606.31270
    
    Taxonomy: L2. Updated object(s): 3.6 Scaffold Code; 4.1 Tool Set; 3.4 Harness / Workflow.

97. Learning to Choose: An Empowerment-Guided Multi-Agent System with semantic communication for Adaptive Method Selection
    [Paper] · arXiv:2605.30042
    
    Taxonomy: L2. Updated object(s): 3.3 Harness / Workflow; 2.1 Experience Memory.

98. LEVI: Stronger Search Architectures Can Substitute for Larger LLMs in Evolutionary Search
    [Paper] · arXiv:2605.09764
    
    Taxonomy: L2. Updated object(s): 9.1 Program Code; 1.2 Task Prompt.

99. Libra: Training the Environment for Agentic Information Retrieval
    [Paper] · arXiv:2607.00016
    
    Taxonomy: L2. Updated object(s): 2.2 Knowledge Memory; 8.3 Environment.

100. LLaMEA-BO: A Large Language Model Evolutionary Algorithm for Automatically Generating Bayesian Optimization Algorithms
     [Paper] · arXiv:2505.21034
     
     Taxonomy: L2. Updated object(s): 9.2 Algorithm; 9.1 Program Code.

101. LLM as a Tool, Not an Agent: Code-Mined Tree Transformations for Neural Architecture Search
     [Paper] · arXiv:2604.16555
     
     Taxonomy: L2. Updated object(s): 9.3 External Artifact.

102. LLM-Evolved Pattern Generators for Optimal Classical Planning
     [Paper] · arXiv:2606.02438
     
     Taxonomy: L2. Updated object(s): 9.1 Program Code.

103. LLM-Guided Evolutionary Program Synthesis for Quasi-Monte Carlo Design
     [Paper] · arXiv:2510.03650
     
     Taxonomy: L2. Updated object(s): 9.2 Algorithm.

104. LLM-Guided Search for Deletion-Correcting Codes
     [Paper] · arXiv:2504.00613
     
     Taxonomy: L2. Updated object(s): 9.2 Algorithm.

105. LoongFlow: Directed Evolutionary Search via a Cognitive Plan-Execute-Summarize Paradigm
     [Paper] · arXiv:2512.24077
     
     Taxonomy: L2. Updated object(s): 9.2 Algorithm; 9.1 Program Code.

106. Mathematical exploration and discovery at scale
     [Paper] · arXiv:2511.02864
     
     Taxonomy: L2. Updated object(s): 9.1 Program Code; 9.2 Algorithm.

107. Mechanistically Eliciting Latent Behaviors in Language Models
     [Paper] · arXiv:2606.29604
     
     Taxonomy: L2. Updated object(s): 5.2 Adapter.

108. Medical Heuristic Learning: An LLM-Driven Framework for Interpretable and Auditable Clinical Decision Rules
     [Paper] · arXiv:2606.16337
     
     Taxonomy: L2. Updated object(s): 9.1 Program Code.

109. MEMENTO: Leveraging Web as a Learning Signal for Low-Data Domains
     [Paper] · arXiv:2605.29795
     
     Taxonomy: L2. Updated object(s): 2.3 Procedural Memory; 2.2 Knowledge Memory.

110. MEMO: Memory-Augmented Model Context Optimization for Robust Multi-Turn Multi-Agent LLM Games
     [Paper] · arXiv:2603.09022
     
     Taxonomy: L2. Updated object(s): 1.2 Task Prompt; 2.2 Knowledge Memory.

111. Meta-Harness: End-to-End Optimization of Model Harnesses
     [Paper] · arXiv:2603.28052
     
     Taxonomy: L2. Updated object(s): 3.6 Scaffold Code.

112. Middo: Model-Informed Dynamic Data Optimization for Enhanced LLM Fine-Tuning via Closed-Loop Learning
     [Paper] · arXiv:2508.21589
     
     Taxonomy: L2. Updated object(s): 8.1 Experience Data; 5.1 Model Weights.

113. Mining Generalizable Activation Functions
     [Paper] · arXiv:2602.05688
     
     Taxonomy: L2. Updated object(s): 9.2 Algorithm.

114. MLEvolve: A Self-Evolving Framework for Automated Machine Learning Algorithm Discovery
     [Paper] · arXiv:2606.06473
     
     Taxonomy: L2. Updated object(s): 9.1 Program Code; 9.2 Algorithm.

115. MobEvolve: An Agentic Self-Evolving Heuristic System for Interpretable Human Mobility Generation
     [Paper] · arXiv:2606.01640
     
     Taxonomy: L2. Updated object(s): 9.1 Program Code; 2.1 Experience Memory.

116. OpenComputer: Verifiable Software Worlds for Computer-Use Agents
     [Paper] · arXiv:2605.19769
     
     Taxonomy: L2. Updated object(s): 7.3 Verifier; 8.1 Experience Data; 8.3 Environment.

117. Organize then Retrieve: Hierarchical Memory Navigation for Efficient Agents
     [Paper] · arXiv:2606.11680
     
     Taxonomy: L2. Updated object(s): 4.4 Skill Library; 5.1 Model Weights.

118. PACEvolve++: Improving Test-time Learning for Evolutionary Search Agents
     [Paper] · arXiv:2605.07039
     
     Taxonomy: L2. Updated object(s): 5.1 Model Weights; 9.1 Program Code.

119. Persona Generators: Generating Diverse Synthetic Personas for Arbitrary Contexts
     [Paper] · arXiv:2602.03545
     
     Taxonomy: L2. Updated object(s): 9.1 Program Code.

120. Polaris: A Gödel Agent Framework for Small Language Models through Experience-Abstracted Policy Repair
     [Paper] · arXiv:2603.23129
     
     Taxonomy: L2. Updated object(s): 3.6 Scaffold Code; 2.3 Procedural Memory.

121. ProfiLLM: Utility-Aligned Agentic User Profiling for Industrial Ride-Hailing Dispatch
     [Paper] · arXiv:2606.18803
     
     Taxonomy: L2. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

122. PromptAgent: Strategic Planning with Language Models Enables Expert-level Prompt Optimization
     [Paper] · arXiv:2310.16427
     
     Taxonomy: L2. Updated object(s): 1.2 Task Prompt.

123. PromptWizard: Task-Aware Prompt Optimization Framework
     [Paper] · arXiv:2405.18369
     
     Taxonomy: L2. Updated object(s): 1.1 Instruction; 1.4 Prompt & Context.

124. ProTeGi: Automatic Prompt Optimization with 'Gradient Descent' and Beam Search
     [Paper] · arXiv:2305.03495
     
     Taxonomy: L2. Updated object(s): 1.2 Task Prompt.

125. PYTHALAB-MERA: Validation-Grounded Memory, Retrieval, and Acceptance Control for Frozen-LLM Coding Agents
     [Paper] · arXiv:2605.08468
     
     Taxonomy: L2. Updated object(s): 2.1 Experience Memory; 1.5 Context; 4.4 Skill Library.

126. QUBE: Enhancing Automatic Heuristic Design via Quality-Uncertainty Balanced Evolution
     [Paper] · arXiv:2412.20694
     
     Taxonomy: L2. Updated object(s): 9.1 Program Code; 6.5 Meta-optimization.

127. QueenBee Planner: Skill-Evolving Communication Topologies for Token-Efficient LLM Multi-Agent Systems
     [Paper] · arXiv:2606.27492
     
     Taxonomy: L2. Updated object(s): 2.3 Procedural Memory; 3.5 Multi-agent.

128. RankEvolve: Automating the Discovery of Retrieval Algorithms via LLM-Driven Evolution
     [Paper] · arXiv:2602.16932
     
     Taxonomy: L2. Updated object(s): 9.2 Algorithm.

129. Recursive Harness Self-Improvement
     [Paper] · arXiv:2607.15524
     
     Taxonomy: L2. Updated object(s): 3.1 Workflow; 3.5 Multi-agent.

130. Reflection-Based Task Adaptation for Self-Improving VLA
     [Paper] · arXiv:2510.12710
     
     Taxonomy: L2. Updated object(s): 5.1 Model Weights; 7.2 Reward / Fitness; 6.4 Curriculum.

131. REFLEX: Reflective Evolution from LLM Experience
     [Paper] · arXiv:2606.16496
     
     Taxonomy: L2. Updated object(s): 9.1 Program Code; 4.4 Skill Library.

132. Repeated post-training is not Self-improving: Diagnosing Scientific Amnesia in Continual DPO Pipelines
     [Paper] · arXiv:2606.21089
     
     Taxonomy: L2. Updated object(s): 5.1 Model Weights; 6.5 Meta-optimization.

133. Rethinking Code Similarity for Automated Algorithm Design with LLMs
     [Paper] · arXiv:2603.02787
     
     Taxonomy: L2. Updated object(s): 7.2 Reward / Fitness; 3.6 Scaffold Code.

134. Revisiting OPRO: The Limitations of Small-Scale LLMs as Optimizers
     [Paper] · arXiv:2405.10276
     
     Taxonomy: L2. Updated object(s): 1.2 Task Prompt.

135. RLPrompt: Optimizing Discrete Text Prompts with Reinforcement Learning
     [Paper] · arXiv:2205.12548
     
     Taxonomy: L2. Updated object(s): 1.2 Task Prompt.

136. S1-NexusAgent: a Self-Evolving Agent Framework for Multidisciplinary Scientific Research
     [Paper] · arXiv:2602.01550
     
     Taxonomy: L2. Updated object(s): 4.4 Skill Library.

137. Self-Consolidating Language Models: Continual Knowledge Incorporation from Context
     [Paper] · arXiv:2605.07076
     
     Taxonomy: L2. Updated object(s): 5.1 Model Weights; 5.2 Adapter.

138. Self-Harness: Harnesses That Improve Themselves
     [Paper] · arXiv:2606.09498
     
     Taxonomy: L2. Updated object(s): 3.6 Scaffold Code; 1.1 Instruction.

139. Self-Improvement Can Self-Regress: The Rise-and-Collapse Failure Mode of LLM Self-Training
     [Paper] · arXiv:2606.21090
     
     Taxonomy: L2. Updated object(s): 5.1 Model Weights; 6.5 Meta-optimization.

140. SePO: Self-Evolving Prompt Agent for System Prompt Optimization
     [Paper] · arXiv:2606.04465
     
     Taxonomy: L2. Updated object(s): 1.2 Task Prompt; 1.1 Instruction.

141. SGM: A Statistical Godel Machine for Risk-Controlled Recursive Self-Modification
     [Paper] · arXiv:2510.10232
     
     Taxonomy: L2. Updated object(s): 6.3 Train Schedule.

142. SHARP: A Self-Evolving Human-Auditable Rubric Policy for Financial Trading Agents
     [Paper] · arXiv:2605.06822
     
     Taxonomy: L2. Updated object(s): 1.1 Instruction.

143. SIA: Self Improving AI with Harness & Weight Updates
     [Paper] · arXiv:2605.27276
     
     Taxonomy: L2. Updated object(s): 10.2 Full-system / Co-evolution; 3.6 Scaffold Code; 5.1 Model Weights.

144. SignalClaw: LLM-Guided Evolutionary Synthesis of Interpretable Traffic Signal Control Skills
     [Paper] · arXiv:2604.05535
     
     Taxonomy: L2. Updated object(s): 9.1 Program Code; 4.4 Skill Library.

145. Skill-R1: Agent Skill Evolution via Reinforcement Learning
     [Paper] · arXiv:2605.09359
     
     Taxonomy: L2. Updated object(s): 4.4 Skill Library; 5.1 Model Weights.

146. SkillCoach: Self-Evolving Rubrics for Evaluating and Enhancing Agentic Skill-Use
     [Paper] · arXiv:2607.01874
     
     Taxonomy: L2. Updated object(s): 7.1 Judge; 5.1 Model Weights; 6.4 Curriculum.

147. SkillMaster: Toward Autonomous Skill Mastery in LLM Agents
     [Paper] · arXiv:2605.08693
     
     Taxonomy: L2. Updated object(s): 4.4 Skill Library; 5.1 Model Weights.

148. SkillRevise: Improving LLM-Authored Agent Skills via Trace-Conditioned Skill Revision
     [Paper] · arXiv:2606.01139
     
     Taxonomy: L2. Updated object(s): 4.4 Skill Library.

149. SkillSmith: Co-Evolving Skills and Tools for Self-Improving Agent Systems
     [Paper] · arXiv:2606.01314
     
     Taxonomy: L2. Updated object(s): 4.4 Skill Library; 4.3 Tool Code; 2.1 Experience Memory.

150. SOCIA-EVO: Automated Simulator Construction via Dual-Anchored Bi-Level Optimization
     [Paper] · arXiv:2604.17351
     
     Taxonomy: L2. Updated object(s): 2.2 Knowledge Memory; 9.1 Program Code.

151. Support Vector Rubrics: Closing the Gap Between Self-Generated and Human Rubrics
     [Paper] · arXiv:2606.08077
     
     Taxonomy: L2. Updated object(s): 7.1 Judge; 1.3 Prompt & Context.

152. Taming I2V models for Image HOI Editing: A Cognitive Benchmark and Agentic Self-Correcting Framework
     [Paper] · arXiv:2606.19073
     
     Taxonomy: L2. Updated object(s): 2.2 Knowledge Memory; 1.2 Task Prompt.

153. The Time is Here for Just-in-Time Systems: Challenges and Opportunities
     [Paper] · arXiv:2605.24096
     
     Taxonomy: L2. Updated object(s): 9.1 Program Code.

154. Towards AGI A Pragmatic Approach Towards Self Evolving Agent
     [Paper] · arXiv:2601.11658
     
     Taxonomy: L2. Updated object(s): 4.1 Tool Set; 5.1 Model Weights.

155. Trajectory-Informed Memory Generation for Self-Improving Agent Systems
     [Paper] · arXiv:2603.10600
     
     Taxonomy: L2. Updated object(s): 2.3 Procedural Memory; 1.5 Context.

156. Understanding the Challenges in Iterative Generative Optimization with LLMs
     [Paper] · arXiv:2603.23994
     
     Taxonomy: L2. Updated object(s): 9.1 Program Code; 1.2 Task Prompt; 3.6 Scaffold Code.

157. Unifying Temporal and Structural Credit Assignment in LLM-Based Multi-Agent Prompt Optimization
     [Paper] · arXiv:2605.30227
     
     Taxonomy: L2. Updated object(s): 1.1 Instruction.

158. VASO: Formally Verifiable Self-Evolving Skills for Physical AI Agents
     [Paper] · arXiv:2606.05395
     
     Taxonomy: L2. Updated object(s): 4.4 Skill Library; 4.2 Tool Interface.

159. When LLM Reward Design Fails: Diagnostic-Driven Refinement for Sparse Structured RL
     [Paper] · arXiv:2605.28918
     
     Taxonomy: L2. Updated object(s): 9.1 Program Code.

Table-derived extension

Table 5 identifies the following learner-conditioned future-experience loop, complementary to the self-play and autonomous-practice systems already in this section.

1. SIMA 2: A Generalist Embodied Agent for Virtual Worlds
   [Paper] · arXiv:2512.04797
   
   Survey-table rationale: In the full ASKA setting, evaluation feedback directs practice toward weaker skills; later experience is conditioned on the learner's changing capability state. Updated object(s): 8.1 Training / Experience Data; 8.2 Task / Curriculum Generator; 8.3 Environment / Simulator.

2. Absolute Zero: Reinforced Self-play Reasoning with Zero Data
   [Paper] · arXiv:2505.03335
   
   Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

3. Active Zero: Self-Evolving Vision-Language Models through Active Environment Exploration
   [Paper] · arXiv:2602.11241
   
   Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

4. Agent0: Unleashing Self-Evolving Agents from Zero Data via Tool-Integrated Reasoning
   [Paper] · arXiv:2511.16043
   
   Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

5. ANCORA: Learning to Question via Manifold-Anchored Self-Play for Verifiable Reasoning
   [Paper] · arXiv:2604.27644
   
   Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator; 6.4 Curriculum.

6. ASH: Agents that Self-Hone via Embodied Learning
   [Paper] · arXiv:2605.14211
   
   Taxonomy: L3. Updated object(s): 5.1 Model Weights; 2.1 Experience Memory.

7. BenchEvolver: Frontier Task Synthesis via Solution-Centric Evolution
   [Paper] · arXiv:2606.01286
   
   Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

8. Better LLM Reasoning via Dual-Play
   [Paper] · arXiv:2511.11881
   
   Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

9. Close the Loop: Synthesizing Infinite Tool-Use Data via Multi-Agent Role-Playing
   [Paper] · arXiv:2512.23611
   
   Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

10. Concordia: Self-Improving Synthetic Tables for Federated LLMs
    [Paper] · arXiv:2605.09855
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

11. DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning
    [Paper] · arXiv:2505.19850
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

12. Disentangling Intent from Role: Adversarial Self-Play for Persona-Invariant Safety Alignment
    [Paper] · arXiv:2605.01899
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

13. DUEL: Adversarial Self-Play for Multimodal Reasoning
    [Paper] · arXiv:2605.24794
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

14. EnvGen: Generating and Adapting Environments via LLMs for Training Embodied Agents
    [Paper] · arXiv:2403.12014
    
    Taxonomy: L3. Updated object(s): 8.2 Task Generator; 5.1 Model Weights.

15. EnvHarness: Awakening Static Worlds for Agent Learning
    [Paper] · arXiv:2608.19880
    
    Taxonomy: L3. Updated object(s): 8.3 Environment; 4.4 Skill Library.

16. EvoCurr: Self-evolving Curriculum with Behavior Code Generation for Complex Decision-making
    [Paper] · arXiv:2508.09586
    
    Taxonomy: L3. Updated object(s): 8.2 Task Generator; 9.1 Program Code.

17. EvoLMM: Self-Evolving Large Multimodal Models with Continuous Rewards
    [Paper] · arXiv:2511.16672
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

18. EvoVid: Temporal-Centric Self-Evolution for Video Large Language Models
    [Paper] · arXiv:2605.21931
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

19. From Context to Skills: Can Language Models Learn from Context Skillfully?
    [Paper] · arXiv:2604.27660
    
    Taxonomy: L3. Updated object(s): 4.4 Skill Library; 8.2 Task Generator.

20. G-Zero: Self-Play for Open-Ended Generation from Zero Data
    [Paper] · arXiv:2605.09959
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

21. GASP: Guided Asymmetric Self-Play For Coding LLMs
    [Paper] · arXiv:2603.15957
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

22. GenEnv: Difficulty-Aligned Co-Evolution Between LLM Agents and Environment Simulators
    [Paper] · arXiv:2512.19682
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

23. GeoX: Mastering Geospatial Reasoning Through Self-Play and Verifiable Rewards
    [Paper] · arXiv:2605.20006
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

24. iReasoner: Trajectory-Aware Intrinsic Reasoning Supervision for Self-Evolving Large Multimodal Models
    [Paper] · arXiv:2601.05877
    
    Taxonomy: L3. Updated object(s): 5.2 Adapter; 6.1 Objective.

25. Knowledge-Graph Paths as Intermediate Supervision for Self-Evolving Search Agents
    [Paper] · arXiv:2605.05702
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

26. Language Models Need Sleep: Learning to Self-Modify and Consolidate Memories
    [Paper] · arXiv:2606.03979
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 5.2 Adapter; 8.2 Task Generator.

27. Learning to Adapt: Self-Improving Web Agent via Cognitive-Aware Exploration
    [Paper] · arXiv:2605.31365
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

28. Learning to Build the Environment: Self-Evolving Reasoning RL via Verifiable Environment Synthesis
    [Paper] · arXiv:2605.14392
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

29. Lifelong Robot Library Learning: Bootstrapping Composable and Generalizable Skills for Embodied Control with Language Models
    [Paper] · arXiv:2406.18746
    
    Taxonomy: L3. Updated object(s): 4.4 Skill Library; 2.1 Experience Memory.

30. LLM-as-a-Tutor: Policy-Aware Prompt Adaptation for Non-Verifiable RL
    [Paper] · arXiv:2607.04412
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.1 Experience Data; 7.2 Reward / Fitness.

31. MM-Zero: Self-Evolving Multi-Model Vision Language Models From Zero Data
    [Paper] · arXiv:2603.09206
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

32. Multi-Agent Evolve: LLM Self-Improve through Co-evolution
    [Paper] · arXiv:2510.23595
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 7.1 Judge; 8.1 Experience Data.

33. OMNI-EPIC: Open-endedness via Models of Human Notions of Interestingness with Environments Programmed in Code
    [Paper] · arXiv:2405.15568
    
    Taxonomy: L3. Updated object(s): 8.2 Task Generator; 8.3 Environment; 5.1 Model Weights.

34. Paired Open-Ended Trailblazer (POET): Endlessly Generating Increasingly Complex and Diverse Learning Environments and Their Solutions
    [Paper] · arXiv:1901.01753
    
    Taxonomy: L3. Updated object(s): 8.2 Task Generator; 8.3 Environment.

35. PopuLoRA: Co-Evolving LLM Populations for Reasoning Self-Play
    [Paper] · arXiv:2605.16727
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator; 6.5 Meta-optimization.

36. Project Auto-World: Towards Automated Benchmarking of Neural Relational Reasoners
    [Paper] · arXiv:2606.24965
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

37. Propose, Solve, Verify: Self-Play Through Formal Verification
    [Paper] · arXiv:2512.18160
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

38. R-Diverse: Mitigating Diversity Illusion in Self-Play LLM Training
    [Paper] · arXiv:2602.13103
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

39. R-Zero: Self-Evolving Reasoning LLM from Zero Data
    [Paper] · arXiv:2508.05004
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

40. SAGE: Multi-Agent Self-Evolution for LLM Reasoning
    [Paper] · arXiv:2603.15255
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

41. Scaling Self-Play with Self-Guidance
    [Paper] · arXiv:2604.20209
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

42. SEAgent: Self-Evolving Computer Use Agent with Autonomous Learning from Experience
    [Paper] · arXiv:2508.04700
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

43. Search Self-play: Pushing the Frontier of Agent Capability without Supervision
    [Paper] · arXiv:2510.18821
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

44. SEIF: Self-Evolving Reinforcement Learning for Instruction Following
    [Paper] · arXiv:2605.07465
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator; 7.1 Judge.

45. Seirênes: Adversarial Self-Play with Evolving Distractions for LLM Reasoning
    [Paper] · arXiv:2605.11636
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights.

46. Self-Play Only Evolves When Self-Synthetic Pipeline Ensures Learnable Information Gain
    [Paper] · arXiv:2603.02218
    
    Taxonomy: L3. Updated object(s): 8.2 Task Generator; 5.1 Model Weights.

47. Self-Questioning Language Models
    [Paper] · arXiv:2508.03682
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights.

48. SEVA: Self-Evolving Verification Agent with Process Reward for Fact Attribution
    [Paper] · arXiv:2606.29713
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 6.4 Curriculum.

49. SimWorld Studio: Automatic Environment Generation with Evolving Coding Agent for Embodied Agent Learning
    [Paper] · arXiv:2605.09423
    
    Taxonomy: L3. Updated object(s): 8.3 Environment; 8.2 Task Generator; 4.4 Skill Library.

50. SKILLFOUNDRY: Building Self-Evolving Agent Skill Libraries from Heterogeneous Scientific Resources
    [Paper] · arXiv:2604.03964
    
    Taxonomy: L3. Updated object(s): 4.4 Skill Library.

51. SkillHarness: Harnessing Safe Skills for Computer-Use Agents
    [Paper] · arXiv:2606.20636
    
    Taxonomy: L3. Updated object(s): 4.4 Skill Library.

52. Socratic-SWE: Self-Evolving Coding Agents via Trace-Derived Agent Skills
    [Paper] · arXiv:2606.07412
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 2.1 Experience Memory; 6.4 Curriculum.

53. SPARK: Self-Play with Asymmetric Reward from Knowledge Graphs
    [Paper] · arXiv:2605.05546
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

54. STELLA: Self-Evolving LLM Agent for Biomedical Research
    [Paper] · arXiv:2507.02004
    
    Taxonomy: L3. Updated object(s): 4.1 Tool Set; 4.4 Skill Library.

55. Survive or Collapse: The Asymmetric Roles of Data Gating and Reward Grounding in Self-Play RL
    [Paper] · arXiv:2605.22217
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights.

56. Tool-R0: Self-Evolving LLM Agents for Tool-Learning from Zero Data
    [Paper] · arXiv:2602.21320
    
    Taxonomy: L3. Updated object(s): 8.2 Task Generator; 5.1 Model Weights.

57. Towards Healthy Evolution: Exploring the Role and Mechanisms of Human-Agent Interaction in Self-Evolving Systems
    [Paper] · arXiv:2606.06114
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator; 1.1 Instruction.

58. Towards Understanding Self-play for LLM Reasoning
    [Paper] · arXiv:2510.27072
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 7.2 Reward / Fitness.

59. TREX: Automating LLM Fine-tuning via Agent-Driven Tree-based Exploration
    [Paper] · arXiv:2604.14116
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

60. TTCS: Test-Time Curriculum Synthesis for Self-Evolving
    [Paper] · arXiv:2601.22628
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 6.1 Objective.

61. TTSR: Test-Time Self-Evolving via Reflection
    [Paper] · arXiv:2603.03297
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator; 2.1 Experience Memory.

62. Verifier-Backed Hard Problem Generation for Mathematical Reasoning
    [Paper] · arXiv:2605.06660
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.1 Experience Data.

63. Vocabulary Dropout for Curriculum Diversity in LLM Co-Evolution
    [Paper] · arXiv:2604.03472
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

64. Voyager: An Open-Ended Embodied Agent with Large Language Models
    [Paper] · arXiv:2305.16291
    
    Taxonomy: L3. Updated object(s): 4.4 Skill Library; 8.2 Task Generator.

65. WIST: Web-Grounded Iterative Self-Play Tree for Domain-Targeted Reasoning Improvement
    [Paper] · arXiv:2603.22352
    
    Taxonomy: L3. Updated object(s): 5.1 Model Weights; 8.2 Task Generator.

Table-derived extensions

The following additions are explicitly discussed in Table 6 under trajectory distillation, iterative agent-system revision, or selective retention and deployment of updates.

1. DecoEvo: Score-Decoupled Co-Evolution of Solver and Rubric-Generator Skills in Text Space
   [Paper] · arXiv:2607.25675
   
   Survey-table rationale: Solver and rubric skills co-evolve across iterative rounds with score-independent audits. Updated object(s): 7.1 Evaluator / Judge; 4.4 Skill / Macro Library.

2. Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills
   [Paper] · arXiv:2603.25158
   
   Survey-table rationale: Error and success analysts merge patches into portable skill documents for later reuse. Updated object(s): 4.4 Skill / Macro Library.

3. Metis: Bridging Text and Code Memory for Self-Evolving Agents
   [Paper] · arXiv:2606.24151
   
   Survey-table rationale: Text plans and code tools are retained after completed tasks and validated through reuse and sandbox compilation. Updated object(s): 2.3 Procedural Memory; 4.3 Tool Implementation.

4. Library Drift: Diagnosing and Fixing a Silent Failure Mode in Self-Evolving LLM Skill Libraries
   [Paper] · arXiv:2605.19576
   
   Survey-table rationale: Periodic retention retires low-contribution skills and caps the evolving library. Updated object(s): 4.4 Skill / Macro Library.

5. APEX: Autonomous Policy Exploration for Self-Evolving LLM Agents
   [Paper] · arXiv:2605.21240
   
   Survey-table rationale: A milestone dependency graph retains episode outcomes and informs later exploration. Updated object(s): 2.2 Semantic / Knowledge Memory.

6. Learning Personalized Agents from Human Feedback
   [Paper] · arXiv:2602.16173
   
   Survey-table rationale: User clarification confirms and revises persistent preference entrie