mlsanityData to 7 Oct 2026

Synthetic reasoning data and bootstrapping

165 papers0 this weekLow confidence in this grouping: the papers may belong to more than one topic, or the group may changeLow confidence in this grouping: the papers may belong to more than one topic, or the group may change · About

Keywords: reasoning, llms, language, training, mathematical

Key facts. Synthetic reasoning data and bootstrapping: 165 papers on arXiv since 2021 (data through 7 Oct 2026). 32 of its 165 mapped papers (19%) cite “STaR: Bootstrapping Reasoning With Reasoning” (2022, arXiv 2203.14465). 17 of its 165 mapped papers (10%) cite “Large Language Models Can Self-Improve” (2022, arXiv 2210.11610). 1 new paper 17 Sep–7 Oct.

What’s new

17 Sep – 7 Oct 2026

1 new since 17 Sep

  1. Ladders of Thought: A Self-Evolving Curriculum of Progressively Simplified Reasoning Traces

Foundations

Most cited by papers here

  1. STaR: Bootstrapping Reasoning With Reasoning2022cited by 32 here
  2. Large Language Models Can Self-Improve2022cited by 17 here
  3. MetaMath: Bootstrap Your Own Mathematical Questions for Large Language Models2023cited by 15 here
  4. Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Models2023cited by 11 here
  5. R-Zero: Self-Evolving Reasoning LLM from Zero Data2025cited by 10 here

Lineage

The papers this direction builds on, oldest first.

Each paper on the line cites the one before.

  1. 2022STaR: Bootstrapping Reasoning With Reasoningcited by 32 here
  2. 2023Reinforced Self-Training (ReST) for Language Modelingcited by 10 here
  3. 2023Beyond Human Data: Scaling Self-Training for Problem-Solving with Language Modelscited by 11 here
  4. 2024Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Modelscited by 10 here
  5. 2024Self-Rewarding Language Modelscited by 9 here
  6. 2024Best Practices and Lessons Learned on Synthetic Datacited by 5 here
  7. 2024JiuZhang3.0: Efficiently Improving Mathematical Reasoning by Training Small Data Synthesis Modelscited by 6 here
  8. 2024Qwen2.5-Math Technical Report: Toward Mathematical Expert Model via Self-Improvementcited by 8 here
16 more it builds on, not in this chain
  1. 2022Large Language Models Can Self-Improvecited by 17 here
  2. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 31 here
  3. 2023On-Policy Distillation of Language Models: Learning from Self-Generated Mistakescited by 5 here
  4. 2023MetaMath: Bootstrap Your Own Mathematical Questions for Large Language Modelscited by 15 here
  5. 2024MathGenie: Generating Synthetic Data with Question Back-translation for Enhancing Mathematical Reasoning of LLMscited by 8 here
  6. 2024Key-Point-Driven Data Synthesis with its Enhancement on Mathematical Reasoningcited by 6 here
  7. 2024ReST-MCTS*: LLM Self-Training via Process Reward Guided Tree Searchcited by 8 here
  8. 2024Magpie: Alignment Data Synthesis from Scratch by Prompting Aligned LLMs with Nothingcited by 6 here
  9. 2024RL on Incorrect Synthetic Data Scales the Efficiency of LLM Math Reasoning by Eight-Foldcited by 4 here
  10. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 10 here
  11. 2025rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinkingcited by 6 here
  12. 2025Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRscited by 6 here
  13. 2025Absolute Zero: Reinforced Self-play Reasoning with Zero Datacited by 14 here
  14. 2025Synthetic Data RL: Task Definition Is All You Needcited by 4 here
  15. 2025CoT-Self-Instruct: Building high-quality synthetic prompts for reasoning and non-reasoning taskscited by 4 here
  16. 2025R-Zero: Self-Evolving Reasoning LLM from Zero Datacited by 10 here

The 24 most cited of 129.

Browse

30 of 165, ranked 6–35

  1. 6MathGenie: Generating Synthetic Data with Question Back-translation for Enhancing Mathematical Reasoning of LLMs20248
  2. 7Qwen2.5-Math Technical Report: Toward Mathematical Expert Model via Self-Improvement20248
  3. 8rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinking20256
  4. 9Key-Point-Driven Data Synthesis with its Enhancement on Mathematical Reasoning20246
  5. 10JiuZhang3.0: Efficiently Improving Mathematical Reasoning by Training Small Data Synthesis Models20246
  6. 11Synthetic Data RL: Task Definition Is All You Need20254
  7. 12CoT-Self-Instruct: Building high-quality synthetic prompts for reasoning and non-reasoning tasks20254
  8. 13RL on Incorrect Synthetic Data Scales the Efficiency of LLM Math Reasoning by Eight-Fold20244
  9. 14Self-Evolving Curriculum for LLM Reasoning20253
  10. 15Self-Training with Direct Preference Optimization Improves Chain-of-Thought Reasoning20243
All 30
  1. 16Advancing Mathematical Reasoning in Language Models: The Impact of Problem-Solving Data, Data Synthesis Methods, and Training Stages20252
  2. 17Socratic-Zero : Bootstrapping Reasoning via Data-Free Agent Co-evolution20252
  3. 18TTCS: Test-Time Curriculum Synthesis for Self-Evolving20261
  4. 19R-Diverse: Mitigating Diversity Illusion in Self-Play LLM Training20261
  5. 20CHIMERA: Compact Synthetic Data for Generalizable LLM Reasoning20261
  6. 21BOLT: Bootstrap Long Chain-of-Thought in Language Models without Distillation20251
  7. 22SQL-o1: A Self-Reward Heuristic Dynamic Search Method for Text-to-SQL20251
  8. 23Self-rewarding correction for mathematical reasoning20251
  9. 24Synthetic Data Generation & Multi-Step RL for Reasoning & Tool Use20251
  10. 25Genius: A Generalizable and Purely Unsupervised Self-Training Framework For Advanced Reasoning20251
  11. 26RV-Syn: Rational and Verifiable Mathematical Reasoning Data Synthesis based on Structured Function Library20251
  12. 27ExeSQL: Self-Taught Text-to-SQL Models with Execution-Driven Bootstrapping for SQL Dialects20251
  13. 28Self-Training Large Language Models with Confident Reasoning20251
  14. 29SPFT-SQL: Enhancing Large Language Model for Text-to-SQL Parsing by Self-Play Fine-Tuning20251
  15. 30Learning to Pose Problems: Reasoning-Driven and Solver-Adaptive Data Synthesis20251
  16. 31Self-training Large Language Models through Knowledge Detection20241
  17. 32Self-training Language Models for Arithmetic Reasoning20241
  18. 33Source2Synth: Synthetic Data Generation and Curation Grounded in Real Data Sources20241
  19. 34ReGenesis: LLMs can Grow into Reasoning Generalists via Self-Improvement20241
  20. 35Advancing Large Language Model Attribution through Self-Improving20241