mlsanityData to 14 Oct 2026

TTSR: Test-Time Self-Evolving via Reflection

arXiv 2603.03297 ↗First posted 6 Feb 2026cs.CL cs.AI cs.LGcited by 2

On the map: Synthetic data and self-improvement for LLMs → Synthetic reasoning data and bootstrapping

Abstract (arXiv)

Test-time training (TTT) adapts large language models (LLMs) during inference using only unlabeled test inputs. Existing methods, however, face two major bottlenecks on hard reasoning tasks: (1) lack of learnable samples, as self-generated pseudo-labels on difficult questions are often noisy and yield unstable rewards; and (2) inefficient exploration, as performance gains depend on repeatedly sampling many rollouts without explicit diagnosis of why previous attempts fail. We propose TTSR (Test-Time Self-Reflection), a self-evolving framework based on a reflect-then-synthesize paradigm. A single pretrained model alternates between a textitStudent role and a textitTeacher role: the Student solves test questions and updates, while the Teacher analyzes failed trajectories and synthesizes targeted variant questions closer to the Student's capability frontier. TTSR further maintains a cross-iteration textitweakness memory and compiles persistent weaknesses into a lightweight textitstrategy note prepended to subsequent Student inputs, so diagnostic knowledge can guide exploration and gradually fade as weaknesses are resolved. Experiments on challenging mathematical reasoning benchmarks show consistent test-time improvements, strong cross-backbone generalization, and transfer to general-domain reasoning tasks.

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Placed in Synthetic reasoning data and bootstrapping, one of 13 directions in Synthetic data and self-improvement for LLMs. Builds on 29 of its 37 references that are in the map. First posted on arXiv 6 Feb 2026.

Builds on

29 of 37 references.

Closest ancestors · 4

Same direction, newest first

  1. 2026TTCS: Test-Time Curriculum Synthesis for Self-Evolving30 Jan
  2. 2025R-Zero: Self-Evolving Reasoning LLM from Zero Data7 Aug
  3. 2025Genius: A Generalizable and Purely Unsupervised Self-Training Framework For Advanced Reasoning11 Apr
  4. 2022Large Language Models Can Self-Improve20 Oct

Other references · 25

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  2. 2025Qwen3 Technical Reportcited by 8,820
  3. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
  4. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,879
  5. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,724
  6. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,986
  7. 2024MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmarkcited by 1,848
  8. 2022Solving Quantitative Reasoning Problems with Language Modelscited by 1,392
  9. 2024OlympiadBench: A Challenging Benchmark for Promoting AGI with Olympiad-Level Bilingual Multimodal Scientific Problemscited by 1,136
  10. 2023Large Language Models Cannot Self-Correct Reasoning Yetcited by 1,017
  11. 2024Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Modelscited by 557
  12. 2025SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wildcited by 511
13 more
  1. 2025Absolute Zero: Reinforced Self-play Reasoning with Zero Datacited by 344
  2. 2024ProcessBench: Identifying Process Errors in Mathematical Reasoningcited by 292
  3. 2025LIMR: Less is More for RL Scalingcited by 139
  4. 2026Learning to Discover at Test Timecited by 115
  5. 2025OctoThinker: Mid-training Incentivizes Reinforcement Learning Scalingcited by 113
  6. 2024CriticBench: Benchmarking LLMs for Critique-Correct Reasoningcited by 88
  7. 2025Self-Challenging Language Model Agentscited by 59
  8. 2025SeRL: Self-Play Reinforcement Learning for Large Language Models with Limited Datacited by 51
  9. 2025VisPlay: Self-Evolving Vision-Language Models from Imagescited by 47
  10. 2025Language Self-Play For Data-Free Trainingcited by 43
  11. 2025SPC: Evolving Self-Play Critic via Adversarial Games for LLM Reasoningcited by 42
  12. 2025Learning to Solve and Verify: A Self-Play Framework for Code and Test Generationcited by 35
  13. 2025First SFT, Second RL, Third UPT: Continual Improving Multi-Modal LLM Reasoning via Unsupervised Post-Trainingcited by 34

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026AgentStream: How Well Do Self-Evolving LLM Agents Perform Under Streaming Tasks?cited by 4
  2. 2026A Survey on Self-Improving Test-Time Intelligence: Feedback-Driven Adapting, Learning, and Scaling at Inferencecited by 3

Similar, not cited

Close in wording, no citation either way

  1. 2026TEMPO: Scaling Test-time Training for Large Reasoning Models
  2. 2026Learning to Self-Evolve
  3. 2026Test-Time Learning with an Evolving Library
  4. 2025Self-Reflective Generation at Test Time
  5. 2025Continuous Self-Improvement of Large Language Models by Test-time Training with Verifier-Driven Sample Selection
  6. 2026Test-Time Meta-Adaptation with Self-Synthesis
  7. 2026Rethinking Expressivity and Efficiency in Test-Time Training
  8. 2025The Art of Scaling Test-Time Compute for Large Language Models