Elamur Research⌕
☰
home / papers / 2026 / Beyond Meta-Reasoning: Metacognitive Consolidation for Self-Improving LLM Reasoning

Beyond Meta-Reasoning: Metacognitive Consolidation for Self-Improving LLM Reasoning

Paper · 2026 · Ziqing Zhuang, Linhai Zhang, Jiasheng Si et al.

Summary

The paper addresses a gap in LLM meta-reasoning. Most current approaches are episodic: they plan or intervene within one problem and then discard what they learned. As a result, the same mistakes recur and the costly verification and retry routines run again from scratch. The authors propose Metacognitive Consolidation, which draws on Anderson's idea of proceduralization: experience from past reasoning episodes is turned into reusable procedural knowledge.

Their instantiation, MC², has two loops around a frozen backbone:

  • Inner loop (Meta-Reasoning Orchestrator, MRO): solving is split among a Reasoner, a Monitor and a Controller. This produces explicit, attributable traces of each attempt, its critique and its correction.
  • Outer loop (MetaCognitive Accumulator, MCA): these traces are distilled at several timescales into evolving meta-knowledge. That knowledge rewrites the role-specific prompts used on later instances.

Experiments across several benchmarks and backbone models report consistent gains over meta-reasoning baselines, and accuracy rises as experience accumulates (e.g., on MATH-500). The work matters because it converts per-instance test-time compute into lasting improvement without any fine-tuning.

Key points

  • MRO loop: the Monitor's report contains a verdict, the location of the first suspicious step and a short critique. The Controller then chooses ACCEPT, PATCH or RESTART, and the loop runs for at most N iterations.

  • MCA hierarchy: knowledge is consolidated in three levels:

    • instance-level reflections, built by a deterministic extractor;
    • batch-level micro-lessons, formed through in-batch distillation;
    • slowly evolving global meta-knowledge.

    Separate streams are kept for the Reasoner, Monitor and Controller.

  • Learning mechanism: all learning happens at test time through prompt-based policy updates (P_t → P_t+1). Model parameters stay unchanged, so the accumulated knowledge remains inspectable.

  • Positioning: the method differs from memory approaches such as ReasoningBank, Reflexion, Buffer of Thoughts and ReasonFlux. Those store task-level solution content, whereas MC² stores knowledge about how to monitor and control reasoning.

  • Stated limitations: consolidation is task-specific, with no cross-task transfer yet. The multiple roles add inference overhead. Internalizing the knowledge into weights is left for future work. The excerpt omits the detailed result tables, so specific scores are not reported here.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🔬The Alan Turing InstituteLondon · United Kingdom29 works · 4,368 citations🎓King's College LondonLondon · United Kingdom42 works · 1,617 citations🎓Southeast UniversityNanjing · China28 works · 354 citations🎓Qilu University of TechnologyJinan · China4 works · 3 citations🎓Shandong Academy of SciencesJinan · China2 works · 1 citation

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 2Research subjects and languages the work is about.Language models & pretraining · 3,320 worksReasoning · 1,426 works

Cite

Cite as Ziqing Zhuang, Linhai Zhang, Jiasheng Si, Deyu Zhou and Yulan He. 2026. Beyond Meta-Reasoning: Metacognitive Consolidation for Self-Improving LLM Reasoning. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1095

@inproceedings{zhuang2026beyond,
  title = {{Beyond Meta-Reasoning: Metacognitive Consolidation for Self-Improving LLM Reasoning}},
  author = {Ziqing Zhuang, Linhai Zhang, Jiasheng Si, Deyu Zhou and Yulan He},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1095},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1095}
}

References

31 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Rakefet Ackerman and Valerie A. Thompson. 2017. Meta-reasoning: Monitoring and control of thinking and reasoning. Trends in Cognitive Sciences, 21(8):607–617. find
  2. John R Anderson. 1982. Acquisition of cognitive skill. Psychological review, 89(4):369. find
  3. Ali Behrouz, Meisam Razaviyayn, Peilin Zhong, and Vahab Mirrokni. 2025a. Nested learning: The illusion of deep learning architectures. In The Thirty-ninth Annual Conference on Neural Information Processing Systems. find
  4. Ali Behrouz, Peilin Zhong, and Vahab Mirrokni. 2025b. Titans: Learning to memorize at test time. In The Thirty-ninth Annual Conference on Neural Information Processing Systems. find
  5. Wenhu Chen, Ming Yin, Max Ku, Pan Lu, Yixin Wan, Xueguang Ma, Jianyu Xu, Xinyi Wang, and Tony Xia. 2023. TheoremQA: A theorem-driven question answering dataset. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 7889–7901, Singapore. Association for Comp…
  6. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. 2021. Training verifiers to solve math word problems. Preprint, arXiv:2110.14168.
  7. Peizhong Gao, Ao Xie, Shaoguang Mao, Wenshan Wu, Yan Xia, Haipeng Mi, and Furu Wei. 2024. Meta reasoning for large language models. arXiv preprint arXiv:2406.11698. — [arXiv:2406.11698](https://arxiv.org/abs/2406.11698) find
  8. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, and 1 others. 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948.
  9. Yancheng He, Shilong Li, Jiaheng Liu, Weixun Wang, Xingyuan Bu, Ge Zhang, Z.y. Peng, Zhaoxiang Zhang, Zhicheng Zheng, Wenbo Su, and Bo Zheng. 2025. Can large language models detect errors in long chain-of-thought reasoning? In Proceedings of the 63rd Annual Meeting of the Association for Computation…
  10. Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. 2021. Measuring mathematical problem solving with the MATH dataset. In Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 2).
  11. Qin Liu, Wenxuan Zhou, Nan Xu, James Y. Huang, Fei Wang, Sheng Zhang, Hoifung Poon, and Muhao Chen. 2025. Metascale: Test-time scaling with evolving meta-thoughts. Preprint, arXiv:2503.13447.
  12. Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, Shashank Gupta, Bodhisattwa Prasad Majumder, Katherine Hermann, Sean Welleck, Amir Yazdanbakhsh, and Peter Clark. 2023. Self-refine: Iterative refinement with… find
  13. OpenAI, Aaron Jaech, Adam Kalai, Adam Lerer, Adam Richardson, Ahmed El-Kishky, Aiden Low, Alec Helyar, Aleksander Madry, Alex Beutel, Alex Carney, Alex Iftimie, Alex Karpenko, Alex Tachard Passos, Alexander Neitz, Alexander Prokofiev, Alexander Wei, Allison Tam, Ally Bennett, and 243 others. 2024. O… — OpenAI o1 System Card
  14. Siru Ouyang, Jun Yan, I-Hung Hsu, Yanfei Chen, Ke Jiang, Zifeng Wang, Rujun Han, Long T. Le, Samira Daruki, Xiangru Tang, Vishy Tirumalashetty, George Lee, Mahsan Rofouei, Hangfei Lin, Jiawei Han, Chen-Yu Lee, and Tomas Pfister. 2025. Reasoningbank: Scaling agent self-evolving with reasoning memory.… — [arXiv:2509.25140](https://arxiv.org/abs/2509.25140) find
  15. Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik R Narasimhan, and Shunyu Yao. 2023. Reflexion: language agents with verbal reinforcement learning. In Thirty-seventh Conference on Neural Information Processing Systems.
  16. Charlie Victor Snell, Jaehoon Lee, Kelvin Xu, and Aviral Kumar. 2025. Scaling LLM test-time compute optimally can be more effective than scaling parameters for reasoning. In The Thirteenth International Conference on Learning Representations. find
  17. Yuan Sui, Yufei He, Tri Cao, Simeng Han, Yulin Chen, and Bryan Hooi. 2025. Meta-reasoner: Dynamic guidance for optimized inference-time reasoning in large language models. Preprint, arXiv:2502.19918.
  18. Mirac Suzgun and Adam Tauman Kalai. 2024. Meta-prompting: Enhancing language models with task-agnostic scaffolding. arXiv preprint arXiv:2401.12954.
  19. Hexiang Tan, Fei Sun, Sha Liu, Du Su, Qi Cao, Xin Chen, Jingang Wang, Xunliang Cai, Yuanzhuo Wang, Huawei Shen, and Xueqi Cheng. 2025. Too consistent to detect: A study of self-consistent errors in LLMs. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages…
  20. Ziyu Wan, Yunxiang LI, Xiaoyu Wen, Yan Song, Hanjing Wang, Linyi Yang, Mark Schmidt, Jun Wang, Weinan Zhang, Shuyue Hu, and Ying Wen. 2025. ReMA: Learning to meta-think for LLMs with multi-agent reinforcement learning. In The Thirty-ninth Annual Conference on Neural Information Processing Systems. find
  21. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc V Le, Ed H. Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2023. Self-consistency improves chain of thought reasoning in language models. In The Eleventh International Conference on Learning Representations.
  22. Yuqing Wang and Yun Zhao. 2024. Metacognitive prompting improves understanding in large language models. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 1914–1926, Mexico… find
  23. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, brian ichter, Fei Xia, Ed H. Chi, Quoc V Le, and Denny Zhou. 2022. Chain of thought prompting elicits reasoning in large language models. In Advances in Neural Information Processing Systems.
  24. Yangzhen Wu, Zhiqing Sun, Shanda Li, Sean Welleck, and Yiming Yang. 2025. Inference scaling laws: An empirical analysis of compute-optimal inference for LLM problem-solving. In The Thirteenth International Conference on Learning Representations. find
  25. Jun Xu, Xinkai Du, Yu Ao, Peilong Zhao, Yang Li, Ling Zhong, Lin Yuan, Zhongpu Bo, Xiaorui Wang, Mengshu Sun, and 1 others. 2025. Thinker: Training llms in hierarchical thinking for deep search via multi-turn interaction. arXiv preprint arXiv:2511.07943. — [arXiv:2511.07943](https://arxiv.org/abs/2511.07943) find
  26. Hanqi Yan, Linhai Zhang, Jiazheng Li, Zhenyi Shen, and Yulan He. 2025. Position: LLMs need a bayesian meta-reasoning framework for more robust and generalizable reasoning. In Forty-second International Conference on Machine Learning Position Paper Track. find
  27. Chenxu Yang, Qingyi Si, Mz Dai, Dingyu Yao, Mingyu Zheng, Minghui Chen, Zheng Lin, and Weiping Wang. 2025. Test-time prompt intervention. arXiv preprint arXiv:2508.02511. — [arXiv:2508.02511](https://arxiv.org/abs/2508.02511) find
  28. Ling Yang, Zhaochen Yu, Tianjun Zhang, Shiyi Cao, Minkai Xu, Wentao Zhang, Joseph E. Gonzalez, and Bin CUI. 2024. Buffer of thoughts: Thought-augmented reasoning with large language models. In The Thirty-eighth Annual Conference on Neural Information Processing Systems. find
  29. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Thomas L. Griffiths, Yuan Cao, and Karthik R Narasimhan. 2023. Tree of thoughts: Deliberate problem solving with large language models. In Thirty-seventh Conference on Neural Information Processing Systems. find
  30. Zijing Zhang, Ziyang Chen, Mingxiao Li, Zhaopeng Tu, and Xiaolong Li. 2025. Rlvmr: Reinforcement learning with verifiable meta-reasoning rewards for robust long-horizon agents. arXiv preprint arXiv:2507.22844. — [arXiv:2507.22844](https://arxiv.org/abs/2507.22844) find
  31. Jiaru Zou, Ling Yang, Jingwen Gu, Jiahao Qiu, Ke Shen, Jingrui He, and Mengdi Wang. 2025. Reasonflux-PRM: Trajectory-aware PRMs for long chain-of-thought reasoning in LLMs. In The Thirty-ninth Annual Conference on Neural Information Processing Systems. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhuang-2026-beyond-meta-reasoning-metacognitive-consolidation/