ACL Findings2026
Beyond 'Aha!': Toward Systematic Meta-Abilities Alignment in Large Reasoning Models
Zhiyuan Hu, Yibo Wang, Hanze Dong, Yuhui Xu, Amrita Saha, Caiming Xiong, Bryan Hooi, Junnan Li
Abstract
Large reasoning models (LRMs) already possess a latent capacity for long chain-of-thought reasoning. Prior work has shown that outcomebased reinforcement learning (RL) can incidentally elicit advanced reasoning behaviors such as self-correction, backtracking, and verification-phenomena often referred to as the model's "aha moment". However, the timing and consistency of these emergent behaviors remain unpredictable and uncontrollable, limiting the scalability and reliability of LRMs' reasoning capabilities. To address these limitations, we move beyond reliance on prompts and unpredictable "aha moments". Instead, we explicitly align models with three meta-abilities: deduction, induction, and abduction, using automatically generated, self-verifiable tasks. Our three-stage pipeline (individual alignment, parameter-space merging, domain-specific reinforcement learning) boosts performance by over 10% relative to instruction-tuned baselines. Furthermore, domain-specific RL from the aligned checkpoint yields an additional gain in performance ceiling for both 7B and 32B models across math, coding, and science benchmarks, showing that explicit meta-ability alignment offers a scalable and dependable foundation for reasoning. Code and data can be found in Software and Data part in submission page.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
5 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
Cited by top-tier papers5
- MemGen: Weaving Generative Latent Memory for Self-Evolving AgentsGuibin Zhang, Muxin Fu, Shuicheng YanICLR 2026 · 102 citations
- EvoTest: Evolutionary Test-Time Learning for Self-Improving Agentic SystemsYufei He, Juncheng Liu, Yue Liu, Yibo Li et al.ICLR 2026 · 36 citations
- ReasonAny: Incorporating Reasoning Capability to Any Model via Simple and Effective Model MergingJunyao Yang, Chen Qian, Wen Shen, Yong Liu et al.ACL 2026 · 1 citation
- MetaAct-RL: Training Language Models for Reasoning Through Meta-Action-Based Reinforcement LearningZhiheng Xi, Yuhui Wang, Yiwen Ding, Guanyu Li et al.AAAI 2026
- What Characterizes Effective Reasoning? Revisiting Length, Review, and Structure of CoTYunzhen Feng, Julia Kempe, Cheng Zhang, Parag Jain et al.ICML 2026
Builds on2
Related papers
- Metaphor Reasoning is Meta-reasoningQianyu He, Junting Lu, Yikai Zhang, Siyu Yuan et al.ACL 2026
- AceReason-Nemotron: Advancing Math and Code Reasoning through Reinforcement LearningYang Chen, Zhuolin Yang, Zihan Liu, Chankyu Lee et al.NeurIPS 2025 · 79 citations
- Provable Benefits of RLVR over SFT for Reasoning Models: Learning to Backtrack EfficientlyStanley Wei, Juno KimICML 2026
- PretrainZero: Reinforcement Active Learning on Pretraining DataXingrun Xing, Zhiyuan Fan, Jie Lou, Guoqi Li et al.ICML 2026
- Unveiling the Compositional Ability Gap in Vision-Language Reasoning ModelTianle Li, Jihai Zhang, Yongming Rao, Yu ChengNeurIPS 2025 · 17 citations
