ACL Findings2026
Beyond Memorization: Extending Reasoning Depth with Recurrence, Memory and Test-Time Compute Scaling
Ivan Rodkin, Daniil Orel, Konstantin Smirnov, Arman Bolatov, Bilal Elbouardi, Besher Hassan, Yuri Kuratov, Aydar Bulatov, Preslav Nakov, Timothy Baldwin, Artem Shelmanov, Mikhail Burtsev
摘要
Reasoning is a core capability of large language models (LLMs), yet how multi-step reasoning is learned and executed remains unclear. We study this question in a controlled cellular-automata (1dCA) framework that excludes memorization by using disjoint training and test rules. Given a short state sequence, the model is required to infer the hidden local rule and then chain it to predict multiple future steps. Our evaluation shows that LLMs largely fail to reliably solve a natural-language proxy of the proposed task. We find that most neural architectures trained from scratch can learn rule inference and achieve high next-step accuracy, but performance drops sharply as the required number of intermediate reasoning steps increases. Experiments show that increasing model depth is crucial, and extending effective depth via recurrence, memory, or test-time compute improves results but remains bounded. Code is available on github: https://github.com/RodkinIvan/associative- recurrent-memory-transformer/tree/ACT.
问 Lune
几分钟读透这篇 Paper,不用再花一下午。
Lune 能读完全文,也能读引用它的顶会 Paper,讲清它的结论、证据和局限,每一点都注明出自哪篇 Paper。不用硬啃,也不靠猜。
脉络
- Tree of Thoughts: Deliberate Problem Solving with Large Language Models 引用了 Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
- Chain-of-Thought Prompting Elicits Reasoning in Large Language Models 引用了 Training language models to follow instructions with human feedback
它引用的 Paper21
- Training language models to follow instructions with human feedbackLong Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida 等NeurIPS 2022 · 24,707 次引用
- Chain-of-Thought Prompting Elicits Reasoning in Large Language ModelsJason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma 等NeurIPS 2022 · 22,562 次引用
- Tree of Thoughts: Deliberate Problem Solving with Large Language ModelsShunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran 等NeurIPS 2023 · 5,068 次引用
- Efficiently Modeling Long Sequences with Structured State SpacesAlbert Gu, Karan Goel, Christopher RéICLR 2022 · 3,482 次引用
- Compressive Transformers for Long-Range Sequence ModellingJack W. Rae, Anna Potapenko, Siddhant M. Jayakumar, Chloe Hillier 等ICLR 2020 · 833 次引用
相关 Paper
- Reasoning with Latent Thoughts: On the Power of Looped TransformersNikunj Saunshi, Nishanth Dikkala, Zhiyuan Li, Sanjiv Kumar 等ICLR 2025
- Language models can learn implicit multi-hop reasoning, but only if they have lots of training dataYuekun Yao, Yupei Du, Dawei Zhu, Michael Hahn 等EMNLP 2025
- How reinforcement learning after next-token prediction facilitates learningNikolaos Tsilivis, Eran Malach, Karen Ullrich, Julia KempeICLR 2026 · 9 次引用
- ChainGPT: Dual-Reasoning Model with Recurrent Depth and Multi-Rank State UpdatesYunao Zheng, Xiaojie Wang, Lei Ren, Chen WeiICLR 2026
- The Illusion of Diminishing Returns: Measuring Long Horizon Execution in LLMsAkshit Sinha, Arvindh Arun, Shashwat Goel, Steffen Staab 等ICLR 2026 · 64 次引用
