arXiv AI

LongReasonArena: A Long Reasoning Benchmark for Large Language Models

LongReasonArena is a new benchmark that evaluates the long reasoning abilities of large language models, rather than just their comprehension of long inputs. It presents tasks that require multi-step algorithms involving retrieval and backtracking, with reasoning lengths that can be scaled up to one million tokens. Evaluation shows that even advanced models like GPT‑5 struggle, achieving only 27.8% accuracy on the hardest level, and accuracy decreases linearly with the logarithm of expected reasoning steps.

arXiv AI
Sep 7

Harnessing the Reasoning Economy: A Survey of Efficient Reasoning for Large Language Models

The paper surveys efficient reasoning in large language models, contrasting fast intuitive (System 1) and slow deep (System 2) reasoning. It analyzes why System 2 is computationally costly yet more accurate, and why System 1 is efficient but less effective. The survey covers causes of inefficiency, patterns of reasoning behavior, and potential solutions to balance performance and computational budgets, offering actionable insights and an open‑source repository for ongoing research.

By Rui Wang, Hongru Wang, Boyang Xue, Jianhui Pang, Shudong Liu, Yi Chen, Jiahao Qiu, Derek Fai Wong, Heng Ji, Kam-Fai Wong
arXiv AI
Jun 15

Fractured Chain-of-Thought Reasoning

arXiv:2505. 12992v4 Announce Type: replace-cross Abstract: Inference-time scaling techniques have significantly bolstered the reasoning capabilities of large language models (LLMs) by harnessing additional computational effort at inference without retraining.

By Baohao Liao, Hanze Dong, Yuhui Xu, Doyen Sahoo, Christof Monz, Junnan Li, Caiming Xiong
arXiv Machine Learning
Oct 22

LongZero: Enhancing Long-Context Reasoning in LLMs from Zero Data

LongZero is a reinforcement learning method that enables large language models to improve long‑context reasoning without external data. It works by having the model first generate a compact, structured question‑answer pair with supporting evidence, then expanding that evidence into a long, multi‑document context. The approach combines self‑written multi‑hop reasoning and multi‑round retrieval tasks, trained with GRPO, and outperforms baseline RL methods on five long‑context reasoning benchmarks.

By Zhennan Wan, Jianfei Chen
arXiv Computation and Language
Oct 9

Probing for Long-Horizon Deductive Reasoning Capabilities in Language Models with Prolog

The paper introduces ProloNg, a synthetic benchmark designed to test long-horizon deductive reasoning in language models using Prolog. It evaluates eight reasoning models from five leading LLM families on problems with reasoning depths up to 22 and context lengths of 62,000 tokens. Results show that model performance drops sharply as reasoning depth increases, with most models performing near chance beyond a depth of 10.

By Hadeel Al-Negheimish, Jasna Ilieva, Yoon Kim
arXiv AI
Aug 26

Parason: Revealing Subtask and Trial Parallelism in LLM Reasoning

Parason is a new framework that discovers and exploits both subtask and trial parallelism in large language model (LLM) reasoning. By converting sequential reasoning traces into structured parallel trajectories and training with Parallelism-Aware Group Relative Policy Optimization, it balances accuracy, latency, and parallelism. Experiments on mathematical reasoning benchmarks such as AIME24 and AIME25 show that Parason achieves an average acceleration of about 1.7× while maintaining competitive accuracy.

By Zhengyang Zhang, Zijian Zhang, Jiaxuan Gao, Shusheng Xu, Yi Wu, Song Han, Ligeng Zhu