Algorithmic reasoning is a fundamental cognitive ability that plays a pivotal
role in problem-solving and decision-making processes. Reinforcement Learning
(RL) has demonstrated remarkable proficiency in tasks such as motor control,
handling perceptual input, and managing stochastic environments. These
advancements have been enabled in part by the availability of benchmarks. In
this work we introduce PUZZLES, a benchmark based on Simon Tatham's Portable
Puzzle Collection, aimed at fostering progress in algorithmic and logical
reasoning in RL. PUZZLES contains 40 diverse logic puzzles of adjustable sizes
and varying levels of complexity; many puzzles also feature a diverse set of
additional configuration parameters. The 40 puzzles provide detailed
information on the strengths and generalization capabilities of RL agents.
Furthermore, we evaluate various RL algorithms on PUZZLES, providing baseline
comparisons and demonstrating the potential for future research. All the
software, including the environment, is available at
this https URL

算法推理在问题解决和决策过程中起着重要作用，强化学习在运动控制、处理感知输入和管理随机环境等任务中表现出卓越的技能。本研究引入了 PUZZLES，这是一个基于 Simon Tatham 的便携式拼图集的基准，旨在促进算法和逻辑推理在强化学习中的进展。PUZZLES 包含 40 个不同大小和复杂程度的多样逻辑拼图，许多拼图还具有一组多样的配置参数。这 40 个拼图提供了关于强化学习代理的优势和泛化能力的详细信息。此外，我们在 PUZZLES 上评估了各种强化学习算法，提供了基准比较，并展示了未来研究的潜力。所有软件和环境都可在此 https URL 网址上获得。

PUZZLES：神经算法推理的基准

PUZZLES: A Benchmark for Neural Algorithmic Reasoning

Large language models (LLMs), such as GPT-3 and GPT-4, have demonstrated
exceptional performance in various natural language processing tasks and have
shown the ability to solve certain reasoning problems. However, their reasoning
capabilities are limited and relatively shallow, despite the application of
various prompting techniques. In contrast, formal logic is adept at handling
complex reasoning, but translating natural language descriptions into formal
logic is a challenging task that non-experts struggle with. This paper proposes
a neuro-symbolic method that combines the strengths of large language models
and answer set programming. Specifically, we employ an LLM to transform natural
language descriptions of logic puzzles into answer set programs. We carefully
design prompts for an LLM to convert natural language descriptions into answer
set programs in a step by step manner. Surprisingly, with just a few in-context
learning examples, LLMs can generate reasonably complex answer set programs.
The majority of errors made are relatively simple and can be easily corrected
by humans, thus enabling LLMs to effectively assist in the creation of answer
set programs.

本文提出了一种神经符号方法，将大型语言模型和答案集编程的优势相结合，以将逻辑谜题的自然语言描述转化为答案集程序。

利用大型语言模型生成答案集程序

Leveraging Large Language Models to Generate Answer Set Programs

This work delves into the realm of logic puzzles by focusing on the Knight
and Knave problems popularized by Raymond Smullyan in his book series "What is
the Name of This Book?". The puzzles revolve around characters known as Knights
(truth-tellers) and Knaves (liars), challenging solvers to determine the true
identity of each person based on their statements. This paper explores the
utilization of Python algorithms to automate the process of solving these
puzzles, offering a computational approach that enhances efficiency and
accessibility. In this work, we aim to develop a Python algorithm capable of
parsing and analyzing the statements provided in the Knight and Knave puzzles.
A logical reasoning framework is integrated within the algorithm to deduce the
identities of the characters based on their statements. The algorithm processes
the input statements, create a knowledge base, and make deductions following
the rules of Knight and Knave logic. The developed algorithm is thoroughly
tested on various instances of Knight and Knave puzzles, comparing its results
to known solutions and manual approaches. We further expand the scope of the
problem by introducing a Normal (who can sometimes lie and sometimes say the
truth).

该研究使用 Python 算法自动解决骑士和诡辩者的逻辑难题，并整合了逻辑推理框架，通过分析陈述来确定每个人的真实身份，并通过与已知解决方案和手动方法进行比较进行了全面测试。

使用算法解决逻辑谜题

What is the Title of this Paper? Solving logic puzzles using algorithms

For a natural language problem that requires some non-trivial reasoning to
solve, there are at least two ways to do it using a large language model (LLM).
One is to ask it to solve it directly. The other is to use it to extract the
facts from the problem text and then use a theorem prover to solve it. In this
note, we compare the two methods using ChatGPT and GPT4 on a series of logic
word puzzles, and conclude that the latter is the right approach.

本文通过在逻辑单词谜题上比较使用 ChatGPT 和 GPT-4 直接解决问题和提取问题文本事实并使用定理证明器解决两种方法，并得出后者是正确的方法来研究了利用大语言模型解决自然语言问题需要一些非平凡的推理。