Analysing Mathematical Reasoning Abilities of Neural Models
arXiv:1904.01557
Abstract
Mathematical reasoning---a core ability within human intelligence---presents some unique challenges as a domain: we do not come to understand and solve mathematical problems primarily on the back of experience and evidence, but on the basis of inferring, learning, and exploiting laws, axioms, and symbol manipulation rules. In this paper, we present a new challenge for the evaluation (and eventually the design) of neural architectures and similar system, developing a task suite of mathematics problems involving sequential questions and answers in a free-form textual input/output format. The structured nature of the mathematics domain, covering arithmetic, algebra, probability and calculus, enables the construction of training and test splits designed to clearly illuminate the capabilities and failure-modes of different architectures, as well as evaluate their ability to compose and relate knowledge and learned processes. Having described the data generation process and its potential future expansions, we conduct a comprehensive analysis of models from two broad classes of the most powerful sequence-to-sequence architectures and find notable differences in their ability to resolve mathematical problems and generalize their knowledge.
References in corpus (11)
- Adam: A Method for Stochastic Optimization
- Adaptive Computation Time for Recurrent Neural Networks
- Learning to Execute
- Neural GPUs Learn Algorithms
- Learning a SAT Solver from Single-Bit Supervision
- Inferring Algorithmic Patterns with Stack-Augmented Recurrent Nets
- Measuring abstract reasoning in neural networks
- Program Induction by Rationale Generation : Learning to Solve and Explain Algebraic Word Problems
- Relational recurrent neural networks
- Can Neural Networks Understand Logical Entailment?
- Annotating Derivations: A New Evaluation Strategy and Dataset for Algebra Word Problems
Cited by in corpus (30)
- The Pile: An 800GB Dataset of Diverse Text for Language Modeling
- Measuring Mathematical Problem Solving With the MATH Dataset
- Scaling Laws for Autoregressive Generative Modeling
- Generative Language Modeling for Automated Theorem Proving
- Neural Symbolic Regression that Scales
- Can neural networks do arithmetic? A survey on the elementary numerical skills of state-of-the-art deep learning models
- MiniF2F: a cross-system benchmark for formal Olympiad-level mathematics
- Bongard-LOGO: A New Benchmark for Human-Level Concept Learning and Reasoning
- Compositionality decomposed: how do neural networks generalise?
- Emergent Symbols through Binding in External Memory
- Towards Question Format Independent Numerical Reasoning: A Set of Prerequisite Tasks
- AR-LSAT: Investigating Analytical Reasoning of Text
- Generating Out of Distribution Adversarial Attack using Latent Space Poisoning
- KANDINSKYPatterns -- An experimental exploration environment for Pattern Analysis and Machine Intelligence
- Human-like machine thinking: Language guided imagination
- Pointer Value Retrieval: A new benchmark for understanding the limits of neural network generalization
- Learning to solve arithmetic problems with a virtual abacus
- An Empirical Investigation of Contextualized Number Prediction
- Distilling Wikipedia mathematical knowledge into neural network models
- Towards Tractable Mathematical Reasoning: Challenges, Strategies, and Opportunities for Solving Math Word Problems
- Towards Coinductive Models for Natural Language Understanding. Bringing together Deep Learning and Deep Semantics
- Machine Number Sense: A Dataset of Visual Arithmetic Problems for Abstract and Relational Reasoning
- Latent Execution for Neural Program Synthesis
- Adversarial Examples for Evaluating Math Word Problem Solvers
- Toward Building Science Discovery Machines
- Pretrained Language Models are Symbolic Mathematics Solvers too!
- Disambiguating Symbolic Expressions in Informal Documents
- SCIMAT: Science and Mathematics Dataset
- Math Operation Embeddings for Open-ended Solution Analysis and Feedback
- Teaching Autoregressive Language Models Complex Tasks By Demonstration