Improving Coherence and Consistency in Neural Sequence Models with Dual-System, Neuro-Symbolic Reasoning

By Maxwell Nye, Michael Henry Tessler, Josh Tenenbaum, Brenden M. Lake. 2021

149 citations recorded by the source corpus.

Human reasoning can often be understood as an interplay between two systems: the intuitive and associative ("System 1") and the deliberative and logical ("System 2"). Neural sequence models -- which have been increasingly successful at performing complex, structured tasks -- exhibit the advantages and failure modes of System 1: they are fast and learn patterns from data, but are often inconsistent and incoherent. In this work, we seek a lightweight, training-free means of improving existing System 1-like sequence models by adding System 2-inspired logical reasoning. We explore several variations on this theme in which candidate generations from a neural sequence model are examined for logical consistency by a symbolic reasoning module, which can either accept or reject the generations. Our approach uses neural inference to mediate between the neural System 1 and the logical System 2. Results in robust story generation and grounded instruction-following show that this approach can increase the coherence and accuracy of neurally-based generations.

Subfields: Artificial Intelligence, Natural Language Processing, Prompting & Reasoning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

In Prompting & Reasoning

Citations in the corpus

Self-Consistency Improves Chain of Thought Reasoning in Language Models (2022, methodology): “Nye et al. (2021) improve the logical consistency of samples from a System 1 model by adding a System 2-inspired logical reasoning module.”
Selection-Inference: Exploiting Large Language Models for Interpretable Logical Reasoning (2022, background): “Although some approaches have attempted to bridge the gap between deep learning and symbolic AI by converting problems into formal logic (Nye et al., 2021b) and using existing solvers to help produce an answer, this process can be brittle and tends not to scale well.”
PAL: Program-aided Language Models (2022, methodology): “Nye et al. (2021b) first generate candidate sentences using a language model; then, another language model (GPT-3) is used to derive logical constraints entailed by each candidate; these constraints are then cross-checked with a predefined list of facts to rank the candidates.”
Human-like intuitive behavior and reasoning biases emerged in large language models but disappeared in ChatGPT (2022, methodology): “Our approach falls within a quickly growing category of studies employing classic psychological tests and experiments to probe LLM ‘psychological’ processes, such as judgment, decision-making and cognitive biases 14,22–24 .”
Evaluating Large Language Models in Generating Synthetic HCI Research Data: a Case Study (2023, background): “This can be mitigated to some degree by automatically detecting and regenerating undesired content [54, 74] and researchers are developing "debiasing" approaches [83].”
Using cognitive psychology to understand GPT-3 (2022, result): “For all three items of the CRT, GPT-3 responded with the intuitive but incorrect answer, as has been observed in earlier work 22 .”
Maieutic Prompting: Logically Consistent Reasoning with Recursive Explanations (2022, methodology): “, 2021a), or an auxiliary symbolic module for the generation to be consistent with the world model (Nye et al., 2021b) and performing discrete operations (Chen et al.”