Debiasing Large Language Models via Adaptive Causal Prompting with Sketch-of-Thought

TL;DR

This paper introduces ACPS, a novel prompting framework that uses causal models and concise reasoning sketches to improve large language model reasoning, reducing token usage and enhancing generalisability across tasks.

Contribution

The paper proposes a new adaptive causal prompting method with Sketch-of-Thought that improves reasoning efficiency and generalisability without task-specific retraining.

Findings

ACPS outperforms existing prompts in accuracy and robustness.

It significantly reduces token usage and inference costs.

Demonstrates effectiveness across multiple benchmarks and models.

Abstract

Despite notable advancements in prompting methods for Large Language Models (LLMs), such as Chain-of-Thought (CoT), existing strategies still suffer from excessive token usage and limited generalisability across diverse reasoning tasks. To address these limitations, we propose an Adaptive Causal Prompting with Sketch-of-Thought (ACPS) framework, which leverages structural causal models to infer the causal effect of a query on its answer and adaptively select an appropriate intervention (i.e., standard front-door and conditional front-door adjustments). This design enables generalisable causal reasoning across heterogeneous tasks without task-specific retraining. By replacing verbose CoT with concise Sketch-of-Thought, ACPS enables efficient reasoning that significantly reduces token usage and inference cost. Extensive experiments on multiple reasoning benchmarks and LLMs demonstrate that ACPS consistently outperforms existing prompting baselines in terms of accuracy, robustness, and computational efficiency.