Iterative Forward Tuning Boosts In-Context Learning in Language Models

TL;DR

This paper introduces a two-stage iterative framework called Deep-Thinking that enhances in-context learning in large language models by allowing multiple rounds of information processing without training, leading to improved performance.

Contribution

The study proposes a novel two-stage framework with an iterative attention mechanism that significantly boosts in-context learning performance in LLMs.

Findings

Outperforms vanilla ICL methods across benchmarks

Effective in tasks with difficult demonstration selection

Enhances understanding without additional training

Abstract

Despite the advancements in in-context learning (ICL) for large language models (LLMs), current research centers on specific prompt engineering, such as demonstration selection, with the expectation that a single iteration of demonstrations processing can generalize effectively to a given test sample. However, this perspective overlooks the potential benefits derived from multiple iterations involving demonstrations, a practice aligning more closely with the iterative decision-making process exhibited by humans, who often learn through analogy. In this study, we introduce a novel two-stage framework to boost ICL in LLMs. Specifically, our framework delineates the ICL process into two distinct stages: Deep-Thinking and test stages. The Deep-Thinking stage incorporates a unique attention mechanism, i.e., iterative enhanced attention, which enables multiple rounds of information accumulation. This mechanism operates by manipulating the Key-Value matrices without training, fostering enhanced understanding capabilities in LLMs by thinking demonstrations multiple times. We evaluated Deep-Thinking across a range of benchmarks and LLMs, showing its superior performance over vanilla ICL methods and its effectiveness in challenging tasks where demonstration selection is infeasible.