Probing the Capacity of Language Model Agents to Operationalize Disparate Experiential Context Despite Distraction

TL;DR

This paper evaluates the ability of state-of-the-art large language models to correctly reason over complex, distractor-laden experiential contexts, revealing significant limitations in their decision-making capabilities under challenging conditions.

Contribution

The study introduces the OEDD corpus for testing LLM reasoning with distracting information and demonstrates that current models struggle with complex, multi-faceted experiential reasoning tasks.

Findings

LLMs perform worse than random when faced with distractors in long contexts.

Performance declines significantly with increased input length and complexity.

Current LLMs have limited capacity to operationalize disparate experiential information.

Abstract

Large language model (LLM) agents show promise in an increasing number of domains. In many proposed applications, it is expected that the agent reasons over accumulated experience presented in an input prompt. We propose the OEDD (Operationalize Experience Despite Distraction) corpus, a human-annotator-validated body of scenarios with pre-scripted agent histories where the agent must make a decision based on disparate experiential information in the presence of a distractor. We evaluate three state-of-the-art LLMs (GPT-3.5 Turbo, GPT-4o, and Gemini 1.5 Pro) using a minimal chain-of-thought prompting strategy and observe that when (1) the input context contains over 1,615 tokens of historical interactions, (2) a crucially decision-informing premise is the rightful conclusion over two disparate environment premises, and (3) a trivial, but distracting red herring fact follows, all LLMs perform worse than random choice at selecting the better of two actions. Our code and test corpus are publicly available at: https://github.com/sonnygeorge/OEDD .