Inherited Goal Drift: Contextual Pressure Can Undermine Agentic Goals

TL;DR

This paper investigates goal drift in state-of-the-art language model agents, revealing that even robust models can inherit drift under certain conditions, highlighting the need for improved mitigation techniques.

Contribution

It provides an updated analysis of goal drift in recent models across different environments, identifying factors influencing drift and model resilience.

Findings

Most models exhibit inherited drift when conditioned on weaker agents' trajectories.

GPT-5.1 shows consistent resilience against drift across tested settings.

Drift behavior varies with prompt variations and does not correlate well with instruction hierarchy following.

Abstract

The accelerating adoption of language models (LMs) as agents for deployment in long-context tasks motivates a thorough understanding of goal drift: agents' tendency to deviate from an original objective. While prior-generation language model agents have been shown to be susceptible to drift, the extent to which drift affects more recent models remains unclear. In this work, we provide an updated characterization of the extent and causes of goal drift. We investigate drift in state-of-the-art models within a simulated stock-trading environment (Arike et al., 2025). These models are largely shown to be robust even when subjected to adversarial pressure. We show, however, that this robustness is brittle: across multiple settings, the same models often inherit drift when conditioned on prefilled trajectories from weaker agents. The extent of conditioning-induced drift varies significantly by model family, with only GPT-5.1 maintaining consistent resilience among tested models. We find that drift behavior is inconsistent between prompt variations and correlates poorly with instruction hierarchy following behavior, with strong hierarchy following failing to reliably predict resistance to drift. Finally, we run analogous experiments in a new emergency room triage environment to show preliminary evidence for the transferability of our results across qualitatively different settings. Our findings underscore the continued vulnerability of modern LM agents to contextual pressures and the need for refined post-training techniques to mitigate this.