When Agents Look the Same: Quantifying Distillation-Induced Similarity in Tool-Use Behaviors

TL;DR

This paper introduces two metrics, Response Pattern Similarity and Action Graph Similarity, to quantify behavioral homogenization in distilled language models, distinguishing mandatory task behaviors from autonomous preferences.

Contribution

It proposes novel metrics to measure non-mandatory behavioral patterns and demonstrates their effectiveness in differentiating model convergence and diversity.

Findings

Within-family model pairs have higher Action Graph Similarity than cross-family pairs.

Kimi-K2 (thinking) exceeds previous models in node and dependency similarity.

AGS effectively distinguishes teacher-specific convergence from general model improvements.

Abstract

Model distillation is a primary driver behind the rapid progress of LLM agents, yet it often leads to behavioral homogenization. Many emerging agents share nearly identical reasoning steps and failure modes, suggesting they may be distilled echoes of a few dominant teachers. Existing metrics, however, fail to distinguish mandatory behaviors required for task success from non-mandatory patterns that reflect a model's autonomous preferences. We propose two complementary metrics to isolate non-mandatory behavioral patterns: \textbf{Response Pattern Similarity (RPS)} for verbal alignment and \textbf{Action Graph Similarity (AGS)} for tool-use habits modeled as directed graphs. Evaluating 18 models from 8 providers on $\tau$-Bench and $\tau^2$-Bench against Claude Sonnet 4.5 (thinking), we find that within-family model pairs score 5.9 pp higher in AGS than cross-family pairs, and that Kimi-K2 (thinking) reaches 82.6\% $S_{\text{node}}$ and 94.7\% $S_{\text{dep}}$, exceeding Anthropic's own Opus 4.1. A controlled distillation experiment further confirms that AGS distinguishes teacher-specific convergence from general improvement. RPS and AGS capture distinct behavioral dimensions (Pearson $r$ = 0.491), providing complementary diagnostic signals for behavioral convergence in the agent ecosystem. Our code is available at https://github.com/Syuchin/AgentEcho.