Dissecting model behavior through agent trajectories

2026-06-15

AI Summary

This work addresses the performance gap between intent and execution that hinders large language models in autonomous programming by introducing the customizable Simple Strands Agent (SSA) framework, which uniformly supports mainstream models including Claude, Gemini, GPT, Grok, and Qwen. Leveraging 138,000 agent trajectories, we propose fine-grained behavioral metrics—such as edit frequency, test activity, and phase transitions—within the code state space, moving beyond conventional pass@1 evaluation paradigms. Our experiments reproduce or surpass official model results on benchmarks including SWE-Pro, SWE-Verified, and Terminal-Bench-2, while systematically uncovering distinct problem-solving strategies employed by different state-of-the-art models.

Abstract

AI agent performance is not just a modeling problem, it is fundamentally a systems problem. The advanced capabilities of models are realized through agent harnesses. Therefore, a gap between model assumptions and harness behavior can easily prevent the model's full capabilities from translating into agent performance. We formalize this as the `intent-execution' gap: the mismatch between what the model intends and what the harness executes, and vice versa. We argue that minimizing this intent-execution gap is as important as other aspects of harness design such as tools and execution loops. To illustrate the impact of this harness-model alignment, we develop a simple and customizable harness called `Simple Strands Agent' (SSA). SSA aims to find the bulk of common patterns which generalize across different model families (such as Claude, Gemini, GPT, Grok, Qwen), as well as a small number of model-specific preferences. We make two contributions: (i) we $\textbf{reproduce or improve on the pass@1}$ performance reported by diverse model-provider families on popular agentic benchmarks (SWE-Pro, SWE-Verified and Terminal-Bench-2), and (ii) building on an $\textbf{analysis of 138k trajectories generated by SSA}$, we look beyond the $\texttt{pass@1}$ numbers which tend to be relatively even across frontier models. By representing agent trajectories in code state-spaces, we observe model-level differences in problem-solving behavior. Finer-grained metrics such as edit frequency, testing activity, and phase-transitions reveal how individual models allocate effort across different stages of autonomous problem solving.

Problem

• intent-execution gap
• agent trajectories
• model-harness alignment
• AI agent performance
• autonomous problem solving

Innovation

• intent-execution gap
• agent trajectories
• harness-model alignment
• code state-spaces
• autonomous problem solving