Beyond the Answer: Decoding the Behavior of LLMs as Scientific Reasoners

TL;DR

This paper explores how prompt optimization influences the reasoning behavior of LLMs, revealing model-specific heuristics and emphasizing the importance of interpretability for safe and effective collaboration.

Contribution

It introduces a systematic prompt optimization method using Genetic Pareto to analyze reasoning heuristics and transferability across LLMs.

Findings

Prompt gains often relate to model-specific heuristics.

Optimized prompts reveal local, non-generalizable logic.

Understanding these heuristics aids interpretability and collaboration.

Abstract

As Large Language Models (LLMs) achieve increasingly sophisticated performance on complex reasoning tasks, current architectures serve as critical proxies for the internal heuristics of frontier models. Characterizing emergent reasoning is vital for long-term interpretability and safety. Furthermore, understanding how prompting modulates these processes is essential, as natural language will likely be the primary interface for interacting with AGI systems. In this work, we use a custom variant of Genetic Pareto (GEPA) to systematically optimize prompts for scientific reasoning tasks, and analyze how prompting can affect reasoning behavior. We investigate the structural patterns and logical heuristics inherent in GEPA-optimized prompts, and evaluate their transferability and brittleness. Our findings reveal that gains in scientific reasoning often correspond to model-specific heuristics that fail to generalize across systems, which we call "local" logic. By framing prompt optimization as a tool for model interpretability, we argue that mapping these preferred reasoning structures for LLMs is an important prerequisite for effectively collaborating with superhuman intelligence.