Cognitive Loop via In-Situ Optimization: Self-Adaptive Reasoning for Science

TL;DR

This paper introduces CLIO, a self-adaptive reasoning framework for AI that enhances transparency, steerability, and accuracy in scientific discovery by enabling models to self-formulate, adapt, and provide insights into their reasoning process.

Contribution

The paper presents CLIO, a novel in-situ optimization approach allowing large language models to self-regulate reasoning, improve accuracy, and offer transparency without additional training.

Findings

CLIO improves GPT-4.1's accuracy by 13.82% on biology and medicine questions.

Oscillations in internal uncertainty measures correlate with answer accuracy.

CLIO's open design enables better understanding and control of AI reasoning processes.

Abstract

The capacity for artificial intelligence (AI) to formulate, evolve, and test altered thought patterns under dynamic conditions indicates advanced cognition that is crucial for scientific discovery. The existing AI development landscape falls into two categories: 1) frameworks over non-reasoning models that natively incorporate opinions on how humans think, and 2) reasoning models that abstract precise control of the reasoning intuition away from end users. While powerful, for scientists to maximize utility of AI in scientific discovery, they not only require accuracy and transparency in reasoning, but also steerability. Hence, we introduce an alternative approach that enables deep and precise control over the reasoning process called: a cognitive loop via in-situ optimization (CLIO). CLIO enables large language models (LLMs) to self-formulate ways of approaching a problem, adapt behavior when self-confidence is low, and ultimately provide scientists with a final belief or answer. Through CLIO's open design, scientists can observe uncertainty levels, understand how final belief states are formulated using graph structures, and interject corrections. Without any further post-training, OpenAI's GPT-4.1 with CLIO yields an accuracy of 22.37\% in text-based biology and medicine questions on Humanity's Last Exam (HLE). This yields a 13.82\% net or 161.64\% relative increase when compared to the base GPT-4.1 model and surpasses OpenAI's o3 performance in high and low reasoning effort modes. We further discovered that oscillations within internal uncertainty measures are key in determining the accuracy of CLIO's results, revealing how its open design and internal mechanisms can provide insight and control into scientific decision-making processes.