Therefore I am. I Think

📰 ArXiv cs.AI

Research suggests large language models make decisions before generating reasoning chains

advanced Published 2 Apr 2026

Action Steps

Investigate pre-generation activations in large language models to understand decision-making processes
Use linear probes to decode tool-calling decisions from these activations
Analyze the timing of decision-making in relation to reasoning chain generation
Evaluate the implications of early-encoded decisions on model transparency and explainability

Who Needs to Know This

AI researchers and engineers can benefit from understanding how language models make decisions, which can inform the development of more transparent and explainable AI systems

Key Insight

💡 Detectable, early-encoded decisions shape chain-of-thought in reasoning models

Key Takeaways

Research suggests large language models make decisions before generating reasoning chains

Full Article

Title: Therefore I am. I Think

Abstract:
arXiv:2604.01202v1 Announce Type: new Abstract: We consider the question: when a large language reasoning model makes a choice, did it think first and then decide to, or decide first and then think? In this paper, we present evidence that detectable, early-encoded decisions shape chain-of-thought in reasoning models. Specifically, we show that a simple linear probe successfully decodes tool-calling decisions from pre-generation activations with very high confidence, and in some cases, even befor

Read full paper → ← Back to Reads