Computing with speculations

State of Thought Enables Endogenous Reasoning

Test-time compute has emerged as a major approach to improving the capabilities of Large Language Models (LLMs). However, existing test-time reasoning paradigms rely heavily on externally imposed control, either through fixed reasoning programs or through costly expansion in constrained search spaces, limiting both gen...

A Computational Model of Backward Reasoning in Human Problem Solving

When solving a problem with a clear goal, people often break down the problem into subgoals, a form of reasoning known as backward reasoning. The cognitive mechanisms of backward reasoning are poorly understood. Inspired by classic ideas from Newell and Simon, we conceptualize backward reasoning as search through a s...

A Minimal Framework for Knowledge Representation Sufficient to Support Analogical Reasoning and Other Forms of Inductive Inference

Induction and Inquiry via Probabilistic Reasoning over Language and Code

The results suggest that one way humans continually grow their knowledge is by mentally representing many hypotheses spanning language-like and program-like representations, then revising those hypotheses to approximate Bayesian updates, while a bottom-up neural mechanism (an LLM) makes inference both tractable and lea...

Emergent Reasoning in Large Language Models: A Systematic Evaluation Across Task Complexity

It is argued that emergent reasoning in LLMs is a product of three factors: model size, prompting approach, and evaluation metric, and proposed implications for designing benchmarks and assessing capabilities are proposed.

Hegelian Self-Reflection for Improved Mathematical and Symbolic Reasoning, and Scientific Discovery in Large Language Models