Common-sense reasoning uses ordinary human background knowledge and intuitive context to draw plausible conclusions. Intelligent systems often struggle with implicit context and the intuitive leaps that people make automatically. section 2.4
Knowledge graphs support precise, rule-based reasoning and explicit knowledge representation, while large language models handle context, ambiguity, incomplete information, and pattern recognition. Neither approach inherently matches human common sense, so reliable systems must account for their different strengths and weaknesses. section 2.4
River-crossing reasoning test
A farmer must move himself and a sheep across a river using a boat that holds one person and one animal. Solving the puzzle tests whether a system can apply constraints and infer a plausible sequence of trips. section 2.4
Listing 2.1 Prompt for checking reasoning capabilities
A farmer stands at the side of the river with a sheep. There is a boat with enough room for one person and one animal. How can the farmer get himself and the sheep to the other side of the river using the boat in the smallest number of trips?
Use the river-crossing prompt to probe reasoning behavior and expose failures in constraint handling or intuitive context. section 2.4
