Depth recurrent model transfers knowledge across new rule sets in puzzles
Knowledge transfer between simple discrete worlds
Artificial Intelligence
Summary
Understanding how to adapt to new rules is a basic part of intelligence. The authors studied this by training a model to solve puzzles like Sudoku, but with different combinations of rules. They tested how well the model could handle new puzzles, unseen rule combinations, or entirely new rules based on descriptions or examples. They found that descriptions worked best when recombining known rules, but examples helped more when the model encountered completely new rule instances. This research offers a way to better test how models learn and generalize in reasoning tasks.
What this means in practice
- •For ai system developers: Build AI agents that adapt to new rule-based tasks by training on combinations of simple rules and using rule descriptions or examples for rapid generalization.
- •For game designers: Create puzzle games with flexible rule sets that AI opponents can learn and adapt to quickly using combined rule information.
Authors
Silvère Gangloff, Jan Hula
Abstract
Adapting to a new world and reasoning correctly about which outcomes are possible under rules one has not met before is a basic capacity underlying general intelligence. We study it in small discrete worlds: sets of grid configurations consistent with a shared set of rules, concretely Sudoku-like puzzles built from a common base and a handful of extra rules. We train a single depth-recurrent model on many combinations of rules and test it at four levels of novelty: a new puzzle, a new combination of known rules, a pairing of rules never seen together, and a rule instance not seen at all during training. The model receives the active rules either as a symbolic description or as two solved examples, and a model that gets no rule information is used as a baseline. At recombining rules it was trained on, the description matches or exceeds examples ($99\%$ against $95\%$ on new combinations). On rule instances whose components never co-occurred in training, the description falls below the no-information baseline while examples stay above it. The framework allows us to test many such generalization regimes in more systematic way than benchmarks such as ARC-AGI and provides new ground for evaluating depth-recurrent reasoning models, which are typically tested on saturated Sudoku benchmarks.