Fine-tuning and in-context learning (ICL) are two prevalent methods in imbuing large language models with task-specific knowledge. It is commonly believed that fine-tuning can surpass ICL given sufficient training samples as it allows the model to adjust its internal parameters based on the data. However, this paper presents a counterintuitive finding: For tasks with implicit patterns, ICL captures these patterns significantly better than fine-tuning. We developed several datasets featuring implicit patterns, such as sequences determining answers through parity or identifying reducible terms in calculations. We then evaluated the models' understanding of these patterns under both fine-tuning and ICL across models ranging from 0.5B to 7B parameters. The results indicate that models employing ICL can quickly grasp deep patterns and significantly improve accuracy. In contrast, fine-tuning, despite utilizing thousands of times more training samples than ICL, achieved only limited improvements. We also proposed circuit shift theory from a mechanistic interpretability's view to explain why ICL wins.

The why and how the human brain develops religious beliefs may stem from our ability to learn, a new study reveals. Researches found individuals who can unconsciously predict complex patterns in the environment believe there is a god who creates order and intervenes in an otherwise chaotic universe. The study used a cognitive test to measure implicit pattern learning, which showed a sequence of dots appeared and disappeared on a computer screen. Participants were told to push a button when a dot appeared, but some learned the distinct patter and were able to predict when it would appear - and sometimes before. The data showed those who noted they had faith in a higher power performed better overall during the experiment.