When Context Sticks: Studying Interference in In-Context Learning
For researchers studying in-context learning in transformers, this work quantifies interference effects and training curriculum impacts, though it is incremental as it uses synthetic tasks and known phenomena.
This paper studies context stickiness in in-context learning, showing that earlier examples in a prompt interfere with a transformer's ability to adapt to later tasks. Using synthetic regression tasks, they find that more preceding linear examples degrade quadratic predictions, and training curricula significantly affect resilience, with sequential training yielding the fastest recovery.
This paper investigates context stickiness in in-context learning (ICL), a phenomenon where earlier examples in a prompt interfere with a transformer's ability to adapt to later tasks. Using synthetic regression tasks over linear and quadratic functions, we examine how models trained under sequential, mixed, and random curricula handle abrupt task switches during inference. By sweeping over structured combinations of misleading linear examples followed by recovery quadratic examples, we quantify how prior context biases prediction error and how quickly models realign. Our results show strong evidence of persistent interference: more preceding linear examples reliably degrade quadratic predictions, while additional quadratic examples reduce error but with diminishing returns. We further find that training curricula significantly modulate resilience, with sequential training on the target function class yielding the fastest recovery, and surprisingly, random training producing the least robust behavior.