Attention Overflow: Language Model Input Blur during Long-Context Missing Items Recommendation
This addresses a practical limitation for recommendation systems using LLMs with lengthy user histories, though it is incremental in analyzing an existing problem.
The paper identifies that large language models (LLMs) degrade in performance when suggesting missing items from long lists, starting to repeat items already in the input at around 100 items for mid-2024 models, as shown in synthetic and movie recommendation tests.
Large language models (LLMs) can suggest missing elements from items listed in a prompt, which can be used for list completion or recommendations based on users' history. However, their performance degrades when presented with too many items, as they start to suggest items already included in the input list. This occurs at around 100 items for mid-2024 flagship LLMs. We evaluate this phenomenon on both synthetic problems (e.g., finding missing numbers in a given range of shuffled integers) and realistic movie recommendation scenarios. We refer to this issue as \textit{attention overflow}, as preventing repetition requires attending to all items simultaneously. Although iterative loops can mitigate this problem, their costs increase with the repetition rate, affecting the language models' ability to derive novelty from lengthy inputs.