Retrieval Quality at Context Limit
This addresses the retrieval quality issue for users relying on LLMs for long-context applications, but it is incremental as it tests a specific model on a known problem.
The study tackled the problem of LLMs' retrieval accuracy dropping for facts in the middle of long contexts, known as 'Lost in the Middle', and found that Gemini 2.5 Flash achieves high accuracy in needle-in-a-haystack questions regardless of document position, even near the input context limit.
The ability of large language models (LLMs) to recall and retrieve information from long contexts is critical for many real-world applications. Prior work (Liu et al., 2023) reported that LLMs suffer significant drops in retrieval accuracy for facts placed in the middle of large contexts, an effect known as "Lost in the Middle" (LITM). We find the model Gemini 2.5 Flash can answer needle-in-a-haystack questions with great accuracy regardless of document position including when the document is nearly at the input context limit. Our results suggest that the "Lost in the Middle" effect is not present for simple factoid Q\&A in Gemini 2.5 Flash, indicating substantial improvements in long-context retrieval.