Abstract--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. Large language models (LLMs) have rapidly advanced in their ability to process and reason over long textual contexts, enabling applications in summarization, retrieval-augmented Q&A, document understanding, and more.

Despite recent advances in Vision-Language Models (VLMs), long-video understanding remains a challenging problem. Although state-of-the-art long-context VLMs can process around 1000 input frames, they still struggle to effectively leverage this sequence length, and succumb to irrelevant distractors within the context window. We present Temporal Chain of Thought, an inference strategy for video question-answering that curates the model's input context. We use the VLM itself to iteratively identify and extract the most relevant frames from the video, which are then used for answering. We demonstrate how leveraging more computation at inference-time to select the most relevant context leads to improvements in accuracy, in agreement with recent work on inference-time scaling of LLMs. Moreover, we achieve state-of-the-art results on 4 diverse video question-answering datasets, showing consistent improvements with 3 different VLMs. In particular, our method shines on longer videos which would not otherwise fit within the model's context window: On longer videos of more than 1 hour on LVBench, our approach using a context window of 32K outperforms the same VLM using standard inference with a 700K context window by 2.8 points.

Large language models (LLMs) have become essential tools in computer science, especially for tasks involving code understanding and generation. However, existing work does not address many of the unique challenges presented by code written for government applications. In particular, government enterprise software is often written in legacy languages like MUMPS or assembly language code (ALC) and the overall token lengths of these systems exceed the context window size for current commercially available LLMs. Additionally, LLMs are primarily trained on modern software languages and have undergone limited testing with legacy languages, making their ability to understand legacy languages unknown and, hence, an area for empirical study. This paper examines the application of LLMs in the modernization of legacy government code written in ALC and MUMPS, addressing the challenges of input limitations. We investigate various code-chunking methods to optimize the generation of summary module comments for legacy code files, evaluating the impact of code-chunking methods on the quality of documentation produced by different LLMs, including GPT-4o, Claude 3 Sonnet, Mixtral, and Llama 3. Our results indicate that LLMs can select partition points closely aligned with human expert partitioning. We also find that chunking approaches have significant impact on downstream tasks such as documentation generation. LLM-created partitions produce comments that are up to 20% more factual and up to 10% more useful than when humans create partitions. Therefore, we conclude that LLMs can be used as suitable replacements for human partitioning of large codebases during LLM-aided modernization.

As the context limits of Large Language Models (LLMs) increase, the range of possible applications and downstream functions broadens. In many real-world tasks, decisions depend on details scattered across collections of often disparate documents containing mostly irrelevant information. Long-context LLMs appear well-suited to this form of complex information retrieval and reasoning, which has traditionally proven costly and time-consuming. However, although the development of longer context models has seen rapid gains in recent years, our understanding of how effectively LLMs use their context has not kept pace. To address this, we conduct a set of retrieval experiments designed to evaluate the capabilities of 17 leading LLMs, such as their ability to follow threads of information through the context window. Strikingly, we find that many models are remarkably threadsafe: capable of simultaneously following multiple threads without significant loss in performance. Still, for many models, we find the effective context limit is significantly shorter than the supported context length, with accuracy decreasing as the context window grows. Our study also highlights the important point that token counts from different tokenizers should not be directly compared -- they often correspond to substantially different numbers of written characters. We release our code and long-context experimental data.

Large language models (LLMs) have shown remarkable performance in various natural language processing tasks. However, a primary constraint they face is the context limit, i.e., the maximum number of tokens they can process. Previous works have explored architectural changes and modifications in positional encoding to relax the constraint, but they often require expensive training or do not address the computational demands of self-attention. In this paper, we present Hierarchical cOntext MERging (HOMER), a new training-free scheme designed to overcome the limitations. HOMER uses a divide-and-conquer algorithm, dividing long inputs into manageable chunks. Each chunk is then processed collectively, employing a hierarchical strategy that merges adjacent chunks at progressive transformer layers. A token reduction technique precedes each merging, ensuring memory usage efficiency. We also propose an optimized computational order reducing the memory requirement to logarithmically scale with respect to input length, making it especially favorable for environments with tight memory restrictions. Our experiments demonstrate the proposed method's superior performance and memory efficiency, enabling the broader use of LLMs in contexts requiring extended context. Code is available at https://github.com/alinlab/HOMER.