This is the code repository of Insights into LLM Long-Context Failures: When Transformers Know but Don't Tell. [EMNLP 2024 Findings]

Large Language Models (LLMs) exhibit positional bias, struggling to utilize information from the middle or end of long contexts. Our study explores LLMs' long-context reasoning by probing their hidden representations. We find that while LLMs encode the position of target information, they often fail to leverage this in generating accurate responses. This reveals a disconnect between information retrieval and utilization, a "know but don't tell" phenomenon. We further analyze the relationship between extraction time and final accuracy, offering insights into the underlying mechanics of transformer models.

If you find this work helpful, please consider citing the following:

@misc{lu2024insightsllmlongcontextfailures,
      title={Insights into LLM Long-Context Failures: When Transformers Know but Don't Tell}, 
      author={Taiming Lu and Muhan Gao and Kuai Yu and Adam Byerly and Daniel Khashabi},
      year={2024},
      eprint={2406.14673},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2406.14673}, 
}