Clarification on usage of Infinite Retrieval paper

[whisper-bye]

The README references the paper "Infinite Retrieval: Attention Enhanced LLMs in Long-Context Processing". Could you please clarify where and how the techniques from this paper are used in the project?

[SergiiShcherbak]

@whisper-bye The paper "Infinite Retrieval: Attention Enhanced LLMs in Long-Context Processing" is linked in the README's "Focused document analysis" section to illustrate the broader trend of research aimed at improving long-context capabilities in LLMs. This aligns with ContextGem’s architectural choice of leveraging LLMs' native long context windows directly - rather than relying on RAG techniques - for focused, document-level information extraction.

To clarify: ContextGem does not implement the InfiniRetri method or modify underlying LLM attention mechanisms, as it is a model-agnostic framework designed to work with existing LLM APIs. The techniques described in the paper are relevant for LLM providers and researchers working at the model architecture level, not at the orchestration or framework level.

In light of your feedback, it seems that linking the paper in this context could cause confusion, and I’ll plan to remove that link in a future documentation update. Thanks for pointing this out.