Lossless at What Length? Near-Lossless KV Cache Compression Is a Claim Indexed by Decode Exposure, Scoring Statistic and Budget Accounting, the Decode-Aware Line Has Answered the First Index Mostly on Final-Answer Accuracy, and Divergence From the Full Cache Is Not by Itself the Defect

APA

Mahendrakar, P. (2026). Lossless at What Length? Near-Lossless KV Cache Compression Is a Claim Indexed by Decode Exposure, Scoring Statistic and Budget Accounting, the Decode-Aware Line Has Answered the First Index Mostly on Final-Answer Accuracy, and Divergence From the Full Cache Is Not by Itself the Defect [Journal article]. Life of Research. https://doi.org/10.5281/zenodo.23343116

Copy APA

BibTeX

@misc{lor1062026,
  title        = {Lossless at What Length? Near-Lossless KV Cache Compression Is a Claim Indexed by Decode Exposure, Scoring Statistic and Budget Accounting, the Decode-Aware Line Has Answered the First Index Mostly on Final-Answer Accuracy, and Divergence From the Full Cache Is Not by Itself the Defect},
  author       = {Pranay Mahendrakar},
  year         = {2026},
  publisher    = {Life of Research},
  doi          = {10.5281/zenodo.23343116},
  url          = {https://lifeofresearch.com/articles/106},
  keywords     = {KV cache compression, token eviction, KV cache quantization, long-context inference, reasoning models, evaluation methodology, instruction following, LLM serving},
  note         = {LOR-00106}
}

Copy BibTeX More on the record page