A Gap to a Reference Is Not a Loss: What Text-Capability Comparisons Between Vision-Language Models and Their Language Backbones Measure, Why Independent Audits and Vendor Reports Disagree in Sign, and the Controlled Trajectory That Would Assign the Cause

APA

Mahendrakar, P. (2026). A Gap to a Reference Is Not a Loss: What Text-Capability Comparisons Between Vision-Language Models and Their Language Backbones Measure, Why Independent Audits and Vendor Reports Disagree in Sign, and the Controlled Trajectory That Would Assign the Cause [Preprint]. Life of Research. https://doi.org/10.5281/zenodo.23858880

Copy APA

BibTeX

@misc{lor1792026,
  title        = {A Gap to a Reference Is Not a Loss: What Text-Capability Comparisons Between Vision-Language Models and Their Language Backbones Measure, Why Independent Audits and Vendor Reports Disagree in Sign, and the Controlled Trajectory That Would Assign the Cause},
  author       = {Pranay Mahendrakar},
  year         = {2026},
  publisher    = {Life of Research},
  doi          = {10.5281/zenodo.23858880},
  url          = {https://lifeofresearch.com/preprints/179},
  keywords     = {vision-language models, catastrophic forgetting, instruction following, attention sinks, QK normalization, evaluation methodology, multimodal training data, model merging},
  note         = {LOR-00179}
}

Copy BibTeX More on the record page