{"ID":2884485,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2508.05905","arxiv_id":"2508.05905","title":"The Fourth State: Signed-Zero Ternary for Stable LLM Quantization (and More)","abstract":"Quantization is usually regarded as a means to trade quality of performance for reduced compute requirements, i.e., as a suboptimal approximation. However, if examined in terms of a fixed overall resource budget, a very different perspective arises. We introduce Signed-Zero Ternary (SZT), a 2-bit quantization that deterministically provides gradient information with no forward-path penalty. Our analysis provides evidence that it may improve information density compared to non-quantized alternatives.","short_abstract":"Quantization is usually regarded as a means to trade quality of performance for reduced compute requirements, i.e., as a suboptimal approximation. However, if examined in terms of a fixed overall resource budget, a very different perspective arises. We introduce Signed-Zero Ternary (SZT), a 2-bit quantization that dete...","url_abs":"https://arxiv.org/abs/2508.05905","url_pdf":"https://arxiv.org/pdf/2508.05905v1","authors":"[\"Jeffrey Uhlmann\"]","published":"2025-08-08T00:01:29Z","proceeding":"cs.LG","tasks":"[\"cs.LG\"]","methods":"[\"Large Language Model\"]","has_code":false}