Together AI 发布 YAQA 量化算法，显著降低模型输出偏差

AI 摘要

Together AI 发布了新的后训练量化方法 YAQA，通过直接最小化 KL 散度来保留原始模型输出，与现有量化器（如 QTIP）兼容，可将 KL 散度降低超过 30%，并在下游任务上取得最优性能。该方法利用 Fisher 信息矩阵和 Kronecker 分解来近似 Hessian，从而在保持模型质量的同时实现高效量化。

正文节选

We’re excited to announce YAQA (Yet Another Quantization Algorithm), a new weight-only LLM post-training quantization method that quantizes models to directly preserve the original model’s outputs. YAQA is quantizer-agnostic, meaning that it can be used with both hardware-accelerated datatypes and special memory-bound quantizers like QTIP. Across a wide range of models and quantizers, YAQA consistently reduces the KL divergence to the original model by >30% over existing rounding algorithms, res

发布时间：—

抓取时间：2026-08-03 01:13

来源机构：Together AI