关于training phase1的任务

[anthony123]

你好，请教一下，放出的代码，training phase1只训练了Interleaved User Persona Grounding data 中的任务吗？

[Silent-majority]

为啥看代码只有Sequential Preference Modeling

[zhangrz22]

@anthony123
In industrial settings, Phase 1 uses Interleaved User Persona Grounding data to align the model with user-specific contexts.
However, for the open-source dataset, we only utilized the user interaction sequences for training.