返回
Ssignal
12
·7天前·1 signals
Archived topic · 归档话题,来源已停止追踪

Same GRPO recipe on three from-scratch LLMs (353M/316M/672M) gave three different outcomes, with no clean relationship to scale [P]

模型发布开源代码

热度趋势

趋势数据积累中

百分比基于当前可用热度信号,而非评论数或独立用户人数。