Unverified50% confidenceSolutionExact time
The Motif team encountered abnormal activation issues in the early stages of pretraining, the root cause being anomalous weight values (such as 2 or -2) caused by weight initialization, which was resolved after fixing the initialization method.
1
Sources
50%
Confidence
Long-term
Relevance
10/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn the post-training phase, Motif adopts a "failure-driven SFT" strategy: identifying model weaknesses through benchmark testing and internal scenarios, then generating targeted datasets for training.68% similarUnverified不同模型在抗诱导能力上的差异很大程度上源于后训练阶段（post-training）的策略差异62% similarVerifiedMethods for addressing overfitting include increasing training data, regularization, early stopping, and Dropout.62% similarUnverifiedRLHF和DPO等对齐技术的效果高度依赖预训练基础质量，预训练阶段未习得的能力几乎无法通过后续对齐补救62% similarUnverified解决灾难性遗忘的方案是在训练时混入10%-20%的原始预训练数据62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/973084API
curl https://kongchang.com/api/v1/knowledge/claims/973084MCP
get_claim(id=973084)