ENTITY PROFILE / 实体档案
LoRA GRPO
LoRA GRPO, an LoRA reinforcement‐learning from human feedback method using VERL, was evaluated with AIME‑2024, boosting mean@8 from 0.56 to 0.85, with training–rollout KL ≈2×10⁻⁴. Funders VC backed Berachain NGRAVE followed a similar phishing campaign that targeted SecuX users on Oct. 16.· Graphiti 生成的实体摘要
RELATIONS
2关系CURRENT
2当前有效EVENTS
2相关事件REPORTS
2来源报道暂无可公开的邻近关系
按需展开实体的全部已公开关系，保留探索过的节点与连线。
关系变化
按生效时间- NGRAVEFOLLOWED SIMILAR PHISHING CAMPAIGN AGAINSTLoRA GRPO
NGRAVE followed a similar phishing campaign that targeted SecuX users on Oct. 16.
- JSCBACKEDLoRA GRPO
Funders VC backed Berachain
关系是 Graphiti 从报道中提取的事实陈述，不是原文引述；失效时间表示后来的报道显示这条关系已不成立。