Unverified50% confidenceFactExact time
Unsloth 第一时间支持了 Google DeepMind 的新多模态嵌入模型 EmbeddingGemma 2，可将文本、代码、图像、视频和音频转成向量
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/6/2026
First Seen
Valid until: 1/4/2027
Sources
Related Entities
Related Claims
UnverifiedGoogle DeepMind的Genie项目尝试从无标注视频中学习可交互的世界模型77% similarUnverifiedGoogle DeepMind推出了基于JAX的MJX（MuJoCo XLA）和Brax，将物理仿真编译为可在GPU/TPU上高效执行的计算图75% similarVerifiedGoogle DeepMind提出了Nested Learning（嵌套学习）相关方法71% similarVerifiedOpenAI、Anthropic和Google DeepMind均已在各自模型中集成了类似的智能体能力框架71% similarUnverifiedUnsloth 新增了对 DeepSeek-V4-Flash 模型的完整支持，并带有 Thinking（思维链）开关切换能力70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/981695API
curl https://kongchang.com/api/v1/knowledge/claims/981695MCP
get_claim(id=981695)