Unverified50% confidenceBenchmarkExact time
After training a 24-layer, 10B-A1B model on 708B tokens, the full Multi-View MoE improves the average score on nine downstream tasks by 1.18 percentage points over the baseline.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/30/2026
First Seen
Valid until: 1/28/2027
Sources
Related Entities
Related Claims
UnverifiedMulti-View MoE has so far been validated only at the 10B-A1B scale, and whether the gains hold for larger models and larger numbers of routers remains to be seen.73% similarUnverifiedMoE models activate only approximately 10 out of 512 experts per token, with an effective activated parameter count of approximately 20–30B.66% similarUnverifiedMoE架构每次推理仅激活少数专家，训练成本相对同等能力稠密模型大幅降低，10B激活1B相比10B稠密模型训练成本低10倍且推理更快65% similarUnverifiedThe figure "pretraining throughput reaching 170% of a single H100" is jointly affected by batch size, sequence length, model scale, precision format, and interconnect bandwidth, and is not a general conclusion; the ratio may change with a different model or hyperparameters.65% similarUnverified在100帧生成任务的FVD指标上，欧拉运动引导取得76.18，优于最强基线的79.2064% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1115291API
curl https://kongchang.com/api/v1/knowledge/claims/1115291MCP
get_claim(id=1115291)