Unverified50% confidenceFactExact time
DeepSeek-V2引入了无辅助损失的负载均衡策略以减少训练干扰
1
Sources
50%
Confidence
Long-term
Relevance
7/24/2026
First Seen
Sources
Related Claims
UnverifiedDeepSeek V3 introduced an FP8 mixed-precision training framework during the pre-training phase, reducing training memory usage and communication overhead.78% similarVerifiedDeepSeek achieves near top-tier performance while significantly reducing training costs through a Mixture of Experts (MoE) architecture and reinforcement learning optimization.75% similarVerifiedDeepSeek introduced innovations such as auxiliary-loss-free load balancing in its MoE routing strategy, improving expert utilization.74% similarUnverifiedDeepSeek-V3以极低的训练成本实现了顶尖推理能力73% similarVerifiedDeepSeek-V3 uses dynamic bias terms rather than additional loss functions to achieve load balancing, avoiding the negative impact of auxiliary losses on model quality.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/604702API
curl https://kongchang.com/api/v1/knowledge/claims/604702MCP
get_claim(id=604702)