Unverified50% confidenceFactExact time
ML4's training was completed in Europe, using about 3,800 NVIDIA Grace Blackwell GPUs.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTraining simultaneously across 4 harnesses improved Liquid AI's LFM2.5-2.6B model score from 42% to 54%.53% similarUnverified以 GPT-3 为例，r=4 时 LoRA 可将可训练参数从 175B 降至约 4.7M，减少约 37,000 倍53% similarUnverified一个rank=4的LoRA文件对于SD1.5模型约只有3-6MB，LoRA设计使训练参数量减少95%以上52% similarUnverifiedGLM-5.3-Flash 在 Toolathlon Verified 得78.4分，OfficeQA Pro 得62.4分，均领先 Claude Opus 4.8。52% similarUnverifiedTaking GPT-3's attention layers as an example, the original weight matrix is approximately 12,288×12,288 (about 150 million parameters), while a LoRA decomposition with rank r=8 requires training only about 200,000 parameters, a compression ratio exceeding 700×.52% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1009651API
curl https://kongchang.com/api/v1/knowledge/claims/1009651MCP
get_claim(id=1009651)