Unverified50% confidenceOpinionExact time
In the context of large models, '10T' may also refer to the total number of tokens in training data rather than the parameter count.
1
Sources
50%
Confidence
Long-term
Relevance
9/12/2026
First Seen
Sources
Related Entities
Related Claims
Unverified针对有限意图集合训练的轻量级专用模型参数量通常在10M以下，其推理延迟可比通用LLM调用低2-3个数量级74% similarUnverifiedMicrosoft's Phi series of small models are trained on carefully designed synthetic 'textbook-quality' data, surpassing models ten times their size on multiple benchmarks with fewer than 3B parameters.69% similarUnverifiedFine-tuning on a few hundred real data samples on top of synthetic pre-training can often improve mAP by 10 to 20 percentage points.67% similarUnverifiedDiffusionGemma的监督微调步骤成本不到原模型训练总开销的10%66% similarUnverifiedResearch shows that on certain open-source benchmarks, the n-gram overlap rate between the training data of leading large models and the evaluation data can be as high as 15%–30%.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/913789API
curl https://kongchang.com/api/v1/knowledge/claims/913789MCP
get_claim(id=913789)