Unverified50% confidenceBenchmarkExact time
Using less than 25% of labeled data, DySENSE can generate labeled high-fidelity synthetic energy consumption data and improve downstream prediction performance by 8% mIoU.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/16/2026
First Seen
Valid until: 2/14/2027
Sources
Related Entities
Related Claims
Unverified靶向数据工程中,针对性数据占总微调数据的 5%-15% 通常能取得良好效果,过高比例可能导致模型在其他维度退化77% similarUnverifiedCompared with existing methods, DySENSE improves prediction performance by 8.7% (accuracy) and reduces prediction error by 9%–18% (CVRMSE, ASHRAE standard).72% similarUnverifiedIn the ablation study, when no data is used and only parametric memory is relied on, even the best model performs poorly; after providing early-quarter data, performance improves by about 25 percentage points; but there is still a 12-16 percentage point gap from using the latest data.72% similarUnverifiedFor small models with limited parameter counts, data quality can have an even greater impact on final performance than data quantity.72% similarUnverified参数量在1亿以下的轻量模型经领域数据微调后可达95%以上意图分类准确率,单次推理延迟控制在5-20ms以内72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1174137API
curl https://kongchang.com/api/v1/knowledge/claims/1174137MCP
get_claim(id=1174137)