Unverified50% confidenceFactExact time
自2024年以来，多项研究和产业实践显示单纯增加参数量带来的收益正在递减，训练成本呈指数增长而基准测试分数提升趋于平缓
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/28/2026
First Seen
Valid until: 11/26/2026
Sources
Related Entities
Related Claims
UnverifiedSince the second half of 2024, the industry has broadly observed that the improvement margins of mainstream large models on standard evaluation benchmarks have narrowed significantly, a phenomenon referred to as 'Scaling Law deceleration.'69% similarUnverifiedA 2024 study indicates that even when AI-generated content accounts for only 10–20% of training data, measurable quality degradation can occur after multiple generations of training.68% similarUnverified测试时计算扩展的边际收益并非线性，存在明显的收益递减区间67% similarUnverified自回归缩放定律的边际收益正在递减，单纯增加模型规模和数据量带来的提升已微乎其微67% similarUnverifiedIndustry estimates suggest that the inference cost of mainstream LLMs drops approximately 75% every 18 months, but demand is growing faster, causing total costs to continue rising67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/815973API
curl https://kongchang.com/api/v1/knowledge/claims/815973MCP
get_claim(id=815973)