Unverified70% confidenceFactExact time

研究发现CoT推理能力只在参数量超过约100B的超大规模模型中自发涌现

2

Sources

70%

Confidence

Long-term

Relevance

7/10/2026

First Seen

Sources

GPT-5.6三模型架构解析：Sol/Terra/Luna定价与性能全览

bilibiliDP_IT7/9/2026

Related Claims

UnverifiedCoT效果在参数量超过约100B的模型上尤为显著，体现为「涌现能力」81% similar Unverified研究表明CoT提示在参数量低于100B的模型上几乎无效81% similar VerifiedCoT的涌现性表现为仅当模型参数量超过约1000亿时才能显著提升性能，小模型上几乎无效78% similar Verified思维链效果在参数量超过100B的模型上尤为显著71% similar Unverified参数量更大、推理步骤更多的模型，其GPU算力消耗可能是轻量模型的10-50倍68% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/470873

API

curl https://kongchang.com/api/v1/knowledge/claims/470873

MCP

get_claim(id=470873)