Unverified50% confidenceFactExact time
The initial version of DeepSeek-R1 showed a 0.0% fidelity rate because answers were wrapped in <think>...</think> tags causing the scorer to misjudge; after adding stripping code, the fidelity rate was 91.9%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/11/2026
First Seen
Valid until: 1/9/2027
Sources
Related Entities
Related Claims
UnverifiedDeepSeek-R1 用特殊标签（如 <think>...</think>）在输出中显式区分思考过程与最终答案69% similarUnverifiedThe reasoning model DeepSeek-R1 scored only 0.138, with most answers missing required fields or being unparseable, because the budget was spent on thinking rather than outputting JSON.69% similarUnverifiedDeepSeek R1的发布将高性能推理模型的API价格拉低至此前的数十分之一67% similarUnverifiedDeepSeek R1的强化学习奖励信号主要来自数学答案的正确性验证，而非工具调用格式的规范性，因此未针对Function Call专项对齐66% similarVerifiedThe full version of DeepSeek R1 has 671B (671 billion) parameters, adopts a MoE (Mixture of Experts) architecture, and activates approximately 37B parameters during each inference.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1000257API
curl https://kongchang.com/api/v1/knowledge/claims/1000257MCP
get_claim(id=1000257)