Unverified50% confidenceFactExact time
DeepSeek 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
1
Sources
50%
Confidence
Long-term
Relevance
7/18/2026
First Seen
Sources
Related Claims
Cite This Claim
Stable URI
https://kongchang.com/claim/549627API
curl https://kongchang.com/api/v1/knowledge/claims/549627MCP
get_claim(id=549627)