Unverified50% confidenceFactExact time
Wonder Full's figures used xhigh reasoning effort.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/12/2026
First Seen
Valid until: 1/10/2027
Sources
Related Entities
Related Claims
UnverifiedXH mode has a noticeably longer thinking time than HIGH mode, corresponding to the Test-Time Compute research direction.64% similarUnverified对于逻辑推理任务，以更多Token和算力换取逻辑严密性的XH模式权衡通常是值得的63% similarUnverifiedThe core idea of inference-time compute is to have the model generate longer chains of thought at inference time, improving accuracy on complex tasks through self-verification and multi-path search, which has a methodological connection to the success of Monte Carlo Tree Search in game-playing AI—AlphaGo surpassed human players through large-scale tree search at inference time.61% similarUnverifiedGrok's Thinking mode is a reasoning-enhanced mode under xAI.60% similarUnverified一些推理增强型模型（如DeepSeek-R1、Kimi K3）会生成非常长的思维链，有时推理过程的Token数量是最终答案的数十倍57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1011233API
curl https://kongchang.com/api/v1/knowledge/claims/1011233MCP
get_claim(id=1011233)