Unverified50% confidenceBenchmarkExact time
研究数据表明，在数学竞赛级别问题上，启用深度推理的模型准确率可从不足30%提升至80%以上
1
Sources
50%
Confidence
Long-term
Relevance
8/8/2026
First Seen
Sources
Related Claims
UnverifiedExperiments at institutions including MIT show that Multi-Agent Debate (MAD) can improve mathematical reasoning and fact-checking accuracy by 10%–30%.74% similarUnverifiedExperimental data shows that when key information is located at the beginning and end of the input sequence, the model's retrieval accuracy can be as high as over 90%, but when located in the middle, it may drop sharply to below 60%.74% similarVerifiedWhen the draft model prediction accuracy of Speculative Decoding reaches 60–80%, the overall inference speed can be improved by 2–3 times.73% similarUnverified参数量在1亿以下的轻量模型经领域数据微调后可达95%以上意图分类准确率，单次推理延迟控制在5-20ms以内72% similarVerifiedChain-of-thought can improve accuracy by more than 40% on complex tasks such as mathematical reasoning71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/705817API
curl https://kongchang.com/api/v1/knowledge/claims/705817MCP
get_claim(id=705817)