Expired50% confidenceBenchmarkExact time
In actual testing, the acceptance length for GSM8K was 6 (the paper-reported value is 6.1), and the acceptance length for MTBench was 3.65 (the paper value is 3.64), highly consistent with the official data.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/9/2026
First Seen
Valid until: 10/7/2026(expired)
Sources
DeepSeek DSpec Hands-On: How Speculative Decoding Accelerates LLM Inference by 6x
bilibili程序员-智能译站7/8/2026
Related Claims
Unverified在 GSM8K 基准上，标准思维链准确率约为 56%-63%（GPT-3 规模），PAL 可提升至 72% 以上62% similarUnverifiedGLM-5 scored 77.8% on the SWE-Bench Verified benchmark, placing it in the top tier among open-source models.59% similarUnverifiedGSM8K发布时GPT-3的zero-shot准确率仅约5%，加入few-shot示例也不超过35%58% similarVerifiedIn the Terminal Bench 2.0 test, K2.6 scored 66.7, slightly higher than both GPT-5.4 and Claude Opus 4.6.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/362549API
curl https://kongchang.com/api/v1/knowledge/claims/362549MCP
get_claim(id=362549)