Unverified50% confidenceBenchmarkExact time
Saluki 27B shows a regression compared to the original in competitive mathematics and complex reasoning tasks
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/11/2026
First Seen
Valid until: 1/9/2027
Sources
Related Entities
Related Claims
Unverified低延迟与高质量推理往往难以兼得，前后台分离架构通过任务分级路由解决这一矛盾59% similarUnverifiedThe application of the Mixture of Experts (MoE) architecture has broken the traditional trade-off between parameter scale and inference speed.58% similarUnverifiedo1在某个任务中最终分数低于Claude 3.5，体现了强推理模型的'过度推理'问题——初始阶段出现偏差后会沿错误路径深入56% similarUnverifiedAccumulated floating-point errors may cause rewound results to be inconsistent with the original state, thereby breaking puzzle logic.56% similarUnverifiedIQ2_XXS档位在数学计算、复杂多步推理和高事实准确性问答任务上会出现明显退化，更适合快速原型验证、文本摘要、简单对话等场景55% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/996995API
curl https://kongchang.com/api/v1/knowledge/claims/996995MCP
get_claim(id=996995)