Unverified50% confidenceBenchmarkExact time
According to the STALE research, most dedicated memory frameworks score below 10% on the task of recognizing the failure of their own memory.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/14/2026
First Seen
Valid until: 1/12/2027
Sources
Related Entities
Related Claims
VerifiedDuring the autoregressive decoding phase, GPU compute utilization is often less than 10%, making it a memory-intensive operation.73% similarUnverifiedAccording to the ForgetEval research, when leading memory systems were scored on test cases written by people other than the authors themselves, internal test scores of 63%-68% plummeted to 28%-33%.72% similarUnverified少样本方式在格式化输出、分类任务和规则遵循方面的幻觉率可降低40%-70%71% similarUnverified11 of the 18 models correctly answered at least 96% of memory read questions, but the gap widened significantly on the memory write task (maintaining the passport).71% similarUnverifiedTest data shows that only about 10% of agent tasks genuinely require the highest-tier model, while the remaining 90% show no measurable quality loss when using cheaper tiers.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1030577API
curl https://kongchang.com/api/v1/knowledge/claims/1030577MCP
get_claim(id=1030577)