Unverified50% confidenceBenchmarkExact time
On the recall-focused portion of the StuLife benchmark, JAZ's invoke outperforms Letta (MemGPT) by 8% at half the cost.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/24/2026
First Seen
Valid until: 12/23/2026
Sources
Related Entities
Related Claims
UnverifiedGPT models perform poorly when recall ability is tested in isolation, correctly recalling only about 50% of facts across all categories including spatial information, transcription content, and physical experiments, with open-source models performing even worse.60% similarUnverified选择性重计算由Korthikanti等人在2022年的Megatron-LM论文中系统研究，仅对注意力激活值做重计算可在额外计算开销不超过10%的条件下接近全量重计算的显存节省59% similarUnverifiedIn traditional LLM inference, the actual utilization rate of the KV cache is often less than 50%, while PagedAttention can raise it to nearly 100%.59% similarUnverifiedCognition has released coding model Sway 2, achieving 50.0% on the Frontier Code benchmark, claiming to be within 1 point of stronger models while costing 64% less.59% similarUnverifiedEvaluations such as LLM-as-a-Judge incur double or even multiple times the token overhead, because tokens are consumed both to generate the answer being evaluated and to have another model score it.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/946059API
curl https://kongchang.com/api/v1/knowledge/claims/946059MCP
get_claim(id=946059)