Unverified50% confidenceBenchmarkExact time
When the neuron retention rate drops to 50%, RAMPr can reduce resident GPU memory by 36-43%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/11/2026
First Seen
Valid until: 2/9/2027
Sources
Related Entities
Related Claims
UnverifiedAt a 75% neuron retention rate, RAMPr retains on average 91-98% of the original full performance across six models.70% similarUnverifiedFreeToken supports runtime adjustment of the GPU expert cache size; tests show that reducing the cache from 58% to 40% results in only about 10% performance loss, with performance dropping sharply below 20%.70% similarUnverifiedAfter adopting the subscription pass-through mechanism at the Router layer, ablation experiments showed that the Router's number of decisions decreased by about 30% and cached tokens decreased by 37%.67% similarVerifiedDuring the autoregressive decoding phase, GPU compute utilization is often less than 10%, making it a memory-intensive operation.66% similarUnverifiedScaling from 1 to 56 sessions means a reduction of approximately 98% in GPU occupancy per call.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1158063API
curl https://kongchang.com/api/v1/knowledge/claims/1158063MCP
get_claim(id=1158063)