Unverified50% confidenceBenchmarkExact time
In the RULER-16K experiments, KVFetch improved the average score across 13 tasks by +8.4, with gains concentrated on tasks requiring sequential access.
1
Sources
50%
Confidence
Long-term
Relevance
10/8/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAfter using TailSFT, the pass@16 metric on coding tasks during the SFT phase showed an absolute improvement of up to 16.8 points.66% similarUnverifiedIn equal-budget controlled experiments on the RULER-16K dataset, KVFetch restored the verbatim copying score from 0.8 to 78.4.65% similarUnverifiedClaude achieved a task completion rate of 14.9% on the OSWorld benchmark, representing the highest level at the time.63% similarUnverifiedIn the InferenceX public benchmark, Jalapeño achieved 1.5 to 1.9 times the tasks per watt at peak throughput compared to NVIDIA's reference system, with end-to-end latency reduced by 1.7 to 3.6 times.60% similarUnverifiedThe full CCPS achieves or exceeds the Power-SMC baseline in final answer accuracy in 14 out of 15 settings.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/986759API
curl https://kongchang.com/api/v1/knowledge/claims/986759MCP
get_claim(id=986759)