Back to evaluations
Public evaluation

KV Cache Allocation Test

Evaluates OOM preemption rates and latency profiles of the KV-cache planner under continuous load.

Evaluation type
task based
Challenge
KV-Cache Capacity Planner Agent Mesh with CrewAI and Cursor
Difficulty
Intermediate
Rigor
Unspecified

Evaluation overview

How the linked challenge is judged: tasks, benchmarks, and criteria count.

Tasks
1
Benchmarks
0
Criteria
0

Task templates

Inputs and expected outputs.

Task 1

KV Cache Allocation Test

Simulates 1000 long-context requests against a fixed 80GB GPU memory boundary.

Input format

JSON payload containing request arrival rates and context lengths

Output format

JSON containing preemption_count and p95_ttft_ms