CCTest · Blog
EdgeKernelBench Tests Whether LLMs Can Build Fast Edge Kernels
Evaluation & Benchmarks
cctest.ai

EdgeKernelBench Tests Whether LLMs Can Build Fast Edge Kernels

EdgeKernelBench evaluates LLM-generated kernels across heterogeneous edge backends through a unified intermediate representation and hierarchical workloads. Its results expose a persistent gap between functional correctness and real performance, while LEKA improves optimization through layered reasoning.

Read more
CCTest · Blog
Enterprise Agents Must Do More Than Retrieve—They Must Judge Which Records to Trust
Evaluation & Benchmarks
cctest.ai

Enterprise Agents Must Do More Than Retrieve—They Must Judge Which Records to Trust

MMEnterprise-Bench tests multimodal agents against the messy reality of workplace records that are revised, duplicated, and sometimes contradictory. Its results show that retrieving more evidence does not necessarily produce more reliable answers.

Read more