CCTest · Blog
AgentVQA Tests Whether Vision Models Can Truly Understand Agentic Environments
Evaluation & Benchmarks
cctest.ai

AgentVQA Tests Whether Vision Models Can Truly Understand Agentic Environments

AgentVQA brings together 1,000 realistic examples from web agents, robotics, egocentric video, games, and spatial planning. Its evaluation of 22 vision-language models shows that visual grounding and state tracking remain bigger obstacles than high-level reasoning.

Read more
CCTest · Blog
How vLLM Serves GLM-5.3 with Million-Token Contexts on 16 H200s
Inference & Serving
cctest.ai

How vLLM Serves GLM-5.3 with Million-Token Contexts on 16 H200s

vLLM combines Prefill Context Parallelism, Decode Context Parallelism, and Hybrid HiSparse to serve GLM-5.3 across two 8× H200 nodes. On the AgentX benchmark, the setup delivered 1.7–3.2× the per-user speed of data-parallel serving at the same throughput and was the only tested configuration to handle 1M-token contexts.

Read more
CCTest · Blog
SpatialScout Tests Whether VLM Agents Can Explore and Map 3D Scenes
Evaluation & Benchmarks
cctest.ai

SpatialScout Tests Whether VLM Agents Can Explore and Map 3D Scenes

SpatialScout introduces a benchmark in which vision-language agents control a camera, gather visual evidence, and localize objects in a shared 3D coordinate system. Its results show that frontier models can recover rough layouts, but still struggle to produce complete and precise metric maps.

Read more