CCTest · Blog
Following the Bottleneck: How vLLM Optimized MiniMax M3 on AMD MI355X
Inference & Serving
cctest.ai

Following the Bottleneck: How vLLM Optimized MiniMax M3 on AMD MI355X

vLLM’s follow-up on MiniMax M3 shows how serving gains on AMD Instinct MI355X came from repeatedly locating the active bottleneck rather than relying on one breakthrough kernel. The work spans local shapes, redundant computation, metadata movement, backend dispatch, and queue capacity.

Read more
CCTest · Blog
Why LLMs Answer Impossible Questions: Recognition Is Not Refusal
Large Language Models
cctest.ai

Why LLMs Answer Impossible Questions: Recognition Is Not Refusal

A study of instruction-tuned models from 1.7B to 70B parameters suggests that many models recognize structurally unanswerable math and code prompts before generation. Their failure is better explained by a routing gap between recognition and abstention than by an inability to detect the problem.

Read more