A vLLM setting took Kimi K2.6 tool calls from 92% valid to 100% (ubicloud.com)
Ubicloud measured 92% schema accuracy on Kimi K2.6 tool calls with vLLM 0.28, about 2 failures in 25. Turning on constrained decoding explicitly brought it to 100% on the K2VV ToolCall benchmark.
The thread's pushback: the fix was a configuration, not a patch.
Sign in to save