Agent Evaluation
AI Agent 'Pass' ≠ Success — Pretty Numbers Are Self-Deception
A Juejin engineer: Agent evaluation hides infrastructure failures behind pretty success metrics. Buyers should ask: what are we really measuring?
Back to home
Model Evaluation
3 articles tagged with this topic
OpenRouter
Ox Alph Surfaces on OpenRouter Unclaimed — Chinese AI Labs Go 'Naked Release'
'Naked release' = shipping a model with no lab attribution or technical report. Ox Alph on OpenRouter is the latest case — Reddit users suspect Chines
Qwen
Qwen 3.8 Max's Most Underrated Skill: Knowing What NOT to Build
Qwen 3.8 Max wins overseas praise as "best at knowing what not to do." A Chinese bet on planning and restraint over parameter wars.