free

Definition

As a suffix it means "requires no —": training-free, gradient-free, tuning-free, reference-free. The claim is that a step normally taken for granted has been eliminated, which is usually the paper's main selling point and the first thing to check, since the requirement often reappears elsewhere as a hand-tuned threshold or a stronger base model. It does not mean free of charge, nor free as in unconstrained.

Explainers using this term

• Paper Explained: FreeFlow — Optical Flow with a Hierarchical Transformer and No Flow-Specific PartsFreeFlow: A Bias-free Hierarchical Transformer for Optical Flow Estimation
• KV Cache Compression — Quantize, Evict, Share Your Way to Cheaper Long-Context InferenceKIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cache
• CFG and Samplers — What the "Strength" Knob in Generative AI Really DoesClassifier-Free Diffusion Guidance