CFG

The implementation of negative prompting techniques in text-to-image scenarios relies on the Classifier-Free Guidance (CFG) mechanism.

A higher CFG Scale causes the generated result to adhere more strictly to the text description, but too high a value may lead to oversaturated colors and distorted details.

CFG值（Classifier-Free Guidance Scale）通常设置在5-12之间，采样步数一般在20-50步之间取得较好平衡

FSM constraints can guarantee JSON format validity at the grammatical level, but cannot constrain the semantic quality of fields.

When CFG is 1, negative prompts have no effect; if negative prompts are needed, the CFG value can be slightly increased (e.g., to 1.5).

Even with the same prompt, different seeds produce completely different images; once the seed is fixed, even small changes to the number of sampling steps or the CFG guidance strength can cause differences in the output.

The CFG value (Classifier-Free Guidance Scale) controls the degree to which the model adheres to the prompt; a higher value yields greater fidelity but may result in oversaturation.

The Classifier-Free Guidance (CFG) sampling technique, while improving image quality, pushes outputs toward high-density regions of the data distribution, further exacerbating output homogenization.

New-generation models such as FLUX and Z-Image, due to architectural improvements, achieve optimal results with a CFG value of 1 or even lower, whereas traditional Stable Diffusion models typically use CFG values of 7–12.

Logic bugs do not rely on guessing the memory layout or hijacking control flow, so memory mitigation mechanisms such as ASLR and CFG are essentially ineffective against them.

Modern browsers generally deploy multiple layers of protection such as ASLR, CFG, and sandboxes, raising the bar for exploiting memory corruption vulnerabilities.

CFG (Control Flow Guard) verifies whether the target address belongs to a legitimate function entry before an indirect call, blocking a common path by which attackers hijack control flow.

CFG is a parameter controlling the strength of alignment between the generated result and the prompt, usually taking values of 5-9; acceleration LoRAs require CFG=1 because the model has already internalized the guidance signal during distillation, and continuing to use a high CFG would break the step-compression mapping.

The 1920x1088 resolution failed due to OOM because the CFG path requires two forward passes per step, and the high-definition tier doubles memory before improving anything.

The SFG branch only needs to run during the first 20% of image sampling steps, and without using CFG its overhead is only approximately 1.2× the base NFE.

Classifier-free guidance (CFG) runs both a conditional branch and an unconditional branch simultaneously, amplifying the conditional signal by subtracting the unconditional direction from the conditional direction, trading a doubling of inference cost for quality improvement.

Turbo defaults to CFG=1, requiring only one forward inference pass per step.

CFG (Classifier-Free Guidance) requires running both conditional and unconditional forward passes simultaneously at each denoising step, doubling the computation per inference step when enabled.

On Qwen-Image 2.1, a lower CFG value (3) generally produces more natural output with less of an overfitting feel.

Supra2-IMG uses CFG 3.0 as the default display value, which is lower than the 7.0–9.0 commonly used by Stable Diffusion.