Function llama_sampler_init_adaptive_p 

pub unsafe extern "C" fn llama_sampler_init_adaptive_p(
    target: f32,
    decay: f32,
    seed: u32,
) -> *mut llama_sampler

adaptive-p: select tokens near a configurable target probability over time.

the adaptive-p sampler transforms the token probability distribution to favor tokens that fall near a user-configurable probability target.

internally, the sampler maintains an exponential moving average of the ORIGINAL probabilities of selected tokens at each sampling step. it uses this EMA to compute an adapted target probability at each sampling step, thus maintaining the desired target probability over time.

adaptive-p selects a token ID rather than just mutating candidates, so it must be last in the sampler chain (like mirostat, dist, greedy).

only mild truncation before this sampler is recommended. we suggest applying min-p before adaptive-p as the only other active sampler in the chain.

@param target select tokens near this probability (valid range 0.0 to 1.0; negative = disabled) @param decay EMA decay for adaptation; history ≈ 1/(1-decay) tokens (valid range 0.0 - 0.99) @param seed RNG seed

ref: https://github.com/ggml-org/llama.cpp/pull/17927