OptionalcurveOptionalfrequencyOptionalfrequencyOpenAI-style frequency penalty: subtracted once PER OCCURRENCE of the token in the recent window. May be negative.
OptionalgrammarOptionalhlgHLG tone-curve sampling. When enabled, replaces temperature's flat slope
with the piecewise curve (temperature becomes the mid gain). Off/undefined
⇒ the plain temperature path, unchanged. 02d723a:docs/archive/hlg-sampling.md.
OptionallogitAdditive per-token-id logit bias, applied before any penalty
(mlx-lm's logit_bias: {tokenId: bias}).
Optionalminmin-p: keep tokens whose probability ≥ minP · p(top token). mlx-lm's
min_p. Applied after top-p, before XTC/top-k (reference chain order).
OptionalminMinimum number of tokens min-p may never filter out. Default 1.
OptionalpresenceOptionalpresenceOpenAI-style presence penalty: subtracted ONCE from a token's logit if it occurred at all in the recent window. May be negative.
OptionalrepetitionOptionalrepetitionOptionalseedOptionaltemperatureOptionaltopOptionaltopOptionalxtcXTC ("exclude top choices"): with this probability per step, drop every token above the threshold EXCEPT the least likely of them.
OptionalxtcToken ids XTC never removes (mlx-lm's server passes EOS + "\n").
OptionalxtcProbability a token must exceed to be an XTC removal candidate. [0, 0.5].
v2 log-prob transfer-curve sampler. When set, the drawn monotone curve REPLACES temperature+softmax entirely (stochastic, seeded) — see packages/inference/src/sampling/curve.ts. Identity curve ≡ temperature 1.