Server configuration
Store the configuration in a file readable only by the server account. The keys below are public test values, not production secrets:default_enabled and enforce_all work as for Gumbel; see Server configuration. Fields of the other algorithm are rejected.
Request control
Requests turn watermarking on and off with thewatermark object described in Per-request control; with synthid only its enabled field is accepted. Greedy requests (top_k 1 or temperature 0) are never watermarked.
Sampling
A watermarked row is sampled as follows:- Penalties, logit bias, and the grammar mask apply as usual.
- The row keeps its
num_candidateshighest logits, applies its temperature, then its request top-k, top-p, and min-p. - The tournament reweights the remaining candidates by their g-values for the row’s context.
- The stock sampler draws from the result.
SGLANG_RETURN_ORIGINAL_LOGPROB keeps its meaning. A returned sampling mask describes the watermarked support the token was drawn from.
Speculative decoding
SynthID supports EAGLE and EAGLE3 with--speculative-eagle-topk 1, --speculative-use-rejection-sampling, and at most 8 --speculative-num-draft-tokens:
Unsupported configurations
--watermark-algorithm synthid is rejected with:
SGLANG_ENABLE_NV_MEGA_RUNTIME_OPTIMIZATION;--rl-on-policy-target, since the watermark changes the sampling distribution.
--enable-watermark that hold for every algorithm, such as PD disaggregation and custom logit processors, are listed under Compatibility.
State and detection
A request’s context starts empty (zero tokens), not from the prompt, and holds its lastngram_len - 1 generated tokens. A request resumed after retraction replays its generated tokens, so its watermark continues as if uninterrupted. The repeated-context history is a bounded ring: a context can be watermarked again after it is evicted.
To score text, recompute each generated token’s g-values with the same keys, ngram_len, context_history_size, and skip_first_ngram_calls, starting from the same empty context. Calibrate detection thresholds for the output length and the false-positive rate required; a low score is not proof of authorship.