RootAccess4Life/llm-watermark-eval-generations

LLM Watermark Evaluation — Inputs and Generations (Qwen3.5-9B)

Every input given to the generator and every text it produced during full-scale evaluations of the Thicket watermark against published watermarking schemes. All schemes were run with the same model, prompts, sampling settings and seeds, so each row of one config has an exact counterpart in every other config of the same generator (same prompt_id, same sample). The dataset lets readers inspect what went into the model and what came out, compare schemes on identical inputs (quality, diversity, length) and reuse the texts for robustness or detection experiments.

Each config is one stage (split type) and each of its splits is one scheme (the names in the tables below), so all splits of a config share one schema:

from datasets import load_dataset
ds = load_dataset("RootAccess4Life/llm-watermark-eval-generations", "main")                 # DatasetDict, one split per scheme
one = load_dataset("RootAccess4Life/llm-watermark-eval-generations", "main", split="thicket")

ELI5 questions — Qwen3.5-9B

One split per watermark scheme in each of the configs below; the counts are rows.

scheme (split)	scheme and settings	main	dilution_answers	multilingual	benchmarks	attacks	documents
thicket	Thicket (ours): keyed token-pair circle tilt with a likelihood price; κ = 6, price γ = 2 in answers (4 / 1 inside the reasoning delimiter), top-128 candidate shortlist, 256-label circle, first token and every 512-token restart neutral	5,000	200	500	300	–	1,400
textseal	TextSeal (official code), mixing α = 0.1	5,000	200	500	300	–	1,400
gumbelmax	Gumbel-max / Aaronson (TextSeal-repo implementation)	5,000	200	500	300	–	1,400
synthid	SynthID-Text, weighted tournament, depth 10 (TextSeal-repo implementation)	5,000	200	500	300	–	1,400
kgw	KGW (Kirchenbauer et al. 2023): γ = 0.25, δ = 2.0, context width 1	5,000	200	500	300	–	1,400
unigram	Unigram (Zhao et al. 2024): fixed green list, γ = 0.25, δ = 2.0	5,000	200	500	300	–	1,400
none	No watermark (plain sampling; the unwatermarked reference)	5,000	200	500	300	–	1,400
sweet	SWEET (Lee et al. 2024): entropy-gated KGW, γ = 0.25, δ = 2.0, entropy threshold 1.2	5,000	200	500	300	–	1,400
ewd	EWD (Lu et al. 2024): KGW generation (γ 0.25, δ 2.0), entropy-weighted detection	5,000	200	500	300	–	–
kgw_delta1.0	KGW, γ = 0.25, δ = 1.0	5,000	200	500	300	1,600	1,400

Each scheme used one fixed key; the keys are released in keys.json and explained in KEYS.md, so the detection results can be reproduced with the project's code.

Splits

split	inputs	outputs	sampling
main	1,000 English questions (ELI5, test split)	5 independent answers per question (sample 0–4), at most 400 tokens	temperature 0.8, top-p 0.9; thinking disabled
dilution_answers	the first 200 of those questions	1 answer each; the evaluation embeds these answers into 2k–12k-token Wikipedia documents to test detection inside longer text (see documents)	temperature 1.0, top-p 0.95
multilingual	100 questions in each of English, Arabic, Chinese, Hindi, Japanese	1 answer each, with its reasoning trace (reasoning) and final answer (answer); the watermark covers both	temperature 0.7, top-p 0.95; at most 2,000 reasoning tokens, 4,096 tokens in total
benchmarks	300 questions from GSM8K, MATH-500, AIME, MMLU, ARC-Easy, ARC-Challenge, HellaSwag, WinoGrande, HumanEval and MBPP (30 each), with their reference answers	1 answer each (reasoning trace and final answer), used to measure the effect of each watermark on task accuracy	temperature 0.6, top-p 0.95; at most 3,000 reasoning tokens, 4,096 tokens in total
attacks	the first 200 main answers of sample 0	each answer attacked 8 ways (attack, setting): rephrased by Llama-3.2-3B-Instruct with the TextSeal repository's four prompts (mild, moderate, aggressive, extreme), 30% word deletion, 50% WordNet synonym substitution, DIPPER-XXL paraphrase, English→French→English back-translation (NLLB-200-distilled-600M)	rephrasing: temperature 0.5 / 0.8 / 1.0 / 1.2 (mild … extreme), top-p 0.95, at most 1,024 new tokens; DIPPER: lexical diversity 60, order diversity 0, top-p 0.75; edits and sampling with fixed seeds
documents	the 200 dilution_answers texts and unwatermarked English Wikipedia host documents	1,400 documents: each answer embedded whole in a 2k, 4k, 8k or 12k-token document (dilute_*) or cut into k = 2, 3 or 5 pieces scattered over an 8k-token document (fragment_8k_k*); true_spans marks the watermarked text	insertion points drawn with a fixed seed; lengths in generator tokens

Fields

field	meaning
method	watermark scheme (the config name)
model	the generator that produced the text (in attacks and documents: the generator of the source answer)
sample	sample index 0–4 (the five samples use different sampling seeds); in attacks, the sample of the attacked answer
prompt_index	position of the question in the evaluation's list (in attacks: of the attacked answer, which main holds under the same prompt_id and sample; in documents: of the embedded answer, which dilution_answers holds under the same prompt_id)
prompt_id	identifier of the question in its source dataset
question	the question as written in the source dataset
prompt	the exact string given to the model: the question inside the generator's chat template
language	multilingual splits only
benchmark, kind, reference_answer	benchmarks only: source benchmark, answer type and gold answer
text	the generated text; in attacks the attacked text; in documents the whole document
reasoning, answer	multilingual splits and benchmarks: text split at its first </think> (the answer stripped of surrounding whitespace); empty reasoning when the generator has no thinking mode
attack, setting	attacks only: attack family (rephrasing, word_deletion, synonym_substitution, dipper, back_translation) and its strength or parameters
condition, length, k	documents only: dilution / fragmentation condition, target document length in generator tokens, number of watermarked pieces
host_id	documents only: the Wikipedia host document (consecutive English Wikipedia articles concatenated to the target length)
true_spans	documents only: character offsets (start, end, Python string indices) of the watermarked pieces in text
n_tokens	length of text in tokens of the generator's tokenizer

Generation

Model Qwen/Qwen3.5-9B, fp16, PyTorch SDPA attention; batched sampling with fixed per-batch seeds; each watermark applied inside the sampler by its official or reference implementation (TextSeal, Gumbel-max and SynthID-Text from the TextSeal codebase; KGW, Unigram, SWEET and EWD through the Hugging Face transformers watermark logits processor or re-implementations of their papers; Thicket from its authors' code). The unwatermarked reference (none, when present) uses the same sampler without a watermark.

Not included

• GPQA: its 30 questions and the corresponding answers are left out because GPQA's authors ask that its examples not be published online (benchmark-leakage prevention). All other benchmarks are included.
• The negatives of the false-positive tests (Wikipedia passages and unwatermarked same-model answers to other ELI5 questions).
• Copy-paste attack documents (watermarked fragments pasted into Wikipedia text).
• attacks and documents of the runs that finished before per-text records were kept (their attacked texts and documents were not stored; see the tables above).
• Evaluation scores: no p-values, detector outputs or metrics.

Sources and licenses

Generated texts: CC BY 4.0. They are outputs of the generator named in model, and their use is also subject to that model's terms:

• Qwen/Qwen3.5-9B: subject to the Qwen3.5 model license (see the model card).

Attacked texts (attacks, c4_attacks): the rephrasings are outputs of meta-llama/Llama-3.2-3B-Instruct (Built with Llama; Llama 3.2 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved), the paraphrases of kalpeshk2011/dipper-paraphraser-xxl, the back-translations of facebook/nllb-200-distilled-600M (model license CC BY-NC 4.0).

documents / c4_documents: CC BY-SA 4.0. These splits contain text from English Wikipedia (wikimedia/wikipedia, dump 20231101.en; © Wikipedia contributors, https://en.wikipedia.org), redistributed under CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/); the generated passages they embed are marked by true_spans. All other splits: CC BY 4.0.

Inputs keep the licenses of their sources: ELI5 (Reddit questions, as distributed in the Pavithree/eli5 Hugging Face dataset; also the English multilingual questions), the multilingual question pool (shanearora/CaLMQA, CC BY 4.0), GSM8K (MIT), MATH-500 (MIT), AIME (public competition problems), MMLU (MIT), ARC (CC BY-SA 4.0), HellaSwag (MIT), WinoGrande (CC BY), HumanEval (MIT), MBPP (CC BY 4.0).

Citation

Paper under review; citation to be added.