Dataset Viewer
Auto-converted to Parquet Duplicate
The dataset viewer is not available for this split.
Not found.
Error code:   ResponseNotFound

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

LLM Watermark Evaluation โ€” Inputs and Generations

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 โ€” 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; ฮบ = 6, price ฮณ = 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 1,600 1,400
textseal TextSeal (official code), mixing ฮฑ = 0.1 5,000 200 500 300 1,600 1,400
gumbelmax Gumbel-max / Aaronson (TextSeal-repo implementation) 5,000 200 500 300 1,600 1,400
synthid SynthID-Text, weighted tournament, depth 10 (TextSeal-repo implementation) 5,000 200 500 300 1,600 1,400
kgw KGW (Kirchenbauer et al. 2023): ฮณ = 0.25, ฮด = 2.0, context width 1 5,000 200 500 300 1,600 1,400
unigram Unigram (Zhao et al. 2024): fixed green list, ฮณ = 0.25, ฮด = 2.0 5,000 200 500 300 1,600 1,400
none No watermark (plain sampling; the unwatermarked reference) 5,000 200 500 300 โ€“ 1,400
sweet SWEET (Lee et al. 2024): entropy-gated KGW, ฮณ = 0.25, ฮด = 2.0, entropy threshold 1.2 5,000 200 500 300 1,600 1,400
ewd EWD (Lu et al. 2024): KGW generation (ฮณ 0.25, ฮด 2.0), entropy-weighted detection 5,000 200 500 300 1,600 1,400
kgw_delta1.0 KGW, ฮณ = 0.25, ฮด = 1.0 5,000 200 500 300 1,600 1,400
kgw_delta0.7 KGW, ฮณ = 0.25, ฮด = 0.7 (perplexity cost matched to Thicket's) 5,000 200 500 300 1,600 1,400
ewd_delta0.7 EWD at ฮด = 0.7: KGW generation (ฮณ 0.25, ฮด 0.7), entropy-weighted detection (matched distortion) 5,000 200 500 300 1,600 1,400
unbiased ฮณ-reweight / Unbiased watermark (Hu et al. 2024), MarkLLM defaults 5,000 200 500 300 1,600 1,400
dipmark DiPmark (Wu et al. 2024), MarkLLM defaults 5,000 200 500 300 1,600 1,400
sir SIR semantic-invariant watermark (Liu et al. 2024), MarkLLM defaults 5,000 200 500 300 1,600 1,400
exp_edit EXP-edit (Kuditipudi et al. 2024), MarkLLM defaults 5,000 200 500 300 1,600 1,400

C4-RealNews โ€” Qwen3.5-9B

News-article continuations (the KGW / Tr-GoF evaluation setting): configs c4_*, one split per scheme.

scheme (split) scheme and settings c4_realnews c4_dilution_answers c4_attacks c4_documents
thicket Thicket (ours): keyed token-pair circle tilt with a likelihood price; ฮบ = 6, price ฮณ = 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 1,600 1,400
textseal TextSeal (official code), mixing ฮฑ = 0.1 5,000 200 1,600 1,400
gumbelmax Gumbel-max / Aaronson (TextSeal-repo implementation) 5,000 200 1,600 1,400
synthid SynthID-Text, weighted tournament, depth 10 (TextSeal-repo implementation) 5,000 200 โ€“ โ€“
none No watermark (plain sampling; the unwatermarked reference) 5,000 200 1,600 1,400
kgw_delta0.7 KGW, ฮณ = 0.25, ฮด = 0.7 (perplexity cost matched to Thicket's) 5,000 200 1,600 1,400

Other generator models

The ELI5 protocol on other generators, one <scheme>__<model> split each in the ELI5 configs: the same questions, sampling settings and seeds; every prompt re-rendered from question with the generator's own chat template (Gemma has no system role, so the multilingual system prompt opens the user turn); generators without a thinking mode run multilingual and benchmarks without a reasoning budget (the whole generation is the answer, reasoning is empty).

scheme (split) generator scheme and settings main dilution_answers multilingual benchmarks attacks documents
thicket__llama31 meta-llama/Llama-3.1-8B-Instruct Thicket (ours): keyed token-pair circle tilt with a likelihood price; ฮบ = 6, price ฮณ = 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 1,600 1,400
textseal__llama31 meta-llama/Llama-3.1-8B-Instruct TextSeal (official code), mixing ฮฑ = 0.1 5,000 200 500 300 1,600 1,400
gumbelmax__llama31 meta-llama/Llama-3.1-8B-Instruct Gumbel-max / Aaronson (TextSeal-repo implementation) 5,000 200 500 300 1,600 1,400
synthid__llama31 meta-llama/Llama-3.1-8B-Instruct SynthID-Text, weighted tournament, depth 10 (TextSeal-repo implementation) 5,000 200 500 300 1,600 1,400
none__llama31 meta-llama/Llama-3.1-8B-Instruct No watermark (plain sampling; the unwatermarked reference) 5,000 200 500 300 1,600 1,400
kgw_delta0.7__llama31 meta-llama/Llama-3.1-8B-Instruct KGW, ฮณ = 0.25, ฮด = 0.7 (perplexity cost matched to Thicket's) 5,000 200 500 300 1,600 1,400
kgw_delta0.9__llama31 meta-llama/Llama-3.1-8B-Instruct KGW, ฮณ = 0.25, ฮด = 0.9 5,000 200 500 300 1,600 1,400
thicket__gemma2 google/gemma-2-9b-it Thicket (ours): keyed token-pair circle tilt with a likelihood price; ฮบ = 6, price ฮณ = 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 1,600 1,400
textseal__gemma2 google/gemma-2-9b-it TextSeal (official code), mixing ฮฑ = 0.1 5,000 200 500 300 1,600 1,400
gumbelmax__gemma2 google/gemma-2-9b-it Gumbel-max / Aaronson (TextSeal-repo implementation) 5,000 200 500 300 1,600 1,400
synthid__gemma2 google/gemma-2-9b-it SynthID-Text, weighted tournament, depth 10 (TextSeal-repo implementation) 5,000 200 500 300 1,600 1,400
none__gemma2 google/gemma-2-9b-it No watermark (plain sampling; the unwatermarked reference) 5,000 200 500 300 1,600 1,400
kgw_delta0.7__gemma2 google/gemma-2-9b-it KGW, ฮณ = 0.25, ฮด = 0.7 (perplexity cost matched to Thicket's) 5,000 200 500 300 1,600 1,400
thicket__mistral7b mistralai/Mistral-7B-Instruct-v0.3 Thicket (ours): keyed token-pair circle tilt with a likelihood price; ฮบ = 6, price ฮณ = 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 1,600 1,400
textseal__mistral7b mistralai/Mistral-7B-Instruct-v0.3 TextSeal (official code), mixing ฮฑ = 0.1 5,000 200 500 300 1,600 1,400
gumbelmax__mistral7b mistralai/Mistral-7B-Instruct-v0.3 Gumbel-max / Aaronson (TextSeal-repo implementation) 5,000 200 500 300 1,600 1,400
synthid__mistral7b mistralai/Mistral-7B-Instruct-v0.3 SynthID-Text, weighted tournament, depth 10 (TextSeal-repo implementation) 5,000 200 500 300 1,600 1,400
none__mistral7b mistralai/Mistral-7B-Instruct-v0.3 No watermark (plain sampling; the unwatermarked reference) 5,000 200 500 300 1,600 1,400
kgw_delta1.0__mistral7b mistralai/Mistral-7B-Instruct-v0.3 KGW, ฮณ = 0.25, ฮด = 1.0 5,000 200 500 300 1,600 1,400
kgw_delta0.7__mistral7b mistralai/Mistral-7B-Instruct-v0.3 KGW, ฮณ = 0.25, ฮด = 0.7 (perplexity cost matched to Thicket's) 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 0โ€“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 2kโ€“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, Englishโ†’Frenchโ†’English back-translation (NLLB-200-distilled-600M) rephrasing: temperature 0.5 / 0.8 / 1.0 / 1.2 (mild โ€ฆ 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
c4_realnews 1,000 news articles from C4 realnewslike (validation): the first 40 words of each 5 continuations per prefix (sample 0โ€“4), at most 400 tokens; human_continuation holds the article's own next 300 words plain text continuation (no chat template); temperature 0.8, top-p 0.9
c4_dilution_answers the first 200 of those prefixes 1 continuation each (embedded into Wikipedia documents in c4_documents) temperature 1.0, top-p 0.95
c4_attacks the first 200 c4_realnews continuations of sample 0 as attacks as attacks
c4_documents the 200 c4_dilution_answers texts and Wikipedia host documents as documents as documents

Fields

field meaning
method watermark scheme (the config name, without the __<model> suffix of the other-generator configs)
model the generator that produced the text (in attacks and documents: the generator of the source answer)
sample sample index 0โ€“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 (C4: c4rn-val-<line>, the article's line in the C4 validation shard)
question the question as written in the source dataset (C4: the 40-word prefix)
prompt the exact string given to the model: the question inside the generator's chat template; C4: the raw prefix
human_continuation C4 only: the article's next 300 words (human-written)
source_url, source_timestamp C4 only: the article's URL and crawl timestamp
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. The <scheme>__<model> configs use meta-llama/Llama-3.1-8B-Instruct, google/gemma-2-9b-it, mistralai/Mistral-7B-Instruct-v0.3 the same way (fp16), with the prompt handling described above.

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); the C4 articles' human continuations are included as human_continuation.
  • 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).
  • meta-llama/Llama-3.1-8B-Instruct: Built with Llama. Llama 3.1 is licensed under the Llama 3.1 Community License, Copyright ยฉ Meta Platforms, Inc. All Rights Reserved; outputs are also subject to the Llama 3.1 Acceptable Use Policy.
  • google/gemma-2-9b-it: Gemma Terms of Use (ai.google.dev/gemma/terms) and the Gemma Prohibited Use Policy.
  • mistralai/Mistral-7B-Instruct-v0.3: Apache 2.0.

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 ยฉ 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; ยฉ 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). C4 realnewslike (allenai/c4, ODC-BY 1.0; also subject to the Common Crawl terms of use): the prefixes and human_continuation texts.

Citation

Paper under review; citation to be added.

Downloads last month
1,254