| --- |
| pretty_name: LLM Watermark Evaluation — Inputs and Generations |
| license: [cc-by-4.0, cc-by-sa-4.0] |
| language: [en, ar, zh, hi, ja] |
| task_categories: [text-generation] |
| tags: [watermarking, llm-watermarking, ai-generated-text, text-detection, robustness] |
| size_categories: [100K<n<1M] |
| configs: |
| - config_name: main |
| default: true |
| data_files: |
| - split: thicket |
| path: data/thicket/main.parquet |
| - split: textseal |
| path: data/textseal/main.parquet |
| - split: gumbelmax |
| path: data/gumbelmax/main.parquet |
| - split: synthid |
| path: data/synthid/main.parquet |
| - split: kgw |
| path: data/kgw/main.parquet |
| - split: unigram |
| path: data/unigram/main.parquet |
| - split: none |
| path: data/none/main.parquet |
| - split: sweet |
| path: data/sweet/main.parquet |
| - split: ewd |
| path: data/ewd/main.parquet |
| - split: kgw_delta1.0 |
| path: data/kgw_delta1.0/main.parquet |
| - split: kgw_delta0.7 |
| path: data/kgw_delta0.7/main.parquet |
| - split: ewd_delta0.7 |
| path: data/ewd_delta0.7/main.parquet |
| - split: unbiased |
| path: data/unbiased/main.parquet |
| - split: dipmark |
| path: data/dipmark/main.parquet |
| - split: thicket__llama31 |
| path: data/thicket__llama31/main.parquet |
| - split: textseal__llama31 |
| path: data/textseal__llama31/main.parquet |
| - split: none__llama31 |
| path: data/none__llama31/main.parquet |
| - split: kgw_delta0.7__llama31 |
| path: data/kgw_delta0.7__llama31/main.parquet |
| - split: kgw_delta0.9__llama31 |
| path: data/kgw_delta0.9__llama31/main.parquet |
| - split: thicket__gemma2 |
| path: data/thicket__gemma2/main.parquet |
| - split: textseal__gemma2 |
| path: data/textseal__gemma2/main.parquet |
| - split: none__gemma2 |
| path: data/none__gemma2/main.parquet |
| - split: kgw_delta0.7__gemma2 |
| path: data/kgw_delta0.7__gemma2/main.parquet |
| - split: thicket__mistral7b |
| path: data/thicket__mistral7b/main.parquet |
| - split: textseal__mistral7b |
| path: data/textseal__mistral7b/main.parquet |
| - split: none__mistral7b |
| path: data/none__mistral7b/main.parquet |
| - split: kgw_delta1.0__mistral7b |
| path: data/kgw_delta1.0__mistral7b/main.parquet |
| - split: kgw_delta0.7__mistral7b |
| path: data/kgw_delta0.7__mistral7b/main.parquet |
| - config_name: dilution_answers |
| data_files: |
| - split: thicket |
| path: data/thicket/dilution_answers.parquet |
| - split: textseal |
| path: data/textseal/dilution_answers.parquet |
| - split: gumbelmax |
| path: data/gumbelmax/dilution_answers.parquet |
| - split: synthid |
| path: data/synthid/dilution_answers.parquet |
| - split: kgw |
| path: data/kgw/dilution_answers.parquet |
| - split: unigram |
| path: data/unigram/dilution_answers.parquet |
| - split: none |
| path: data/none/dilution_answers.parquet |
| - split: sweet |
| path: data/sweet/dilution_answers.parquet |
| - split: ewd |
| path: data/ewd/dilution_answers.parquet |
| - split: kgw_delta1.0 |
| path: data/kgw_delta1.0/dilution_answers.parquet |
| - split: kgw_delta0.7 |
| path: data/kgw_delta0.7/dilution_answers.parquet |
| - split: ewd_delta0.7 |
| path: data/ewd_delta0.7/dilution_answers.parquet |
| - split: unbiased |
| path: data/unbiased/dilution_answers.parquet |
| - split: dipmark |
| path: data/dipmark/dilution_answers.parquet |
| - split: thicket__llama31 |
| path: data/thicket__llama31/dilution_answers.parquet |
| - split: textseal__llama31 |
| path: data/textseal__llama31/dilution_answers.parquet |
| - split: none__llama31 |
| path: data/none__llama31/dilution_answers.parquet |
| - split: kgw_delta0.7__llama31 |
| path: data/kgw_delta0.7__llama31/dilution_answers.parquet |
| - split: kgw_delta0.9__llama31 |
| path: data/kgw_delta0.9__llama31/dilution_answers.parquet |
| - split: thicket__gemma2 |
| path: data/thicket__gemma2/dilution_answers.parquet |
| - split: textseal__gemma2 |
| path: data/textseal__gemma2/dilution_answers.parquet |
| - split: none__gemma2 |
| path: data/none__gemma2/dilution_answers.parquet |
| - split: kgw_delta0.7__gemma2 |
| path: data/kgw_delta0.7__gemma2/dilution_answers.parquet |
| - split: thicket__mistral7b |
| path: data/thicket__mistral7b/dilution_answers.parquet |
| - split: textseal__mistral7b |
| path: data/textseal__mistral7b/dilution_answers.parquet |
| - split: none__mistral7b |
| path: data/none__mistral7b/dilution_answers.parquet |
| - split: kgw_delta1.0__mistral7b |
| path: data/kgw_delta1.0__mistral7b/dilution_answers.parquet |
| - split: kgw_delta0.7__mistral7b |
| path: data/kgw_delta0.7__mistral7b/dilution_answers.parquet |
| - config_name: multilingual |
| data_files: |
| - split: thicket |
| path: data/thicket/multilingual.parquet |
| - split: textseal |
| path: data/textseal/multilingual.parquet |
| - split: gumbelmax |
| path: data/gumbelmax/multilingual.parquet |
| - split: synthid |
| path: data/synthid/multilingual.parquet |
| - split: kgw |
| path: data/kgw/multilingual.parquet |
| - split: unigram |
| path: data/unigram/multilingual.parquet |
| - split: none |
| path: data/none/multilingual.parquet |
| - split: sweet |
| path: data/sweet/multilingual.parquet |
| - split: ewd |
| path: data/ewd/multilingual.parquet |
| - split: kgw_delta1.0 |
| path: data/kgw_delta1.0/multilingual.parquet |
| - split: kgw_delta0.7 |
| path: data/kgw_delta0.7/multilingual.parquet |
| - split: ewd_delta0.7 |
| path: data/ewd_delta0.7/multilingual.parquet |
| - split: unbiased |
| path: data/unbiased/multilingual.parquet |
| - split: dipmark |
| path: data/dipmark/multilingual.parquet |
| - split: thicket__llama31 |
| path: data/thicket__llama31/multilingual.parquet |
| - split: textseal__llama31 |
| path: data/textseal__llama31/multilingual.parquet |
| - split: none__llama31 |
| path: data/none__llama31/multilingual.parquet |
| - split: kgw_delta0.7__llama31 |
| path: data/kgw_delta0.7__llama31/multilingual.parquet |
| - split: kgw_delta0.9__llama31 |
| path: data/kgw_delta0.9__llama31/multilingual.parquet |
| - split: thicket__gemma2 |
| path: data/thicket__gemma2/multilingual.parquet |
| - split: textseal__gemma2 |
| path: data/textseal__gemma2/multilingual.parquet |
| - split: none__gemma2 |
| path: data/none__gemma2/multilingual.parquet |
| - split: kgw_delta0.7__gemma2 |
| path: data/kgw_delta0.7__gemma2/multilingual.parquet |
| - split: thicket__mistral7b |
| path: data/thicket__mistral7b/multilingual.parquet |
| - split: textseal__mistral7b |
| path: data/textseal__mistral7b/multilingual.parquet |
| - split: none__mistral7b |
| path: data/none__mistral7b/multilingual.parquet |
| - split: kgw_delta1.0__mistral7b |
| path: data/kgw_delta1.0__mistral7b/multilingual.parquet |
| - split: kgw_delta0.7__mistral7b |
| path: data/kgw_delta0.7__mistral7b/multilingual.parquet |
| - config_name: benchmarks |
| data_files: |
| - split: thicket |
| path: data/thicket/benchmarks.parquet |
| - split: textseal |
| path: data/textseal/benchmarks.parquet |
| - split: gumbelmax |
| path: data/gumbelmax/benchmarks.parquet |
| - split: synthid |
| path: data/synthid/benchmarks.parquet |
| - split: kgw |
| path: data/kgw/benchmarks.parquet |
| - split: unigram |
| path: data/unigram/benchmarks.parquet |
| - split: none |
| path: data/none/benchmarks.parquet |
| - split: sweet |
| path: data/sweet/benchmarks.parquet |
| - split: ewd |
| path: data/ewd/benchmarks.parquet |
| - split: kgw_delta1.0 |
| path: data/kgw_delta1.0/benchmarks.parquet |
| - split: kgw_delta0.7 |
| path: data/kgw_delta0.7/benchmarks.parquet |
| - split: ewd_delta0.7 |
| path: data/ewd_delta0.7/benchmarks.parquet |
| - split: unbiased |
| path: data/unbiased/benchmarks.parquet |
| - split: dipmark |
| path: data/dipmark/benchmarks.parquet |
| - split: thicket__llama31 |
| path: data/thicket__llama31/benchmarks.parquet |
| - split: textseal__llama31 |
| path: data/textseal__llama31/benchmarks.parquet |
| - split: none__llama31 |
| path: data/none__llama31/benchmarks.parquet |
| - split: kgw_delta0.7__llama31 |
| path: data/kgw_delta0.7__llama31/benchmarks.parquet |
| - split: kgw_delta0.9__llama31 |
| path: data/kgw_delta0.9__llama31/benchmarks.parquet |
| - split: thicket__gemma2 |
| path: data/thicket__gemma2/benchmarks.parquet |
| - split: textseal__gemma2 |
| path: data/textseal__gemma2/benchmarks.parquet |
| - split: none__gemma2 |
| path: data/none__gemma2/benchmarks.parquet |
| - split: kgw_delta0.7__gemma2 |
| path: data/kgw_delta0.7__gemma2/benchmarks.parquet |
| - split: thicket__mistral7b |
| path: data/thicket__mistral7b/benchmarks.parquet |
| - split: textseal__mistral7b |
| path: data/textseal__mistral7b/benchmarks.parquet |
| - split: none__mistral7b |
| path: data/none__mistral7b/benchmarks.parquet |
| - split: kgw_delta1.0__mistral7b |
| path: data/kgw_delta1.0__mistral7b/benchmarks.parquet |
| - split: kgw_delta0.7__mistral7b |
| path: data/kgw_delta0.7__mistral7b/benchmarks.parquet |
| - config_name: attacks |
| data_files: |
| - split: thicket |
| path: data/thicket/attacks.parquet |
| - split: textseal |
| path: data/textseal/attacks.parquet |
| - split: gumbelmax |
| path: data/gumbelmax/attacks.parquet |
| - split: synthid |
| path: data/synthid/attacks.parquet |
| - split: kgw |
| path: data/kgw/attacks.parquet |
| - split: unigram |
| path: data/unigram/attacks.parquet |
| - split: sweet |
| path: data/sweet/attacks.parquet |
| - split: ewd |
| path: data/ewd/attacks.parquet |
| - split: kgw_delta1.0 |
| path: data/kgw_delta1.0/attacks.parquet |
| - split: kgw_delta0.7 |
| path: data/kgw_delta0.7/attacks.parquet |
| - split: ewd_delta0.7 |
| path: data/ewd_delta0.7/attacks.parquet |
| - split: unbiased |
| path: data/unbiased/attacks.parquet |
| - split: dipmark |
| path: data/dipmark/attacks.parquet |
| - split: thicket__llama31 |
| path: data/thicket__llama31/attacks.parquet |
| - split: textseal__llama31 |
| path: data/textseal__llama31/attacks.parquet |
| - split: none__llama31 |
| path: data/none__llama31/attacks.parquet |
| - split: kgw_delta0.7__llama31 |
| path: data/kgw_delta0.7__llama31/attacks.parquet |
| - split: kgw_delta0.9__llama31 |
| path: data/kgw_delta0.9__llama31/attacks.parquet |
| - split: thicket__gemma2 |
| path: data/thicket__gemma2/attacks.parquet |
| - split: textseal__gemma2 |
| path: data/textseal__gemma2/attacks.parquet |
| - split: none__gemma2 |
| path: data/none__gemma2/attacks.parquet |
| - split: kgw_delta0.7__gemma2 |
| path: data/kgw_delta0.7__gemma2/attacks.parquet |
| - split: thicket__mistral7b |
| path: data/thicket__mistral7b/attacks.parquet |
| - split: textseal__mistral7b |
| path: data/textseal__mistral7b/attacks.parquet |
| - split: none__mistral7b |
| path: data/none__mistral7b/attacks.parquet |
| - split: kgw_delta1.0__mistral7b |
| path: data/kgw_delta1.0__mistral7b/attacks.parquet |
| - split: kgw_delta0.7__mistral7b |
| path: data/kgw_delta0.7__mistral7b/attacks.parquet |
| - config_name: documents |
| data_files: |
| - split: thicket |
| path: data/thicket/documents.parquet |
| - split: textseal |
| path: data/textseal/documents.parquet |
| - split: gumbelmax |
| path: data/gumbelmax/documents.parquet |
| - split: synthid |
| path: data/synthid/documents.parquet |
| - split: kgw |
| path: data/kgw/documents.parquet |
| - split: unigram |
| path: data/unigram/documents.parquet |
| - split: none |
| path: data/none/documents.parquet |
| - split: sweet |
| path: data/sweet/documents.parquet |
| - split: ewd |
| path: data/ewd/documents.parquet |
| - split: kgw_delta1.0 |
| path: data/kgw_delta1.0/documents.parquet |
| - split: kgw_delta0.7 |
| path: data/kgw_delta0.7/documents.parquet |
| - split: ewd_delta0.7 |
| path: data/ewd_delta0.7/documents.parquet |
| - split: unbiased |
| path: data/unbiased/documents.parquet |
| - split: dipmark |
| path: data/dipmark/documents.parquet |
| - split: thicket__llama31 |
| path: data/thicket__llama31/documents.parquet |
| - split: textseal__llama31 |
| path: data/textseal__llama31/documents.parquet |
| - split: none__llama31 |
| path: data/none__llama31/documents.parquet |
| - split: kgw_delta0.7__llama31 |
| path: data/kgw_delta0.7__llama31/documents.parquet |
| - split: kgw_delta0.9__llama31 |
| path: data/kgw_delta0.9__llama31/documents.parquet |
| - split: thicket__gemma2 |
| path: data/thicket__gemma2/documents.parquet |
| - split: textseal__gemma2 |
| path: data/textseal__gemma2/documents.parquet |
| - split: none__gemma2 |
| path: data/none__gemma2/documents.parquet |
| - split: kgw_delta0.7__gemma2 |
| path: data/kgw_delta0.7__gemma2/documents.parquet |
| - split: thicket__mistral7b |
| path: data/thicket__mistral7b/documents.parquet |
| - split: textseal__mistral7b |
| path: data/textseal__mistral7b/documents.parquet |
| - split: none__mistral7b |
| path: data/none__mistral7b/documents.parquet |
| - split: kgw_delta0.7__mistral7b |
| path: data/kgw_delta0.7__mistral7b/documents.parquet |
| --- |
| |
| # LLM Watermark Evaluation — 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: |
|
|
| ```python |
| 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 | 1,600 | 1,400 | |
| | `textseal` | TextSeal (official code), mixing α = 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): γ = 0.25, δ = 2.0, context width 1 | 5,000 | 200 | 500 | 300 | 1,600 | 1,400 | |
| | `unigram` | Unigram (Zhao et al. 2024): fixed green list, γ = 0.25, δ = 2.0 | 5,000 | 200 | 500 | 300 | 1,600 | 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 | 1,600 | 1,400 | |
| | `ewd` | EWD (Lu et al. 2024): KGW generation (γ 0.25, δ 2.0), entropy-weighted detection | 5,000 | 200 | 500 | 300 | 1,600 | 1,400 | |
| | `kgw_delta1.0` | KGW, γ = 0.25, δ = 1.0 | 5,000 | 200 | 500 | 300 | 1,600 | 1,400 | |
| | `kgw_delta0.7` | KGW, γ = 0.25, δ = 0.7 (perplexity cost matched to Thicket's) | 5,000 | 200 | 500 | 300 | 1,600 | 1,400 | |
| | `ewd_delta0.7` | EWD at δ = 0.7: KGW generation (γ 0.25, δ 0.7), entropy-weighted detection (matched distortion) | 5,000 | 200 | 500 | 300 | 1,600 | 1,400 | |
| | `unbiased` | γ-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 | |
|
|
| ## 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; 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 | 1,600 | 1,400 | |
| | `textseal__llama31` | `meta-llama/Llama-3.1-8B-Instruct` | TextSeal (official code), mixing α = 0.1 | 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, γ = 0.25, δ = 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, γ = 0.25, δ = 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; 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 | 1,600 | 1,400 | |
| | `textseal__gemma2` | `google/gemma-2-9b-it` | TextSeal (official code), mixing α = 0.1 | 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, γ = 0.25, δ = 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; 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 | 1,600 | 1,400 | |
| | `textseal__mistral7b` | `mistralai/Mistral-7B-Instruct-v0.3` | TextSeal (official code), mixing α = 0.1 | 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, γ = 0.25, δ = 1.0 | 5,000 | 200 | 500 | 300 | 1,600 | – | |
| | `kgw_delta0.7__mistral7b` | `mistralai/Mistral-7B-Instruct-v0.3` | KGW, γ = 0.25, δ = 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`](keys.json) and explained in [`KEYS.md`](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, 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 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. 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). |
| - 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 © 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 © 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. |
| |
