Constructing Budgeted KGW Evasion through Token-Event Evidence Replacement
Organizations: Independent Researcher
Abstract
KGW watermarks bias sampling to accumulate keyed detection evidence. We construct budgeted evasion through token-event evidence replacement using ordinary phrase rewrites. Our construction jointly selects compositions of LLM-proposed alternatives from an independent model and measures global token-edit distance on complete outputs. At the largest attainable cost within the budget, exact coverage search maximizes removal of original scored token pairs. It uses a known tokenizer without the key, detector feedback or queries to the watermarked model. Exact accounting separates source evidence removed, evidence created and length effects, with initial strength setting the required reduction. Separate signed-event analysis found that mean suppression arose mainly from removing watermark-enriched source events. In controlled, occurrence-counted, one-token-context KGW tests on short passages, 39 of 82 qualified, initially detected sources yielded rewrites with terminal detector score Z < 4 that passed blinded source-relative model review, at a mean actual token-edit cost of 16.0% under a 20% allowance. To examine whether this initial requirement can be inferred, a separate historical 95-passage study estimated initial detector score Z0 from generation traces with correlation 0.793 and an in-sample association beyond length. This aggregate information did not establish improved allocation, and the separate no-trace stopping test did not reliably exploit cheaper reviewed crossings in the candidate bank. Comparisons did not establish an advantage over simpler selection and ordinary constrained rewriting. The construction makes budgeted evidence replacement executable while exposing limits in signed reduction and recognition of sufficient editing.
Figures & tables
| Source outcome | Count |
|---|---|
| Escaped under the original coverage selection | 40 |
| Additional escape available at the same actual cost | 2 |
| Additional escape available only at a lower actual cost | 14 |
| No escaping output in the full allowed candidate bank | 26 |
| Total initially detected qualified sources | 82 |
Appendix figures & tables13 assets
Supplementary material from the paper’s appendix.
Appendix
| Schematic path | Token sequence | Adjacent-token identities |
|---|---|---|
| Source | ||
| Replace | ||
| Replace | ||
| Replace both |
| Object | Blind editor | Evaluator |
|---|---|---|
| Source prose, pinned tokenizer, literal alternatives | Yes | Yes |
| Full-output | Yes | Yes |
| Public-model source surprisals, when enabled | Yes | Yes |
| True key, event states, source/output | No | Yes |
| Original generation trace | No | Diagnostic only |
| Global source-relative quality judgment | After selection | Yes |
| Study | Sources; clusters | Information and role |
|---|---|---|
| Confirmation | 105 of 128; 16/8 | 82 initially detected; final coverage uses tokenizer-only structure. |
| Starting score | 95; 12/8 | Historical CUDA generation observations; aggregate inference. |
| Source nomination | 64; 8/8 | Historical MPS generation observations; candidate-supply diagnostic in the companion. |
| Ordinary rewriting | 16; 10/8 | Exposed development pairs; matched actual cost and mixed model reviewers. |
| Stopping | Same 105; 16/8 | Exposed bank; other-key training scores, without generation traces. |
| Measurement | Cached graph | Build included |
|---|---|---|
| Median paired speedup | ||
| Sources faster than enumeration | 61/105 | 38/105 |
| Sum of source median times (s) | 1.348 | 1.868 |
| Aggregate speedup |
| Policy | Cap | Cost | Raw | Relative | Polished | 95% CI | |
|---|---|---|---|---|---|---|---|
| Coverage | 5% | 4.33% | 13 | 13 | 13 | 0.436 | [0.293, 0.581] |
| 10% | 9.31% | 25 | 24 | 22 | 0.930 | [0.716, 1.142] | |
| 15% | 13.41% | 39 | 37 | 32 | 1.255 | [1.012, 1.515] | |
| 20% | 15.96% | 40 | 39 | 34 | 1.373 | [1.089, 1.660] | |
| Weighted | 5% | 4.33% | 16 | 16 | 15 | 0.496 | [0.360, 0.634] |
| 10% | 9.31% | 29 | 27 | 24 | 0.977 | [0.791, 1.171] |
| Policy | Mean | Relative /105 | Raw /82 | Joint /82 | Polished /82 |
|---|---|---|---|---|---|
| Weighted | 1.333 | 101 | 40 | 39 | 34 |
| Coverage | 1.313 | 100 | 40 | 39 | 34 |
| Fixed hash | 1.305 | 102 | 38 | 37 | 32 |
| 3B weighting | 1.334 | 101 | 39 | 38 | 33 |
| Allowance | Weighted constant | Coverage uniform | Oracle coverage |
|---|---|---|---|
| 5% | |||
| 10% | |||
| 15% | |||
| 20% |
| Quantity | s029 | s001 |
|---|---|---|
| Source/output scored events | ||
| Source centered evidence | ||
| Destroyed-event count | ||
| Homogeneous source reference | ||
| Selected-source excess | ||
| Created contribution |
| Comparison | Extra | 95% interval | |
|---|---|---|---|
| Resolved, primary | 104 | 0.0314 | |
| Operational, with fallback | 105 | 0.0311 | |
| Initially detected | 82 | 0.0070 | |
| Same output length | 104 | 0.0117 | |
| Same | 104 | 0.0209 |
| Policy | Cost (%) | Qualified | Certificate | ||
|---|---|---|---|---|---|
| Coverage-p20 | 15.96 | 40 | 39 | 30 | 34 |
| S1 envelope, target 4 | 15.96 | 40 | 39 | 30 | 34 |
| Training-selected fixed dose | 13.41 | 39 | 37 | 23 | 26 |
| S1 mean, target 4 | 11.55 | 25 | 15–24 | 15 | 18 |
| S1 mean, target 3.5 | 14.35 | 40 | 28–39 | 24 | 27 |
| Operation | Calls | Output-token cap | Identifier / reasoning |
|---|---|---|---|
| Confirmation proposals, Oct. 18 | 1,018 | 1,024 | gpt-5.6-luna / low |
| Source review, Oct. 18 | 128 | 6,000 | gpt-6.1-sol / medium |
| Selected-output review, Oct. 18 | 105 | 6,000 | gpt-6.1-sol / medium |
| Hash completion, Oct. 19 | 39 | 6,000 | gpt-6.1-sol / medium |
| Minimum witnesses, Oct. 25 | 8 | 6,000 | gpt-6.1-sol / medium |
| Coverage completion, Oct. 27 | 5 | 6,000 | gpt-6.1-sol / medium |
| ID | Topic | ID | Topic |
|---|---|---|---|
| p000 | US healthcare costs | p008 | Biblical fathers |
| p001 | Himalayan religious community | p009 | Casting a musical |
| p002 | Church and public-policy debates | p010 | Surface properties and boiling |
| p003 | Parliamentary misconduct | p011 | Alleged attack on a football mascot |
| p004 | Social-media data security | p012 | Diplomatic criminal charges |
| p005 | Television drama adaptation | p013 | Winter television viewing |
LLM Watermarking
Large language model watermarking embeds or detects signals in text and models to trace provenance, ownership, and unauthorized use.
See all 122 papers