How many words must an imitator change to evade a word n-gram human-vs-machine detector that it can read?

What it found

A word 1-2 gram logistic-regression detector (held-out AUROC 0.948 on poetry, reviews and wiki) is attacked by an imitator that knows its weights and deletes or swaps the k most machine-ward word types in each of its own texts. Deleting 10 word types drops AUROC to 0.738, 20 to 0.487 (chance) and 40 to 0.150; swapping 20 gives 0.317. A detector refit on text attacked the same way recovers to 0.94–1.00. Stylised: scripted edits with no fluency check, one seed, one detector family, no LLM run.

A figure from its own output

Paper, code and results

• README and method
• Experiments and results
• Tests

More in Adversarial human imitation

• How many detector queries does an imitator need to evade a human-vs-machine text detector it cannot read?How many detector queries does an imitator need to evade a word n-gram human-vs-machine detector it cannot read? About 100 score queries per text, deleting the 20 best word types, take held-out AUROC from 0.948 to 0.520 (white-box 0.487); blind deletion does nothing. Stylised, no fluency check.
• Does refitting a human-vs-machine text detector on attacked text stop a black-box evader?Does refitting a text detector on attacked text stop a black-box evader? It settles after one refit: fresh-attack AUROC stays near 0.81 (from 0.719) while clean AUROC falls from 0.948 to about 0.90. Stylised, no fluency check.
• Does a prompt-free human-vs-machine detector carry over to a generator or decoding it never saw?Does a prompt-free detector carry over to a generator or decoding it never saw? The mimic-contexts n-gram detector loses 0.01 to 0.02 AUROC on a chat model left out of the fit (0.96 to 0.98 left out) and 0.025 to 0.05 on a left-out decoding (sampling 0.936); averaging over four same-source texts lifts AUROC to 0.999. Preliminary, plain-prompt RAID text only.
• Does adding a zero-shot perplexity score to the n-gram detector help it catch machine text in families it never saw?Does adding a zero-shot perplexity score to the n-gram detector help on unseen families? The two scores are uncorrelated within class, but averaging them lifts held-out AUROC only from 0.926 to 0.935 (paired gain +0.009, interval −0.010 to +0.026); fitted stacks do no better. Preliminary, 720 held-out texts, small scorer, no LLM run.
• How much text does a prompt-free human-vs-machine detector need, and what does it key on?How much text does a prompt-free detector need, and what does it key on? The mimic-contexts n-gram detector on held-out families: AUROC 0.65 at 50 characters, 0.84 at 400, 0.93 at 1500; stripping punctuation or shuffling words costs at most 0.014. Preliminary, plain-prompt RAID text only.
• If a human-vs-machine detector's threshold is set for a 5% false-positive rate, does that rate hold on a kind of text it never saw?If a detector's threshold is set for a 5% false-positive rate, does it hold on an unseen text type? Leave-one-family-out with the mimic-contexts n-gram detector: realised rate 0.00 to 0.54 (abstracts 0.54, books 0.18), and recipes catch 1.7% of machine text. Preliminary, plain-prompt RAID text only.

All 41 Adversarial human imitation projects →

Other research topics

Proper scoring & elicitation28Effort & peer prediction9Verification games & slashing19Detection & tolerances14Markets as routers13Prediction-market design9Paying contributors8Decentralised training systems23Digital twins & sim-to-real164LLM chip design loops40Robot manufacturing loops53Ambient AI assistant46Self-replicating manufacturing34Simulation-first design64Engineered clothing45Task-specific AI accelerators40Algorithmic chip-design loops42Open-source purpose-built robots44The software-engineering loop43Adversarial human imitation41Open-source residential construction41Autonomous multi-angle capture34Sensing, imaging and 3D31Energy supply and dense energy storage39Robot societies among people31Agentic co-design of low-cost robots27Laundry-hanging robots25Machine learning on qualitative data23Robots on the building site27Civilizations & group psychology1History & human sciences1

Produced with AI assistance (Claude). Results are from the code in this repository and are simulated or stylised unless the abstract says otherwise; nothing here implies affiliation with or endorsement by any lab, researcher or company cited. MIT licensed.