November 2026 · Independent humanizer benchmarks. Explore the evidence
← All humanizers

Qwen 2.5 14B Humanizer

Verified by HumanizeBenchWe ran this model ourselves on the benchmark source examples and checked its execution records. Verification confirms the tested model’s provenance; it does not guarantee performance or imply that every detector returned a result.
Open weights33 saved rewrites1 provider exclusion

Qwen 2.5 14B Humanizer has a current composite of 43.12/100 in our shared November evaluation. Readability is 78.1 and rescaled meaning preservation is 21.1. 230 of 231 detector measurements are available.

This entry is the requested peft-lora artifact from arunsingh80475/qwen2.5-14b-humanizer-v4-lora. Related formats are separate roster entries, not necessarily independent underlying models.

Overall score43.12/100
Panel bypass score66.2%
Readability78.1%
Prompts tested33

Detector human-likelihood scores

Mean reported human score on measured outputs. Missing results stay missing. Coverage is shown beside each detector.

GPTZero
29.333/33
ZeroGPT
30.133/33
Copyleaks
66.733/33
Pangram
71.133/33
Grammarly
88.633/33
QuillBot
99.833/33

Performance by category

HumanizeBench v2 per-category values.

Academic essay53.8Application essay41.5Blog post55.7Business email64.1Marketing copy75.9Discussion board56.8News article64.9

Quality and penalties

0.7892Raw meaning cosine
0/33Identical outputs
0/33Refusals
15.0Penalty points
RuleOccurrencesDeduction
Meaning drift25−10
Length deflation5−5

Artifact & evaluation record

Entry
arunsingh80475/qwen2.5-14b-humanizer-v4-lora
Source
Recorded GPU generation
Family
qwen25-humanizer-v4
Pinned revision
960e837fe40d0411eb7a91ff51c5f012ffb5aab2
Scoring
56.5% bypass · 24% meaning · 12% readability · 7.5% consistency, minus fixed penalties.
Measurement notes
NVIDIA quality substitutes; current Polygraph routes; no historical-score fallback.

Inspect the recorded text

The source and rewritten text are available for every sample.

SampleCategorySource modelInput wordsReadabilityDetector scores
S01 Academic essayAcademic essaygpt-5.6-terra31582.37/7
S02 Academic essayAcademic essayclaude-sonnet-529678.07/7
S03 Application essayApplication essaygemini-3.5-flash29173.37/7
S04 Application essayApplication essaygpt-5.6-luna27170.07/7
S05 Blog postBlog postclaude-haiku-4-5-2025100128070.07/7
S06 Blog postBlog postgemini-2.5-flash32980.07/7
S07 Business emailBusiness emailgpt-5.6-terra26280.07/7
S08 Marketing copyMarketing copyclaude-sonnet-529981.37/7
S09 Marketing copyMarketing copygemini-3.5-flash19180.07/7
S10 Discussion boardDiscussion boardgpt-5.6-luna26878.77/7
S11 News articleNews articleclaude-haiku-4-5-2025100130081.37/7
S12 Academic essayAcademic essaygemini-2.5-flash25981.37/7
S13 Academic essayAcademic essaygpt-5.6-terra34981.37/7
S14 Application essayApplication essayclaude-sonnet-531476.77/7
S15 Application essayApplication essaygemini-3.5-flash29275.37/7
S16 Blog postBlog postgpt-5.6-luna31380.07/7
S17 Blog postBlog postclaude-haiku-4-5-2025100128176.06/7
S18 Business emailBusiness emailgemini-2.5-flash18882.37/7
S19 Marketing copyMarketing copygpt-5.6-terra25773.37/7
S20 Marketing copyMarketing copyclaude-sonnet-519180.07/7
S21 Discussion boardDiscussion boardgemini-3.5-flash25180.07/7
S22 News articleNews articlegpt-5.6-luna29872.07/7
S23 Academic essayAcademic essayclaude-haiku-4-5-2025100131270.07/7
S24 Academic essayAcademic essaygemini-2.5-flash25870.07/7
S25 Application essayApplication essaygpt-5.6-terra30080.07/7
S26 Application essayApplication essayclaude-sonnet-532980.07/7
S27 Blog postBlog postgemini-3.5-flash29880.07/7
S28 Blog postBlog postgpt-5.6-luna32380.07/7
S29 Business emailBusiness emailclaude-haiku-4-5-2025100122882.37/7
S30 Marketing copyMarketing copygemini-2.5-flash27781.37/7
S31 Marketing copyMarketing copygpt-5.6-terra19578.07/7
S32 Discussion boardDiscussion boardclaude-sonnet-525280.07/7
S33 News articleNews articlegemini-3.5-flash29582.37/7