Stylometric fingerprints capture sentence length, Markdown usage, punctuation density, and bullet patterns — a coarse but cheap signal that survives even when logprobs are stripped.
Algorithm
Compute 8 features per response, average over ≥ 3 samples, and cosine-compare to per-model centroids stored in lib/fingerprints/. The best match is reported alongside the claimed model's score.
Thresholds
Condition
Verdict contribution
Best match = claimed model
Match
Best match within 0.05 cosine of claimed
Borderline
Otherwise
Mismatch
Limitations
Highly sensitive to prompt phrasing. Use ≥ 3 stable prompts and the same temperature across calls.
Anysingle signal cannot provemalicious behavior. Proxies may show anomalies for legitimate reasons (regional routing, A/B testing, degradation strategies, cache optimization).
Token ratio deviation may result from ChatML wrapping, system prompt injection, or tokenizer version differences — not necessarily intentional inflation.
Model identity judgment is based on statistical fingerprint matching, not cryptographic proof. Quantization, fine-tuning, and post-processing can all alter fingerprints.
MMD distribution tests are sensitive to temperature, sampling parameters, and system prompts. Significant p-values mean distributional difference, not proof of substitution.
Logprobs unavailability is increasingly common (many providers disable it by default in 2025-2026) and does not by itself indicate deception.
ITT rhythm fingerprinting is an early-stage technique. Network jitter, TCP coalescing, and gateway buffering can produce false signals.
This tool generates reference-grade evidence chains, not legal conclusions. Do not make definitive accusations based solely on this report.
The wording in the report refers to statistical "deviations" or "signal inconsistencies". Please do not use this to make fraud or deception claims against any service provider.