data: add trusted reference fingerprint for DeepSeek-V4.1-Flash (official api.deepseek.com)

Collected with this tool against https://api.deepseek.com/v1, model id
deepseek-flash (API name: DeepSeek-V4.1-Flash), standard preset (8 cells x 25
samples), protocol one-token/v1, reasoning disabled via the openai-effort
adapter. 200/200 valid, 0 errors.

Unlike the bundled Zenodo samples, this reference is protocol-matched to the
package battery, so compare()/verify() against it is a direct comparison.
This commit is contained in:
Hermes
2026-10-04 21:06:58 +08:00
parent 94dc44686a
commit a834a1cb8b
2 changed files with 212 additions and 0 deletions
+46
View File
@@ -0,0 +1,46 @@
# Reference fingerprints (trusted, collected with this tool)
These are **first-party** references — collected by `llm-fingerprint-detector`
itself against a provider you trust, using the package's own `one-token/v1`
battery. Unlike `../reference-fingerprints.sample.json` (Zenodo paper data,
`bruckner-zenodo-2026` protocol), these are protocol-matched, so a
`compare()` against them is a direct, non-indicative comparison.
| file | model id | provider | collected | protocol | battery |
|---|---|---|---|---|---|
| `deepseek-v4.1-flash.official.json` | `deepseek-flash` | api.deepseek.com (official) | 2026-10-04 | `one-token/v1` | 8 cells × 25 samples |
## `deepseek-v4.1-flash.official.json`
- **Endpoint:** `https://api.deepseek.com/v1`
- **Model id requested:** `deepseek-flash` (API-reported name: `DeepSeek-V4.1-Flash`)
- **Collected:** 2026-10-04, standard preset, `temperature=1`, reasoning disabled
via the `openai-effort` adapter
- **Result:** 200/200 samples valid, 0 errors, `postReasoning: false`
- **Use it to verify a reseller:** a reseller advertising `deepseek-v4.1-flash`
(or `deepseek-flash`) should land in the *match* band against this file.
```bash
# verify a suspect endpoint against this trusted reference
llm-fingerprint verify \
--base-url https://reseller.example.com/v1 \
--model deepseek-v4.1-flash \
--reference data/references/deepseek-v4.1-flash.official.json
```
### The official fingerprint's most discriminative cells
| cell | official DeepSeek-V4.1-Flash | notes |
|---|---|---|
| `random-number-1-100:en` | `42` ×19 / 25 | overwhelming 42 bias |
| `random-number-1-100:zh` | `42` ×14 / 25 | same bias, Chinese |
| `random-color:en` | `blue` ×25 / 25 | fully deterministic |
| `coin-flip:en` | `heads` ×25 / 25 | fully deterministic |
| `random-number-1-10:en` | `7` ×23 / 25 | near-deterministic |
| `random-letter:en` | `q` ×18 / 25 | q ≫ k ≫ g |
| `random-color:zh` | `蓝` ×20 / 25 | 蓝 ≫ 紫, 橙 |
| `random-animal:en` | `elephant` ×11, `cat` ×9 | elephant dominates |
**Caveat:** references drift as providers ship model updates. Re-collect if the
file is more than ~1–2 months old, or if a verification you trust starts
returning `uncertain`.
@@ -0,0 +1,166 @@
{
"formatVersion": 1,
"protocol": "one-token/v1",
"model": "deepseek-flash",
"collectedAt": "2026-10-04T12:59:45.985Z",
"samplesPerCell": 25,
"postReasoning": false,
"cells": {
"random-number-1-100:en": {
"cellId": "random-number-1-100:en",
"counts": {
"37": 1,
"42": 19,
"47": 2,
"72": 3
},
"validCount": 25,
"invalidCount": 0,
"refusalCount": 0,
"emptyCount": 0,
"errorCount": 0,
"totalCount": 25,
"entropyBits": 1.145235779471061,
"normalizedEntropy": 0.17237516086420482,
"medianLatencyMs": 118.67180000000008,
"meanCompletionTokens": 1,
"meanReasoningTokens": null
},
"random-number-1-100:zh": {
"cellId": "random-number-1-100:zh",
"counts": {
"37": 2,
"42": 14,
"47": 1,
"58": 1,
"73": 6
},
"validCount": 24,
"invalidCount": 1,
"refusalCount": 0,
"emptyCount": 0,
"errorCount": 0,
"totalCount": 25,
"entropyBits": 1.634431504340598,
"normalizedEntropy": 0.24600645433236223,
"medianLatencyMs": 125.12239999999929,
"meanCompletionTokens": 1,
"meanReasoningTokens": null
},
"random-color:en": {
"cellId": "random-color:en",
"counts": {
"blue": 25
},
"validCount": 25,
"invalidCount": 0,
"refusalCount": 0,
"emptyCount": 0,
"errorCount": 0,
"totalCount": 25,
"entropyBits": 0,
"normalizedEntropy": 0,
"medianLatencyMs": 121.21330000000125,
"meanCompletionTokens": 1,
"meanReasoningTokens": null
},
"random-animal:en": {
"cellId": "random-animal:en",
"counts": {
"cat": 9,
"elephant": 11,
"otter": 2,
"octopus": 2,
"owl": 1
},
"validCount": 25,
"invalidCount": 0,
"refusalCount": 0,
"emptyCount": 0,
"errorCount": 0,
"totalCount": 25,
"entropyBits": 1.8205332770550815,
"normalizedEntropy": 0.3225690407125254,
"medianLatencyMs": 121.40209999999934,
"meanCompletionTokens": 1.64,
"meanReasoningTokens": null
},
"random-number-1-10:en": {
"cellId": "random-number-1-10:en",
"counts": {
"4": 2,
"7": 23
},
"validCount": 25,
"invalidCount": 0,
"refusalCount": 0,
"emptyCount": 0,
"errorCount": 0,
"totalCount": 25,
"entropyBits": 0.4021791902022728,
"normalizedEntropy": 0.12106799988273365,
"medianLatencyMs": 122.51369999999997,
"meanCompletionTokens": 1,
"meanReasoningTokens": null
},
"random-letter:en": {
"cellId": "random-letter:en",
"counts": {
"q": 18,
"k": 5,
"g": 1,
"j": 1
},
"validCount": 25,
"invalidCount": 0,
"refusalCount": 0,
"emptyCount": 0,
"errorCount": 0,
"totalCount": 25,
"entropyBits": 1.1771245697587873,
"normalizedEntropy": 0.25042860675688255,
"medianLatencyMs": 122.89889999999991,
"meanCompletionTokens": 1,
"meanReasoningTokens": null
},
"random-color:zh": {
"cellId": "random-color:zh",
"counts": {
"蓝": 20,
"橙": 2,
"紫": 3
},
"validCount": 25,
"invalidCount": 0,
"refusalCount": 0,
"emptyCount": 0,
"errorCount": 0,
"totalCount": 25,
"entropyBits": 0.916118213778296,
"normalizedEntropy": 0.18670035451741826,
"medianLatencyMs": 118.93230000000005,
"meanCompletionTokens": 1,
"meanReasoningTokens": null
},
"coin-flip:en": {
"cellId": "coin-flip:en",
"counts": {
"heads": 25
},
"validCount": 25,
"invalidCount": 0,
"refusalCount": 0,
"emptyCount": 0,
"errorCount": 0,
"totalCount": 25,
"entropyBits": 0,
"normalizedEntropy": 0,
"medianLatencyMs": 121.60199999999986,
"meanCompletionTokens": 2,
"meanReasoningTokens": null
}
},
"meta": {
"tool": "llm-fingerprint-detector"
}
}