Hy3 Behavioral Fingerprint
How Tencent's Hy3 answers one-word probe questions at temperature 1.0 — random numbers, colors, animals, coin flips — measured over 355 valid samples in the open dataset of "One Token Is Enough" (arXiv:2607.10252).
Vendor
Tencent
Data generated
2026-07-08
Probe cells
15
Valid samples
355
Randomness score
16 / 100
Mean normalized entropy across cells, 0–100. 100 would be a uniform-random baseline; real models score far lower.
Signature answers
Favorite random number (1–100)
47
26 / 26 answers
Favorite color
teal
9 / 24 answers
Coin flip: heads
100%
Share of heads across coin-flip cells.
Answer distributions by probe cell
Empirical distributions of normalized one-word answers, per task and language. H is the Shannon entropy in bits; the uniform baseline is log2 of the answer-domain size.
Coin flip · English
n = 24 · H = 0.00 bit · uniform baseline 1.00 bit
| Answer | Count | Share |
|---|---|---|
| heads | 24 | 100.0% |
Coin flip · Chinese
n = 22 · H = 0.00 bit · uniform baseline 1.00 bit
| Answer | Count | Share |
|---|---|---|
| heads | 22 | 100.0% |
Favorite number · English
n = 25 · H = 0.00 bit · uniform baseline 13.29 bit
| Answer | Count | Share |
|---|---|---|
| 42 | 25 | 100.0% |
Favorite number · Chinese
n = 26 · H = 0.00 bit · uniform baseline 13.29 bit
| Answer | Count | Share |
|---|---|---|
| 7 | 26 | 100.0% |
Random animal · English
n = 25 · H = 1.74 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| axolotl | 10 | 40.0% |
| octopus | 7 | 28.0% |
| platypus | 7 | 28.0% |
| tiger | 1 | 4.0% |
Random animal · Chinese
n = 27 · H = 2.79 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| 猫 | 8 | 29.6% |
| 刺猬 | 5 | 18.5% |
| 豹 | 4 | 14.8% |
| 狐 | 3 | 11.1% |
| 狐狸 | 3 | 11.1% |
| 羊 | 1 | 3.7% |
| 企鹅 | 1 | 3.7% |
| 虎 | 1 | 3.7% |
+ 1 more answers
Random city · English
n = 26 · H = 2.11 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| oslo | 15 | 57.7% |
| tokyo | 3 | 11.5% |
| lyon | 2 | 7.7% |
| lisbon | 2 | 7.7% |
| kyoto | 1 | 3.8% |
| os | 1 | 3.8% |
| lima | 1 | 3.8% |
| lagos | 1 | 3.8% |
Random city · Chinese
n = 19 · H = 1.34 bit · uniform baseline 5.64 bit
| Answer | Count | Share |
|---|---|---|
| 重庆 | 14 | 73.7% |
| 成都 | 2 | 10.5% |
| 洛阳 | 1 | 5.3% |
| 贵阳 | 1 | 5.3% |
| 青岛 | 1 | 5.3% |
Random color · English
n = 24 · H = 1.72 bit · uniform baseline 4.91 bit
| Answer | Count | Share |
|---|---|---|
| teal | 9 | 37.5% |
| mauve | 9 | 37.5% |
| magenta | 5 | 20.8% |
| purple | 1 | 4.2% |
Random color · Chinese
n = 23 · H = 1.85 bit · uniform baseline 4.91 bit
| Answer | Count | Share |
|---|---|---|
| 靛 | 10 | 43.5% |
| 蓝 | 6 | 26.1% |
| 青 | 4 | 17.4% |
| 紫 | 3 | 13.0% |
Random letter · English
n = 22 · H = 0.00 bit · uniform baseline 4.70 bit
| Answer | Count | Share |
|---|---|---|
| q | 22 | 100.0% |
Random number 1-10 · English
n = 22 · H = 0.00 bit · uniform baseline 3.32 bit
| Answer | Count | Share |
|---|---|---|
| 7 | 22 | 100.0% |
Random number 1-10 · Chinese
n = 20 · H = 0.00 bit · uniform baseline 3.32 bit
| Answer | Count | Share |
|---|---|---|
| 7 | 20 | 100.0% |
Random number 1-100 · English
n = 26 · H = 0.00 bit · uniform baseline 6.64 bit
| Answer | Count | Share |
|---|---|---|
| 47 | 26 | 100.0% |
Random number 1-100 · Chinese
n = 24 · H = 1.29 bit · uniform baseline 6.64 bit
| Answer | Count | Share |
|---|---|---|
| 47 | 18 | 75.0% |
| 42 | 2 | 8.3% |
| 73 | 2 | 8.3% |
| 37 | 1 | 4.2% |
| 57 | 1 | 4.2% |
Want to verify your API really serves Hy3?
Point the free fingerprint checker at your endpoint: it samples the same probe questions from your browser and compares the distributions against this reference. Your API key never leaves your browser.
Verify your endpointData source & license
Distributions: Tomáš Bruckner, "One Token Is Enough" (arXiv:2607.10252), official dataset Zenodo DOI 10.5281/zenodo.21278557, licensed CC-BY-4.0. Collected via OpenRouter at temperature 1.0 under the paper's fixed minimal one-word system prompt; answer keys are re-normalized with the checker's pipeline (color aliases, number words, coin h/t) so live probes are directly comparable.