hindko-tokenizer / eval /chart_data.json
junaid008's picture
Add evaluation tables and charts
cc35156 verified
Raw History Blame Contribute Delete
3.51 kB
{
"chart_bytes_per_token_test.png": [
{
"label": "hindko-tokenizer (this, 32k)",
"bytes_per_token": 7.277325843823662,
"tokens_vs_released": 1.0
},
{
"label": "urnova-95m (best external)",
"bytes_per_token": 6.061296952279107,
"tokens_vs_released": 1.2006
},
{
"label": "RoBERTa-Urdu (UrduHack)",
"bytes_per_token": 5.708947266954404,
"tokens_vs_released": 1.2747
},
{
"label": "BLOOM",
"bytes_per_token": 5.121690021531185,
"tokens_vs_released": 1.4209
},
{
"label": "Gemma 3 / Gemma 4",
"bytes_per_token": 4.61036885467272,
"tokens_vs_released": 1.5785
},
{
"label": "GPT-4o (o200k)",
"bytes_per_token": 4.337584149418278,
"tokens_vs_released": 1.6777
},
{
"label": "Mistral Tekken (NeMo, Small 4)",
"bytes_per_token": 3.901016504418474,
"tokens_vs_released": 1.8655
},
{
"label": "Llama 4",
"bytes_per_token": 3.7268769536216326,
"tokens_vs_released": 1.9527
},
{
"label": "Qwen 3.5",
"bytes_per_token": 3.5982036581495,
"tokens_vs_released": 2.0225
},
{
"label": "DeepSeek-V3 / R1 / V4",
"bytes_per_token": 3.372220466292344,
"tokens_vs_released": 2.158
},
{
"label": "Llama 3",
"bytes_per_token": 2.64122087382617,
"tokens_vs_released": 2.7553
}
],
"chart_lm_bpb_dev_test.png": {
"confirm": [
{
"id": "R2-A4-SPnat-D2-32k",
"dev_delta_pct": -1.623655,
"dev_ci95": [
-2.107054,
-1.236059
],
"test_delta_pct": -1.0148512414859043,
"test_ci95": [
-1.3337703266599765,
-0.719411106580577
]
},
{
"id": "R2-A10-MinGram-P1r3-D2-48k",
"dev_delta_pct": -1.58176,
"dev_ci95": [
-1.867521,
-1.355869
],
"test_delta_pct": -1.3160397372781005,
"test_ci95": [
-1.467033141440053,
-1.1643594058582694
]
},
{
"id": "R2-A10-MinGram-P1r3-D2-32k",
"dev_delta_pct": -1.414409,
"dev_ci95": [
-1.66217,
-1.19142
],
"test_delta_pct": -1.253180889269175,
"test_ci95": [
-1.4797207661189518,
-1.03708384604859
]
},
{
"id": "R2-A1-P1r3-D2-48k",
"dev_delta_pct": -0.98924,
"dev_ci95": [
-1.211704,
-0.778742
],
"test_delta_pct": -0.850660355728833,
"test_ci95": [
-1.0468293493009588,
-0.6530838548134663
]
},
{
"id": "A1-P1r3-D2-32k",
"dev_delta_pct": -0.810497,
"dev_ci95": [
-1.000743,
-0.657493
],
"test_delta_pct": -0.5660213536703774,
"test_ci95": [
-0.7539943802074569,
-0.38067108247804304
]
}
],
"large": [
{
"id": "R2-A4-SPnat-D2-32k",
"dev_delta_pct": -1.335532,
"dev_ci95": [
-1.775676,
-1.026276
],
"test_delta_pct": -0.7701554030632566,
"test_ci95": [
-1.1521793550485442,
-0.3981661969961223
]
},
{
"id": "R2-A10-MinGram-P1r3-D2-32k",
"dev_delta_pct": -0.715001,
"dev_ci95": [
-0.910685,
-0.524984
],
"test_delta_pct": -0.4672962320113417,
"test_ci95": null
},
{
"id": "R2-A10-MinGram-P1r3-D2-48k",
"dev_delta_pct": -0.436436,
"dev_ci95": [
-0.622803,
-0.286684
],
"test_delta_pct": -0.25657526280652554,
"test_ci95": [
-0.4354746500167368,
-0.08159379637904507
]
},
{
"id": "A1-P1r3-D2-32k",
"dev_delta_pct": 0.027361,
"dev_ci95": [
-0.296706,
0.317218
],
"test_delta_pct": null,
"test_ci95": null
}
]
}
}