Spaces:
Running on Zero
Running on Zero
File size: 19,058 Bytes
d8ea2d7 0155c5c 7b7de26 d8ea2d7 810b119 0155c5c 7b7de26 810b119 d8ea2d7 0155c5c 7b7de26 d8ea2d7 1db83b6 d8ea2d7 810b119 0155c5c 7b7de26 810b119 1db83b6 810b119 0155c5c 7b7de26 810b119 0155c5c 7b7de26 810b119 1db83b6 810b119 76cd6be 0155c5c 76cd6be d8ea2d7 0155c5c d8ea2d7 0155c5c d8ea2d7 1db83b6 d8ea2d7 0155c5c d8ea2d7 0155c5c d8ea2d7 1db83b6 d8ea2d7 0155c5c d8ea2d7 0155c5c d8ea2d7 1db83b6 d8ea2d7 0155c5c d8ea2d7 0155c5c d8ea2d7 1db83b6 d8ea2d7 0155c5c 7b7de26 d8ea2d7 0155c5c 7b7de26 d8ea2d7 0155c5c 7b7de26 d8ea2d7 0155c5c 7b7de26 d8ea2d7 0155c5c 7b7de26 d8ea2d7 0155c5c 7b7de26 d8ea2d7 0155c5c 7b7de26 d8ea2d7 0155c5c 7b7de26 d8ea2d7 0155c5c 7b7de26 d8ea2d7 0155c5c 7b7de26 d8ea2d7 635caac 0155c5c 7b7de26 635caac 0155c5c 7b7de26 635caac 0155c5c 7b7de26 635caac 0155c5c 7b7de26 635caac 0155c5c 7b7de26 635caac d8ea2d7 1db83b6 7b7de26 1db83b6 7b7de26 1db83b6 7b7de26 1db83b6 7b7de26 1db83b6 d8ea2d7 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 | {
"models": [
{
"id": "qwen3_vl_8b_instruct",
"name": "Qwen3-VL-8B-Instruct",
"model_id": "Qwen/Qwen3-VL-8B-Instruct",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"enabled": true,
"prompt": "Transcribe every piece of printed and handwritten text from this document image. Keep paragraph/line layout when obvious. Return plain text only.",
"parameters": {
"max_new_tokens": 1024,
"temperature": 0.1,
"top_p": 0.95,
"do_sample": false
},
"notes": "Best default config for clean extraction and low hallucination."
},
{
"id": "qwen3_vl_4b_instruct",
"name": "Qwen3-VL-4B-Instruct",
"model_id": "Qwen/Qwen3-VL-4B-Instruct",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"enabled": true,
"prompt": "Transcribe every piece of printed and handwritten text from this Hebrew document image. Keep line breaks and text order.",
"parameters": {
"max_new_tokens": 1024,
"temperature": 0.1,
"top_p": 0.95,
"do_sample": false
},
"notes": "Smallest Qwen3-VL baseline for fast OCR baselining."
},
{
"id": "qwen3_vl_8b_thinking",
"name": "Qwen3-VL-8B-Thinking",
"model_id": "Qwen/Qwen3-VL-8B-Thinking",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"enabled": true,
"prompt": "Use deliberate reasoning to first recover layout and then output final OCR text exactly as visible. Hide reasoning and return only plain text.",
"parameters": {
"max_new_tokens": 1536,
"temperature": 0.0,
"top_p": 1.0,
"extra_body": {
"enable_thinking": true
}
},
"notes": "Thinking mode usually helps on challenging handwriting and dense layouts."
},
{
"id": "qwen3_vl_4b_thinking",
"name": "Qwen3-VL-4B-Thinking",
"model_id": "Qwen/Qwen3-VL-4B-Thinking",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"enabled": true,
"prompt": "Use a deliberate OCR pass for printed and handwritten Hebrew text, then output only the final transcription.",
"parameters": {
"max_new_tokens": 1400,
"temperature": 0.0,
"top_p": 1.0,
"extra_body": {
"enable_thinking": true
}
},
"notes": "4B thinking pass for hard handwriting/low-contrast lines."
},
{
"id": "qwen3_vl_30b_a3b_instruct",
"name": "Qwen3-VL-30B-A3B-Instruct",
"model_id": "Qwen/Qwen3-VL-30B-A3B-Instruct",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"enabled": true,
"prompt": "OCR this Hebrew document image and return plain text only, preserving order, lines and spacing.",
"parameters": {
"max_new_tokens": 2048,
"temperature": 0.0,
"top_p": 0.95
},
"notes": "Higher-capacity Qwen3-VL checkpoint for benchmark ceiling."
},
{
"id": "qwen3_vl_30b_a3b_thinking",
"name": "Qwen3-VL-30B-A3B-Thinking",
"model_id": "Qwen/Qwen3-VL-30B-A3B-Thinking",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"enabled": true,
"prompt": "Think through layout and character ambiguity first, then output final OCR text only.",
"parameters": {
"max_new_tokens": 3072,
"temperature": 0.0,
"top_p": 1.0,
"extra_body": {
"enable_thinking": true
}
},
"notes": "Thinking-oriented 30B setup for the hardest mixed handwriting cases."
},
{
"id": "nemotron_ocr_v2",
"name": "Nemotron-OCR-v2",
"model_id": "nvidia/nemotron-ocr-v2",
"provider": "local_transformer",
"enabled": true,
"prompt": "Transcribe all printed and handwritten text in this Hebrew document image. Return only plain text.",
"parameters": {
"max_new_tokens": 1536,
"temperature": 0.0,
"top_p": 0.95
},
"notes": "Multilingual OCR model with strong layout and text recognition behavior."
},
{
"id": "gemma4_e4b_instruct",
"name": "Gemma4 0.4B (E4B) Instruct",
"model_id": "google/gemma-4-E4B-it",
"provider": "local_transformer",
"enabled": true,
"prompt": "You are doing strict OCR. Extract printed and handwritten Hebrew/English text only. Keep every line and punctuation as shown.",
"parameters": {
"max_new_tokens": 1024,
"temperature": 0.0,
"top_p": 0.9
},
"notes": "Gemma4 entry for the 4B family."
},
{
"id": "gemma4_e4b_thinking",
"name": "Gemma4 0.4B (E4B) Thinking",
"model_id": "google/gemma-4-E4B-it",
"provider": "local_transformer",
"enabled": true,
"prompt": "Think carefully through layout, then output only the final document transcription with line breaks preserved.",
"parameters": {
"max_new_tokens": 1536,
"temperature": 0.0,
"top_p": 1.0,
"extra_body": {
"enable_thinking": true
}
},
"notes": "Same checkpoint with higher budget and thinking-oriented system prompt."
},
{
"id": "gemma4_12b_instruct",
"name": "Gemma4 12B Instruct",
"model_id": "google/gemma-4-12B-it",
"provider": "local_transformer",
"enabled": true,
"prompt": "OCR only. Return all readable printed and handwritten text in reading order. Do not describe anything else.",
"parameters": {
"max_new_tokens": 1536,
"temperature": 0.0,
"top_p": 0.95
},
"notes": "Best default for instruction mode."
},
{
"id": "gemma4_12b_thinking",
"name": "Gemma4 12B Thinking",
"model_id": "google/gemma-4-12B-it",
"provider": "local_transformer",
"enabled": true,
"prompt": "Use internal planning for document structure, then output the final OCR text only.",
"parameters": {
"max_new_tokens": 2048,
"temperature": 0.0,
"top_p": 1.0,
"extra_body": {
"enable_thinking": true
}
},
"notes": "Thinking-oriented OCR prompt with larger decoding budget."
},
{
"id": "gemma4_26b4b_instruct",
"name": "Gemma4 26B-A4B Instruct",
"model_id": "google/gemma-4-26B-A4B-it",
"provider": "local_transformer",
"enabled": true,
"prompt": "Transcribe the document exactly. Keep whitespace between words/lines and don't clean the content.",
"parameters": {
"max_new_tokens": 2048,
"temperature": 0.0,
"top_p": 0.9
},
"notes": "Larger Gemma4-A4B OCR baseline."
},
{
"id": "gemma4_26b4b_thinking",
"name": "Gemma4 26B-A4B Thinking",
"model_id": "google/gemma-4-26B-A4B-it",
"provider": "local_transformer",
"enabled": true,
"prompt": "Resolve handwriting and layout ambiguities first, then return only the final cleaned OCR output.",
"parameters": {
"max_new_tokens": 2560,
"temperature": 0.0,
"top_p": 1.0,
"extra_body": {
"enable_thinking": true
}
},
"notes": "Thinking prompt + higher budget."
},
{
"id": "gemma4_31b_instruct",
"name": "Gemma4 31B Instruct",
"model_id": "google/gemma-4-31B-it",
"provider": "local_transformer",
"enabled": true,
"prompt": "Strict OCR pass: output only text from the image, maintaining line structure.",
"parameters": {
"max_new_tokens": 2048,
"temperature": 0.0,
"top_p": 0.95
},
"notes": "Highest-capacity Gemma4 variant in this panel."
},
{
"id": "gemma4_31b_thinking",
"name": "Gemma4 31B Thinking",
"model_id": "google/gemma-4-31B-it",
"provider": "local_transformer",
"enabled": true,
"prompt": "Do internal chain-of-thought for OCR confidence, then output only the final transcription.",
"parameters": {
"max_new_tokens": 3072,
"temperature": 0.0,
"top_p": 1.0,
"extra_body": {
"enable_thinking": true
}
},
"notes": "Thinking-oriented prompt for hard mixed-direction text."
},
{
"id": "chandra2",
"name": "Chandra2 (datalab-to/chandra-ocr-2)",
"model_id": "datalab-to/chandra-ocr-2",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Output only the exact text from the uploaded document, including handwritten and printed parts.",
"parameters": {
"max_new_tokens": 1024,
"temperature": 0.0
},
"notes": "Document OCR and layout-focused model."
},
{
"id": "ronylicha_gigapdf",
"name": "ronylicha/gigapdf-ocr-hebrew",
"model_id": "ronylicha/gigapdf-ocr-hebrew",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Read the Hebrew document image and return only transcribed text.",
"parameters": {
"max_new_tokens": 1024,
"temperature": 0.0
},
"notes": "Hebrew-first OCR model; sometimes better with Hebrew-only content."
},
{
"id": "qunie_v7_mini",
"name": "Qunie-V7-mini",
"model_id": "liskcell/Qunie-V7-mini",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Strict OCR extraction. Output line-by-line text from the whole image.",
"parameters": {
"max_new_tokens": 1536,
"temperature": 0.1,
"top_p": 0.96
},
"notes": "Balanced performance/speed config."
},
{
"id": "openmlkitocr",
"name": "OpenMLKitOCR",
"model_id": "0cve0/OpenMLKitOCR",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "OCR the document and return readable text as-is.",
"parameters": {
"max_new_tokens": 1024,
"temperature": 0.0
},
"notes": "Mobile-oriented OCR stack from ML Kit; tuned for mixed scans."
},
{
"id": "tzefa_word_ocr_trocr",
"name": "Tzefa-Word-OCR-TrOCR",
"model_id": "WARAJA/Tzefa-Word-OCR-TrOCR",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Do OCR on this image. Return only transcribed words in reading order.",
"parameters": {
"max_new_tokens": 1024,
"temperature": 0.0
},
"notes": "Word-level model behavior for clean printed text lines."
},
{
"id": "qunie_v7_pico",
"name": "Qunie-V7-Pico",
"model_id": "liskcell/Qunie-V7-Pico",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Extract all readable text. Prioritize transcription correctness over grammar cleanup.",
"parameters": {
"max_new_tokens": 1024,
"temperature": 0.15,
"top_p": 0.97
},
"notes": "Compact OCR model for faster response."
},
{
"id": "aya_vision_32b",
"name": "Aya Vision 32B",
"model_id": "CohereLabs/aya-vision-32b",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"enabled": true,
"prompt": "Return plain OCR output only from the image, including punctuation and line breaks.",
"parameters": {
"max_new_tokens": 2048,
"temperature": 0.0,
"top_p": 0.95
},
"notes": "High-capacity vision model; strong on multilingual text."
},
{
"id": "aya_vision_8b",
"name": "Aya Vision 8B",
"model_id": "CohereLabs/aya-vision-8b",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"enabled": true,
"prompt": "Return only text from the document image; preserve paragraph breaks where clear.",
"parameters": {
"max_new_tokens": 1536,
"temperature": 0.0,
"top_p": 0.95
},
"notes": "Balanced multilingual OCR speed/quality."
},
{
"id": "phi4_multimodal",
"name": "Phi-4-Multimodal-Instruct",
"model_id": "microsoft/Phi-4-multimodal-instruct",
"provider": "local_transformer",
"pipeline_task": "image-text-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "OCR-only transcription. Output exactly what is printed/handwritten in the image.",
"parameters": {
"max_new_tokens": 1536,
"temperature": 0.0,
"top_p": 0.95
},
"notes": "Keep short and factual output."
},
{
"id": "surya2",
"name": "surya2 (datalab-to/surya-ocr-2)",
"model_id": "datalab-to/surya-ocr-2",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Return only OCR output from this document image, preserving line breaks and order.",
"parameters": {
"max_new_tokens": 1536,
"temperature": 0.0,
"top_p": 0.95
},
"notes": "Specialized OCR pipeline with layout awareness."
},
{
"id": "paddleocr_vl",
"name": "PaddleOCR-VL",
"model_id": "PaddlePaddle/PaddleOCR-VL",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Transcribe all readable printed and handwritten text from this document image in order.",
"parameters": {
"max_new_tokens": 4096,
"temperature": 0.0
},
"notes": "PaddleOCR-VL base checkpoint."
},
{
"id": "paddleocr_vl_15",
"name": "PaddleOCR-VL 1.5",
"model_id": "PaddlePaddle/PaddleOCR-VL-1.5",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Transcribe all readable printed and handwritten text from this document image in order.",
"parameters": {
"max_new_tokens": 4096,
"temperature": 0.0
},
"notes": "PaddleOCR-VL family variant."
},
{
"id": "paddleocr_vl_16",
"name": "PaddleOCR-VL 1.6",
"model_id": "PaddlePaddle/PaddleOCR-VL-1.6",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Transcribe all readable printed and handwritten text from this document image in order.",
"parameters": {
"max_new_tokens": 4096,
"temperature": 0.0
},
"notes": "PaddleOCR-VL family variant."
},
{
"id": "deepseek_ocr_1",
"name": "DeepSeek OCR-1",
"model_id": "deepseek-ai/DeepSeek-OCR",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Transcribe all printed and handwritten text from this Hebrew document image. Output only plain text.",
"parameters": {
"max_new_tokens": 4096,
"temperature": 0.0
},
"notes": "DeepSeek OCR-1 family baseline."
},
{
"id": "deepseek_ocr_2",
"name": "DeepSeek OCR-2",
"model_id": "deepseek-ai/DeepSeek-OCR-2",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Transcribe all printed and handwritten text from this Hebrew document image. Preserve line breaks as visible.",
"parameters": {
"max_new_tokens": 4096,
"temperature": 0.0
},
"notes": "DeepSeek OCR-2 family OCR variant."
},
{
"id": "tesseract_latest",
"name": "Tesseract (local latest)",
"model_id": "tesseract",
"provider": "tesseract",
"enabled": true,
"parameters": {
"lang": "heb+eng",
"oem": 1,
"psm": 6,
"extra_config": "--dpi 300 --psm 6"
},
"prompt": "N/A",
"notes": "Always available local baseline for Hebrew/English and free."
},
{
"id": "cyttic_exp10_trocr_hebrew_matan_full",
"name": "cyttic/exp10-trocr-hebrew-matan-full",
"model_id": "cyttic/exp10-trocr-hebrew-matan-full",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Transcribe all readable printed and handwritten Hebrew text from this document image. Return plain text only, preserve line breaks and spacing.",
"parameters": {
"max_new_tokens": 4096,
"temperature": 0.0,
"top_p": 0.95
},
"notes": "Added for Hebrew-focused TrOCR-style local OCR benchmarking."
},
{
"id": "cyttic_exp23_directfit_unfrozen",
"name": "cyttic/exp23-directfit-unfrozen",
"model_id": "cyttic/exp23-directfit-unfrozen",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "OCR this document image and extract all printed and handwritten Hebrew text faithfully. Return only the plain transcription.",
"parameters": {
"max_new_tokens": 4096,
"temperature": 0.0,
"top_p": 0.95
},
"notes": "Added for offline direct-fit multilingual document OCR benchmarking."
},
{
"id": "cyttic_heb_verifier17_connected",
"name": "cyttic/heb-verifier17-connected",
"model_id": "cyttic/heb-verifier17-connected",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Return only the transcribed text from the uploaded document image (Hebrew document with printed and handwritten text). Preserve line order.",
"parameters": {
"max_new_tokens": 4096,
"temperature": 0.0,
"top_p": 0.95
},
"notes": "Added for verifier-focused Hebrew OCR comparison run."
},
{
"id": "cyttic_exp26_composed1m",
"name": "cyttic/exp26-composed1m",
"model_id": "cyttic/exp26-composed1m",
"provider": "local_transformer",
"pipeline_task": "image-to-text",
"trust_remote_code": true,
"enabled": true,
"prompt": "Transcribe exactly what is visible in the image, including Hebrew text lines and mixed-direction fragments.",
"parameters": {
"max_new_tokens": 4096,
"temperature": 0.0,
"top_p": 0.95
},
"notes": "Added for composed document OCR benchmark."
}
]
}
|