Hugging Face's logo

microsoft
/
Phi-3.5-mini-instruct

Text Generation
Transformers
Safetensors
multilingual
phi3
nlp
code
conversational
custom_code
text-generation-inference
Inference Endpoints
Model card Files Community
28
daekeun-ml commited on
Commit
eacb926
•
1 Parent(s): cd6881a

Updated README.md. Added Korean benchmarks.

Browse files
Files changed (1) hide show
  1. README.md +84 -1
README.md CHANGED
@@ -373,4 +373,87 @@ This project may contain trademarks or logos for projects, products, or services
373
  |-----------|-----------------------|---------------------------------------|--------------------------|---------------------------|------------------|----------------|------------------|-------------------------------|
374
  | English | 94.6 | 94.6 | 85.6 | 94.4 | 37.6 | 63.8 | 92.0 | 98.2 |
375
  | Italian | 86.8 | 84.8 | 76.8 | 83.2 | 16.2 | 37.2 | 85.6 | 97.6 |
376
- | Turkish | 58.6 | 57.2 | 61.6 | 56.6 | 38.4 | 60.2 | 91.4 | 94.6 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
373
  |-----------|-----------------------|---------------------------------------|--------------------------|---------------------------|------------------|----------------|------------------|-------------------------------|
374
  | English | 94.6 | 94.6 | 85.6 | 94.4 | 37.6 | 63.8 | 92.0 | 98.2 |
375
  | Italian | 86.8 | 84.8 | 76.8 | 83.2 | 16.2 | 37.2 | 85.6 | 97.6 |
376
+ | Turkish | 58.6 | 57.2 | 61.6 | 56.6 | 38.4 | 60.2 | 91.4 | 94.6 |
377
+
378
+
379
+ ## Appendix B: Korean benchmarks
380
+
381
+ The prompt is the same as the CLIcK paper prompt. The experimental results below were given with max_tokens=512 (zero-shot), max_tokens=1024 (5-shot), temperature=0.01. No system prompt used.
382
+
383
+
384
+ ### CLIcK
385
+
386
+ #### Accuracy by supercategory
387
+ | supercategory | Phi-3.5-Mini-Instruct | Phi-3.0-Mini-128k-Instruct (June2024) | Llama-3.1-8B-Instruct | GPT-4o | GPT-4o-mini | GPT-4-turbo | GPT-3.5-turbo |
388
+ |:----------------|------------------------:|--------------------------------:|------------------------:|---------:|--------------:|--------------:|----------------:|
389
+ | Culture | 43.77 | 29.74 | 51.15 | 81.89 | 70.95 | 73.61 | 53.38 |
390
+ | Language | 41.38 | 27.85 | 40.92 | 77.54 | 63.54 | 71.23 | 46 |
391
+ | **Overall** | 42.99 | 29.12 | 47.82 | 80.46 | 68.5 | 72.82 | 50.98 |
392
+
393
+ #### Accuracy by category
394
+ | supercategory | category | Phi-3.5-Mini-Instruct | Phi-3.0-Mini-128k-Instruct (June2024) | Llama-3.1-8B-Instruct | GPT-4o | GPT-4o-mini | GPT-4-turbo | GPT-3.5-turbo |
395
+ |:----------------|:------------|------------------------:|--------------------------------:|------------------------:|---------:|--------------:|--------------:|----------------:|
396
+ | Culture | Economy | 61.02 | 28.81 | 66.1 | 94.92 | 83.05 | 89.83 | 64.41 |
397
+ | Culture | Geography | 45.8 | 29.01 | 54.2 | 80.15 | 77.86 | 82.44 | 53.44 |
398
+ | Culture | History | 26.15 | 30 | 29.64 | 66.92 | 48.4 | 46.4 | 31.79 |
399
+ | Culture | Law | 32.42 | 22.83 | 44.29 | 70.78 | 57.53 | 61.19 | 41.55 |
400
+ | Culture | Politics | 54.76 | 33.33 | 59.52 | 88.1 | 83.33 | 89.29 | 65.48 |
401
+ | Culture | Pop Culture | 60.98 | 34.15 | 60.98 | 97.56 | 85.37 | 92.68 | 75.61 |
402
+ | Culture | Society | 54.37 | 31.72 | 65.05 | 92.88 | 85.44 | 86.73 | 71.2 |
403
+ | Culture | Tradition | 47.75 | 31.98 | 54.95 | 87.39 | 74.77 | 79.28 | 55.86 |
404
+ | Language | Functional | 37.6 | 24 | 32.8 | 84.8 | 64.8 | 80 | 40 |
405
+ | Language | Grammar | 27.5 | 23.33 | 22.92 | 57.08 | 42.5 | 47.5 | 30 |
406
+ | Language | Textual | 54.74 | 33.33 | 59.65 | 91.58 | 80.7 | 87.37 | 62.11 |
407
+
408
+ ### HAERAE
409
+
410
+ | category | Phi-3.5-Mini-Instruct | Phi-3.0-Mini-128k-Instruct (June2024) | Llama-3.1-8B-Instruct | GPT-4o | GPT-4o-mini | GPT-4-turbo | GPT-3.5-turbo |
411
+ |:----------------------|------------------------:|--------------------------------:|------------------------:|---------:|--------------:|--------------:|----------------:|
412
+ | General Knowledge | 31.25 | 28.41 | 34.66 | 77.27 | 53.41 | 66.48 | 40.91 |
413
+ | History | 32.45 | 22.34 | 44.15 | 92.02 | 84.57 | 78.72 | 30.32 |
414
+ | Loan Words | 47.93 | 35.5 | 63.31 | 79.88 | 76.33 | 78.11 | 59.17 |
415
+ | Rare Words | 55.06 | 42.96 | 63.21 | 87.9 | 81.98 | 79.01 | 61.23 |
416
+ | Reading Comprehension | 42.95 | 41.16 | 51.9 | 85.46 | 77.18 | 80.09 | 56.15 |
417
+ | Standard Nomenclature | 44.44 | 32.68 | 58.82 | 88.89 | 75.82 | 79.08 | 53.59 |
418
+ | **Overall** | 44.21 | 36.41 | 53.9 | 85.7 | 76.4 | 77.76 | 52.67 |
419
+
420
+ ### KMMLU
421
+
422
+ | supercategory | Phi-3.5-Mini-Instruct | Phi-3.0-Mini-128k-Instruct (June2024) | Llama-3.1-8B-Instruct | GPT-4o | GPT-4o-mini | GPT-4-turbo | GPT-3.5-turbo |
423
+ |:----------------|------------------------:|--------------------------------:|------------------------:|---------:|--------------:|--------------:|----------------:|
424
+ | Applied Science | 35.8 | 31.68 | 37.03 | 61.52 | 49.29 | 55.98 | 38.47 |
425
+ | HUMSS | 31.56 | 26.47 | 37.29 | 69.45 | 56.59 | 63 | 40.9 |
426
+ | Other | 35.45 | 31.01 | 39.15 | 63.79 | 52.35 | 57.53 | 40.19 |
427
+ | STEM | 38.54 | 31.9 | 40.42 | 65.16 | 54.74 | 60.84 | 42.24 |
428
+ | **Overall** | 35.87 | 30.82 | 38.54 | 64.26 | 52.63 | 58.75 | 40.3 |
429
+
430
+ ### KMMLU (5-shot)
431
+
432
+ | supercategory | Phi-3.5-Mini-Instruct | Phi-3.0-Mini-128k-Instruct (June2024)e | Llama-3.1-8B-Instruct | GPT-4o | GPT-4o-mini | GPT-4-turbo | GPT-3.5-turbo |
433
+ |:----------------|------------------------:|--------------------------------:|------------------------:|---------:|--------------:|--------------:|----------------:|
434
+ | Applied Science | 35.8 | 31.68 | 37.03 | 61.52 | 49.29 | 55.98 | 38.47 |
435
+ | HUMSS | 31.56 | 26.47 | 37.29 | 69.45 | 56.59 | 63 | 40.9 |
436
+ | Other | 35.45 | 31.01 | 39.15 | 63.79 | 52.35 | 57.53 | 40.19 |
437
+ | STEM | 38.54 | 31.9 | 40.42 | 65.16 | 54.74 | 60.84 | 42.24 |
438
+ | **Overall** | 35.87 | 30.82 | 38.54 | 64.26 | 52.63 | 58.75 | 40.3 |
439
+
440
+
441
+ ### KMMLU-HARD
442
+
443
+ | supercategory | Phi-3.5-Mini-Instruct | Phi-3.0-Mini-128k-Instruct (June2024) | Llama-3.1-8B-Instruct | GPT-4o | GPT-4o-mini | GPT-4-turbo | GPT-3.5-turbo |
444
+ |:----------------|------------------------:|--------------------------------:|------------------------:|---------:|--------------:|--------------:|----------------:|
445
+ | Applied Science | 27.08 | 26.17 | 26.25 | 37.12 | 22.25 | 29.17 | 21.07 |
446
+ | HUMSS | 20.21 | 24.38 | 20.21 | 41.97 | 23.31 | 31.51 | 19.44 |
447
+ | Other | 23.05 | 24.82 | 23.88 | 40.39 | 26.48 | 29.59 | 22.22 |
448
+ | STEM | 24.36 | 26.91 | 24.64 | 39.82 | 26.36 | 32.18 | 20.91 |
449
+ | **Overall** | 24 | 25.68 | 24.03 | 39.62 | 24.56 | 30.56 | 20.97 |
450
+
451
+ ### KMMLU-HARD (5-shot)
452
+
453
+ | supercategory | Phi-3.5-Mini-Instruct | Phi-3.0-Mini-128k-Instruct (June2024) | Llama-3.1-8B-Instruct | GPT-4o | GPT-4o-mini | GPT-4-turbo | GPT-3.5-turbo |
454
+ |:----------------|------------------------:|--------------------------------:|------------------------:|---------:|--------------:|--------------:|----------------:|
455
+ | Applied Science | 27.08 | 26.17 | 26.25 | 37.12 | 22.25 | 29.17 | 21.07 |
456
+ | HUMSS | 20.21 | 24.38 | 20.21 | 41.97 | 23.31 | 31.51 | 19.44 |
457
+ | Other | 23.05 | 24.82 | 23.88 | 40.39 | 26.48 | 29.59 | 22.22 |
458
+ | STEM | 24.36 | 26.91 | 24.64 | 39.82 | 26.36 | 32.18 | 20.91 |
459
+ | **Overall** | 24 | 25.68 | 24.03 | 39.62 | 24.56 | 30.56 | 20.97 |