File size: 5,111 Bytes
41e1370
 
 
 
 
 
 
 
 
 
 
82700c8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: cc-by-nc-2.0
---
# SOLAR-10.7B-Instruct-v1.0-laser

This version of Solar-10.7B was lasered and perplexity was calculated against gsm8k.

+ Initial Model Perplexity: 12.865185737609863
+ New baseline perplexity: 12.554274559020996

The laser process decreased perplexity by 2.41%
|                                                Model                                                |AGIEval|GPT4All|TruthfulQA|Bigbench|Average|
|-----------------------------------------------------------------------------------------------------|------:|------:|---------:|-------:|------:|
|[SOLAR-10.7B-Instruct-v1.0-laser](https://huggingface.co/macadeliccc/SOLAR-10.7B-Instruct-v1.0-laser)|   46.9|  74.99|     70.64|   43.74|  59.07|

### AGIEval
|             Task             |Version| Metric |Value|   |Stderr|
|------------------------------|------:|--------|----:|---|-----:|
|agieval_aqua_rat              |      0|acc     |29.53|±  |  2.87|
|                              |       |acc_norm|28.35|±  |  2.83|
|agieval_logiqa_en             |      0|acc     |39.78|±  |  1.92|
|                              |       |acc_norm|40.55|±  |  1.93|
|agieval_lsat_ar               |      0|acc     |23.04|±  |  2.78|
|                              |       |acc_norm|21.30|±  |  2.71|
|agieval_lsat_lr               |      0|acc     |51.18|±  |  2.22|
|                              |       |acc_norm|51.76|±  |  2.21|
|agieval_lsat_rc               |      0|acc     |66.54|±  |  2.88|
|                              |       |acc_norm|66.91|±  |  2.87|
|agieval_sat_en                |      0|acc     |78.16|±  |  2.89|
|                              |       |acc_norm|78.16|±  |  2.89|
|agieval_sat_en_without_passage|      0|acc     |50.97|±  |  3.49|
|                              |       |acc_norm|50.00|±  |  3.49|
|agieval_sat_math              |      0|acc     |42.73|±  |  3.34|
|                              |       |acc_norm|38.18|±  |  3.28|

Average: 46.9%

### GPT4All
|    Task     |Version| Metric |Value|   |Stderr|
|-------------|------:|--------|----:|---|-----:|
|arc_challenge|      0|acc     |60.84|±  |  1.43|
|             |       |acc_norm|63.99|±  |  1.40|
|arc_easy     |      0|acc     |83.59|±  |  0.76|
|             |       |acc_norm|81.44|±  |  0.80|
|boolq        |      1|acc     |87.58|±  |  0.58|
|hellaswag    |      0|acc     |68.11|±  |  0.47|
|             |       |acc_norm|85.77|±  |  0.35|
|openbookqa   |      0|acc     |35.40|±  |  2.14|
|             |       |acc_norm|48.40|±  |  2.24|
|piqa         |      0|acc     |80.58|±  |  0.92|
|             |       |acc_norm|80.74|±  |  0.92|
|winogrande   |      0|acc     |77.03|±  |  1.18|

Average: 74.99%

### TruthfulQA
|    Task     |Version|Metric|Value|   |Stderr|
|-------------|------:|------|----:|---|-----:|
|truthfulqa_mc|      1|mc1   |55.45|±  |  1.74|
|             |       |mc2   |70.64|±  |  1.49|

Average: 70.64%

### Bigbench
|                      Task                      |Version|       Metric        |Value|   |Stderr|
|------------------------------------------------|------:|---------------------|----:|---|-----:|
|bigbench_causal_judgement                       |      0|multiple_choice_grade|57.37|±  |  3.60|
|bigbench_date_understanding                     |      0|multiple_choice_grade|62.87|±  |  2.52|
|bigbench_disambiguation_qa                      |      0|multiple_choice_grade|35.66|±  |  2.99|
|bigbench_geometric_shapes                       |      0|multiple_choice_grade|33.15|±  |  2.49|
|                                                |       |exact_str_match      | 0.00|±  |  0.00|
|bigbench_logical_deduction_five_objects         |      0|multiple_choice_grade|26.20|±  |  1.97|
|bigbench_logical_deduction_seven_objects        |      0|multiple_choice_grade|19.71|±  |  1.50|
|bigbench_logical_deduction_three_objects        |      0|multiple_choice_grade|45.00|±  |  2.88|
|bigbench_movie_recommendation                   |      0|multiple_choice_grade|39.00|±  |  2.18|
|bigbench_navigate                               |      0|multiple_choice_grade|51.20|±  |  1.58|
|bigbench_reasoning_about_colored_objects        |      0|multiple_choice_grade|53.90|±  |  1.11|
|bigbench_ruin_names                             |      0|multiple_choice_grade|40.18|±  |  2.32|
|bigbench_salient_translation_error_detection    |      0|multiple_choice_grade|39.98|±  |  1.55|
|bigbench_snarks                                 |      0|multiple_choice_grade|63.54|±  |  3.59|
|bigbench_sports_understanding                   |      0|multiple_choice_grade|68.36|±  |  1.48|
|bigbench_temporal_sequences                     |      0|multiple_choice_grade|65.20|±  |  1.51|
|bigbench_tracking_shuffled_objects_five_objects |      0|multiple_choice_grade|22.48|±  |  1.18|
|bigbench_tracking_shuffled_objects_seven_objects|      0|multiple_choice_grade|18.46|±  |  0.93|
|bigbench_tracking_shuffled_objects_three_objects|      0|multiple_choice_grade|45.00|±  |  2.88|

Average: 43.74%

Average score: 59.07%

Elapsed time: 02:33:24