Hugging Face's logo

NECOUDBFM
/
Jellyfish-8B

Text Generation
Transformers
Safetensors
English
llama
conversational
text-generation-inference
Inference Endpoints
Model card Files Community
HCZhang commited on
Commit
92837d9
·
verified ·
1 Parent(s): 36b9faf

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +35 -25
README.md CHANGED
@@ -43,39 +43,48 @@ If you find our work useful, please give us credit by citing:
43
 
44
  ## Performance on seen tasks
45
 
46
- | Task | Type | Dataset | Non-LLM SoTA<sup>1</sup> | GPT-3 | GPT-3.5<sup>2</sup> | GPT-4<sup>2</sup> | GPT-4o<sup>2</sup> | Table-GPT | Jellyfish-7B | Jellyfish-8B | Jellyfish-13B |
47
- |-----------------|--------|-------------------|-----------------|--------|---------|--------|--------|-----------|--------------|--------------|---------------|
48
- | Error Detection | Seen | Adult | *99.10* | 99.10 | 92.01 | 92.01 | 83.58 | -- | 77.40 | 73.74 | **99.33** |
49
- | Error Detection | Seen | Hospital | 94.40 | **97.80** | 90.74 | 90.74 | 44.76 | -- | 94.51 | 93.40 | *95.59* |
50
- | Error Detection | Unseen | Flights | 81.00 | -- | -- | **83.48** | 66.01 | -- | 69.15 | 66.21 | *82.52* |
51
- | Error Detection | Unseen | Rayyan | 79.00 | -- | -- | *81.95* | 68.53 | -- | 75.07 | 81.06 | **90.65** |
52
- | Data Imputation | Seen | Buy | 96.50 | 98.50 | 98.46 | **100** | **100** | -- | 98.46 | 98.46 | **100** |
53
- | Data Imputation | Seen | Restaurant | 77.20 | 88.40 | *94.19* | **97.67** | 90.70 | -- | 89.53 | 87.21 | 89.53 |
54
- | Data Imputation | Unseen | Flipkart | 68.00 | -- | -- | **89.94** | 83.20 | -- | 87.14 | *87.48* | 81.68 |
55
- | Data Imputation | Unseen | Phone | 86.70 | -- | -- | **90.79** | 86.78 | -- | 86.52 | 85.68 | *87.21* |
56
- | Schema Matching | Seen | MIMIC-III | 20.00 | -- | -- | 40.00 | 29.41 | -- | **53.33** | *45.45* | 40.00 |
57
- | Schema Matching | Seen | Synthea | 38.50 | 45.20 | *57.14* | **66.67** | 6.56 | -- | 55.56 | 47.06 | 56.00 |
58
- | Schema Matching | Unseen | CMS | *50.00* | -- | -- | 19.35 | 22.22 | -- | 42.86 | 38.10 | **59.29** |
59
- | Entity Matching | Seen | Amazon-Google | 75.58 | 63.50 | 66.50 | 74.21 | 70.91 | 70.10 | **81.69** | *81.42* | 81.34 |
60
- | Entity Matching | Seen | Beer | 94.37 | **100** | 96.30 | **100** | 90.32 | 96.30 | **100.00** | **100.00** | 96.77 |
61
- | Entity Matching | Seen | DBLP-ACM | **98.99** | 96.60 | 96.99 | 97.44 | 95.87 | 93.80 | 98.65 | 98.77 | *98.98* |
62
- | Entity Matching | Seen | DBLP-GoogleScholar| *95.70* | 83.80 | 76.12 | 91.87 | 90.45 | 92.40 | 94.88 | 95.03 | **98.51** |
63
- | Entity Matching | Seen | Fodors-Zagats | **100** | **100** | **100** | **100** | 93.62 | **100** | **100** | **100** | **100** |
64
- | Entity Matching | Seen | iTunes-Amazon | 97.06 | *98.20*| 96.40 | **100** | 98.18 | 94.30 | 96.30 | 96.30 | 98.11 |
65
- | Entity Matching | Unseen | Abt-Buy | 89.33 | -- | -- | **92.77** | 78.73 | -- | 86.06 | 88.84 | *89.58* |
66
- | Entity Matching | Unseen | Walmart-Amazon | 86.89 | 87.00 | 86.17 | **90.27** | 79.19 | 82.40 | 84.91 | 85.24 | *89.42* |
67
- | Avg | | | 80.44 | - | - | *84.17* | 72.58 | - | 82.74 | 81.55 | **86.02** |
68
 
69
  _For GPT-3.5 and GPT-4, we used the few-shot approach on all datasets. However, for Jellyfish-13B and Jellyfish-Interpreter, the few-shot approach is disabled on seen datasets and enabled on unseen datasets._
70
  _Accuracy as the metric for data imputation and the F1 score for other tasks._
71
 
 
 
 
 
 
 
 
 
 
72
  ## Performance on unseen tasks
73
 
74
  ### Column Type Annotation
75
 
76
- | Dataset | RoBERTa (159 shots)<sup>1</sup> | GPT-3.5<sup>1</sup> | GPT-4 | Jellfish-13B| Jellyfish-7B | Jellyfish-8B |
77
- | ---- | ---- | ---- | ---- | ---- | ----|----|
78
- | SOTAB | 79.20 | 89.47 | 91.55 | 82.00 | 80.89 | 67.21|
79
 
80
  _Few-shot is disabled for Jellyfish-13B._
81
 
@@ -88,6 +97,7 @@ _Few-shot is disabled for Jellyfish-13B._
88
  | AE-110k | 52.10 | 49.20 | 61.30 | 55.50 | 58.12 | 76.85| 69.78|
89
  | OA-Mine | 50.80 | 55.20 | 62.70 | 68.90 | 55.96 | 76.04| 78.83|
90
 
 
91
 
92
  ## Prompt Template
93
  ```
 
43
 
44
  ## Performance on seen tasks
45
 
46
+ | Task | Type | Dataset | Non-LLM SoTA<sup>1</sup> | GPT-3.5<sup>2</sup> | GPT-4<sup>2</sup> | GPT-4o | Table-GPT | Jellyfish-7B | Jellyfish-8B | Jellyfish-13B |
47
+ |-----------------|--------|-------------------|-----------------|--------|--------|--------|-----------|--------------|--------------|---------------|
48
+ | Error Detection | Seen | Adult | *99.10* | 99.10 | 92.01 | 83.58 | -- | 77.40 | 73.74 | **99.33** |
49
+ | Error Detection | Seen | Hospital | 94.40 | **97.80** | 90.74 | 44.76 | -- | 94.51 | 93.40 | *95.59* |
50
+ | Error Detection | Unseen | Flights | 81.00 | -- | **83.48** | 66.01 | -- | 69.15 | 66.21 | *82.52* |
51
+ | Error Detection | Unseen | Rayyan | 79.00 | -- | *81.95* | 68.53 | -- | 75.07 | 81.06 | **90.65** |
52
+ | Data Imputation | Seen | Buy | 96.50 | 98.50 | **100** | **100** | -- | 98.46 | 98.46 | **100** |
53
+ | Data Imputation | Seen | Restaurant | 77.20 | 88.40 | **97.67** | 90.70 | -- | 89.53 | 87.21 | 89.53 |
54
+ | Data Imputation | Unseen | Flipkart | 68.00 | -- | **89.94** | 83.20 | -- | 87.14 | *87.48* | 81.68 |
55
+ | Data Imputation | Unseen | Phone | 86.70 | -- | **90.79** | 86.78 | -- | 86.52 | 85.68 | *87.21* |
56
+ | Schema Matching | Seen | MIMIC-III | 20.00 | -- | 40.00 | 29.41 | -- | **53.33** | *45.45* | 40.00 |
57
+ | Schema Matching | Seen | Synthea | 38.50 | 45.20 | **66.67** | 6.56 | -- | 55.56 | 47.06 | 56.00 |
58
+ | Schema Matching | Unseen | CMS | *50.00* | -- | 19.35 | 22.22 | -- | 42.86 | 38.10 | **59.29** |
59
+ | Entity Matching | Seen | Amazon-Google | 75.58 | 63.50 | 74.21 | 70.91 | 70.10 | **81.69** | *81.42* | 81.34 |
60
+ | Entity Matching | Seen | Beer | 94.37 | **100** | **100** | 90.32 | 96.30 | **100.00** | **100.00** | 96.77 |
61
+ | Entity Matching | Seen | DBLP-ACM | **98.99** | 96.60 | 97.44 | 95.87 | 93.80 | 98.65 | 98.77 | *98.98* |
62
+ | Entity Matching | Seen | DBLP-GoogleScholar| *95.70* | 83.80 | 91.87 | 90.45 | 92.40 | 94.88 | 95.03 | **98.51** |
63
+ | Entity Matching | Seen | Fodors-Zagats | **100** | **100** | **100** | 93.62 | **100** | **100** | **100** | **100** |
64
+ | Entity Matching | Seen | iTunes-Amazon | 97.06 | *98.20*| **100** | 98.18 | 94.30 | 96.30 | 96.30 | 98.11 |
65
+ | Entity Matching | Unseen | Abt-Buy | 89.33 | -- | **92.77** | 78.73 | -- | 86.06 | 88.84 | *89.58* |
66
+ | Entity Matching | Unseen | Walmart-Amazon | 86.89 | 87.00 | **90.27** | 79.19 | 82.40 | 84.91 | 85.24 | *89.42* |
67
+ | Avg | | | 80.44 | - | *84.17* | 72.58 | - | 82.74 | 81.55 | **86.02** |
68
 
69
  _For GPT-3.5 and GPT-4, we used the few-shot approach on all datasets. However, for Jellyfish-13B and Jellyfish-Interpreter, the few-shot approach is disabled on seen datasets and enabled on unseen datasets._
70
  _Accuracy as the metric for data imputation and the F1 score for other tasks._
71
 
72
+ 1.
73
+ [Ditto](https://arxiv.org/abs/2004.00584) for Entity Matching
74
+ [SMAT](https://www.researchgate.net/publication/353920530_SMAT_An_Attention-Based_Deep_Learning_Solution_to_the_Automation_of_Schema_Matching) for Schema Matching
75
+ [HoloDetect](https://arxiv.org/abs/1904.02285) for Error Detection seen datasets
76
+ [RAHA](https://dl.acm.org/doi/10.1145/3299869.3324956) for Error Detection unseen datasets
77
+ [IPM](https://ieeexplore.ieee.org/document/9458712) for Data Imputation
78
+ 2.
79
+ [Large Language Models as Data Preprocessors](https://arxiv.org/abs/2308.16361)
80
+
81
  ## Performance on unseen tasks
82
 
83
  ### Column Type Annotation
84
 
85
+ | Dataset | RoBERTa (159 shots)<sup>1</sup> | GPT-3.5<sup>1</sup> | GPT-4 | GPT-4o | Jellyfish-7B | Jellyfish-8B | Jellyfish-13B |
86
+ |--------|-----------------|--------|--------|--------|--------------|--------------|---------------|
87
+ | SOTAB | 79.20 | 89.47 | 91.55 | 65.05 | 83 | 76.33 | 82 |
88
 
89
  _Few-shot is disabled for Jellyfish-13B._
90
 
 
97
  | AE-110k | 52.10 | 49.20 | 61.30 | 55.50 | 58.12 | 76.85| 69.78|
98
  | OA-Mine | 50.80 | 55.20 | 62.70 | 68.90 | 55.96 | 76.04| 78.83|
99
 
100
+ 1. Results from [Product Attribute Value Extraction using Large Language Models](https://arxiv.org/abs/2310.12537)
101
 
102
  ## Prompt Template
103
  ```