Commit
•
3b1f4ad
1
Parent(s):
2b4b041
added model comparisons
Browse files
README.md
CHANGED
|
@@ -68,4 +68,90 @@ pipeline = transformers.pipeline(
|
|
| 68 |
|
| 69 |
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
|
| 70 |
print(outputs[0]["generated_text"])
|
| 71 |
-
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
|
| 69 |
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
|
| 70 |
print(outputs[0]["generated_text"])
|
| 71 |
+
```
|
| 72 |
+
|
| 73 |
+
## Benchmark Results
|
| 74 |
+
|
| 75 |
+
Chat2Eco is a major improvement across the board on the benchmarks below compared to the base model, and is the first model to beat the all good benchmarks.
|
| 76 |
+
|
| 77 |
+
## GPT4All:
|
| 78 |
+
```
|
| 79 |
+
| Task |Version| Metric |Value | |Stderr|
|
| 80 |
+
|-------------|------:|--------|-----:|---|-----:|
|
| 81 |
+
|arc_challenge| 0|acc |0.5990|± |0.0143|
|
| 82 |
+
| | |acc_norm|0.6425|± |0.0140|
|
| 83 |
+
|arc_easy | 0|acc |0.8657|± |0.0070|
|
| 84 |
+
| | |acc_norm|0.8636|± |0.0070|
|
| 85 |
+
|boolq | 1|acc |0.8783|± |0.0057|
|
| 86 |
+
|hellaswag | 0|acc |0.6661|± |0.0047|
|
| 87 |
+
| | |acc_norm|0.8489|± |0.0036|
|
| 88 |
+
|openbookqa | 0|acc |0.3440|± |0.0213|
|
| 89 |
+
| | |acc_norm|0.4660|± |0.0223|
|
| 90 |
+
|piqa | 0|acc |0.8324|± |0.0087|
|
| 91 |
+
| | |acc_norm|0.8379|± |0.0086|
|
| 92 |
+
|winogrande | 0|acc |0.7616|± |0.0120|
|
| 93 |
+
```
|
| 94 |
+
Average: 87.25
|
| 95 |
+
|
| 96 |
+
## AGIEval:
|
| 97 |
+
```
|
| 98 |
+
| Task |Version| Metric |Value | |Stderr|
|
| 99 |
+
|------------------------------|------:|--------|-----:|---|-----:|
|
| 100 |
+
|agieval_aqua_rat | 0|acc |0.2402|± |0.0269|
|
| 101 |
+
| | |acc_norm|0.2520|± |0.0273|
|
| 102 |
+
|agieval_logiqa_en | 0|acc |0.4117|± |0.0193|
|
| 103 |
+
| | |acc_norm|0.4055|± |0.0193|
|
| 104 |
+
|agieval_lsat_ar | 0|acc |0.2348|± |0.0280|
|
| 105 |
+
| | |acc_norm|0.2087|± |0.0269|
|
| 106 |
+
|agieval_lsat_lr | 0|acc |0.5549|± |0.0220|
|
| 107 |
+
| | |acc_norm|0.5294|± |0.0221|
|
| 108 |
+
|agieval_lsat_rc | 0|acc |0.6617|± |0.0289|
|
| 109 |
+
| | |acc_norm|0.6357|± |0.0294|
|
| 110 |
+
|agieval_sat_en | 0|acc |0.8010|± |0.0279|
|
| 111 |
+
| | |acc_norm|0.7913|± |0.0284|
|
| 112 |
+
|agieval_sat_en_without_passage| 0|acc |0.4806|± |0.0349|
|
| 113 |
+
| | |acc_norm|0.4612|± |0.0348|
|
| 114 |
+
|agieval_sat_math | 0|acc |0.4909|± |0.0338|
|
| 115 |
+
| | |acc_norm|0.4000|± |0.0331|
|
| 116 |
+
```
|
| 117 |
+
Average: 89.05
|
| 118 |
+
|
| 119 |
+
## BigBench:
|
| 120 |
+
```
|
| 121 |
+
| Task |Version| Metric |Value | |Stderr|
|
| 122 |
+
|------------------------------------------------|------:|---------------------|-----:|---|-----:|
|
| 123 |
+
|bigbench_causal_judgement | 0|multiple_choice_grade|0.6105|± |0.0355|
|
| 124 |
+
|bigbench_date_understanding | 0|multiple_choice_grade|0.7182|± |0.0235|
|
| 125 |
+
|bigbench_disambiguation_qa | 0|multiple_choice_grade|0.5736|± |0.0308|
|
| 126 |
+
|bigbench_geometric_shapes | 0|multiple_choice_grade|0.4596|± |0.0263|
|
| 127 |
+
| | |exact_str_match |0.0000|± |0.0000|
|
| 128 |
+
|bigbench_logical_deduction_five_objects | 0|multiple_choice_grade|0.3500|± |0.0214|
|
| 129 |
+
|bigbench_logical_deduction_seven_objects | 0|multiple_choice_grade|0.2500|± |0.0164|
|
| 130 |
+
|bigbench_logical_deduction_three_objects | 0|multiple_choice_grade|0.5200|± |0.0289|
|
| 131 |
+
|bigbench_movie_recommendation | 0|multiple_choice_grade|0.3540|± |0.0214|
|
| 132 |
+
|bigbench_navigate | 0|multiple_choice_grade|0.5000|± |0.0158|
|
| 133 |
+
|bigbench_reasoning_about_colored_objects | 0|multiple_choice_grade|0.6900|± |0.0103|
|
| 134 |
+
|bigbench_ruin_names | 0|multiple_choice_grade|0.6317|± |0.0228|
|
| 135 |
+
|bigbench_salient_translation_error_detection | 0|multiple_choice_grade|0.2535|± |0.0138|
|
| 136 |
+
|bigbench_snarks | 0|multiple_choice_grade|0.7293|± |0.0331|
|
| 137 |
+
|bigbench_sports_understanding | 0|multiple_choice_grade|0.6744|± |0.0149|
|
| 138 |
+
|bigbench_temporal_sequences | 0|multiple_choice_grade|0.7400|± |0.0139|
|
| 139 |
+
|bigbench_tracking_shuffled_objects_five_objects | 0|multiple_choice_grade|0.2176|± |0.0117|
|
| 140 |
+
|bigbench_tracking_shuffled_objects_seven_objects| 0|multiple_choice_grade|0.1543|± |0.0086|
|
| 141 |
+
|bigbench_tracking_shuffled_objects_three_objects| 0|multiple_choice_grade|0.5200|± |0.0289|
|
| 142 |
+
```
|
| 143 |
+
Average: 87.45
|
| 144 |
+
|
| 145 |
+
# Benchmark Comparison Charts
|
| 146 |
+
|
| 147 |
+
## GPT4All
|
| 148 |
+
|
| 149 |
+
|
| 150 |
+
|
| 151 |
+
## AGI-Eval
|
| 152 |
+
|
| 153 |
+
|
| 154 |
+
|
| 155 |
+
## BigBench Reasoning Test
|
| 156 |
+
|
| 157 |
+
|