ncoop57 commited on
Commit
07730be
1 Parent(s): 741227c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +4 -73
README.md CHANGED
@@ -114,9 +114,8 @@ The dataset is comprised of a mixture of open datasets large-scale datasets avai
114
  | Claude 2 | - |RLHF |8.06| 91.36|
115
  | GPT-4 | -| RLHF |8.99| 95.28|
116
 
117
- ## Other benchmark:
118
- 1. **HuggingFace OpenLLM Leaderboard**
119
- | Metric | Value |
120
  |-----------------------|---------------------------|
121
  | ARC (25-shot) | 47.0 |
122
  | HellaSwag (10-shot) | 74.2 |
@@ -124,76 +123,8 @@ The dataset is comprised of a mixture of open datasets large-scale datasets avai
124
  | TruthfulQA (0-shot) | 46.5 |
125
  | Winogrande (5-shot) | 65.5 |
126
  | GSM8K (5-shot) | 42.3 |
127
-
128
-
129
- 2. **BigBench**:
130
-
131
- - Average: 35.26
132
- - Details:
133
-
134
- | Task | Version | Metric | Value | Stderr |
135
- |-----------------------------------------------------|---------|-------------------------|-------|--------|
136
- | bigbench_causal_judgement | 0 | multiple_choice_grade | 0.5316| 0.0363 |
137
- | bigbench_date_understanding | 0 | multiple_choice_grade | 0.4363| 0.0259 |
138
- | bigbench_disambiguation_qa | 0 | multiple_choice_grade | 0.3217| 0.0291 |
139
- | bigbench_dyck_languages | 0 | multiple_choice_grade | 0.1450| 0.0111 |
140
- | bigbench_formal_fallacies_syllogisms_negation | 0 | multiple_choice_grade | 0.4982| 0.0042 |
141
- | bigbench_geometric_shapes | 0 | multiple_choice_grade | 0.1086| 0.0164 |
142
- | bigbench_hyperbaton | 0 | exact_str_match | 0.0000| 0.0000 |
143
- | bigbench_logical_deduction_five_objects | 0 | multiple_choice_grade | 0.5232| 0.0022 |
144
- | bigbench_logical_deduction_seven_objects | 0 | multiple_choice_grade | 0.2480| 0.0193 |
145
- | bigbench_logical_deduction_three_objects | 0 | multiple_choice_grade | 0.1814| 0.0146 |
146
- | bigbench_movie_recommendation | 0 | multiple_choice_grade | 0.4067| 0.0284 |
147
- | bigbench_navigate | 0 | multiple_choice_grade | 0.2580| 0.0196 |
148
- | bigbench_reasoning_about_colored_objects | 0 | multiple_choice_grade | 0.5990| 0.0155 |
149
- | bigbench_ruin_names | 0 | multiple_choice_grade | 0.4370| 0.0111 |
150
- | bigbench_salient_translation_error_detection | 0 | multiple_choice_grade | 0.3951| 0.0231 |
151
- | bigbench_snarks | 0 | multiple_choice_grade | 0.2265| 0.0133 |
152
- | bigbench_sports_understanding | 0 | multiple_choice_grade | 0.6464| 0.0356 |
153
- | bigbench_temporal_sequences | 0 | multiple_choice_grade | 0.5091| 0.0159 |
154
- | bigbench_tracking_shuffled_objects_five_objects | 0 | multiple_choice_grade | 0.2680| 0.0140 |
155
- | bigbench_tracking_shuffled_objects_seven_objects | 0 | multiple_choice_grade | 0.1856| 0.0110 |
156
- | bigbench_tracking_shuffled_objects_three_objects | 0 | multiple_choice_grade | 0.1269| 0.0080 |
157
-
158
- 3. **AGI Benchmark**:
159
- - Average: 33.23
160
- - Details:
161
- | Task |Version| Metric |Value | |Stderr|
162
- |------------------------------|------:|--------|-----:|---|-----:|
163
- |agieval_aqua_rat | 0|acc |0.2126|± |0.0257|
164
- | | |acc_norm|0.1890|± |0.0246|
165
- |agieval_gaokao_biology | 0|acc |0.2571|± |0.0302|
166
- | | |acc_norm|0.3143|± |0.0321|
167
- |agieval_gaokao_chemistry | 0|acc |0.2464|± |0.0300|
168
- | | |acc_norm|0.2899|± |0.0316|
169
- |agieval_gaokao_chinese | 0|acc |0.2927|± |0.0291|
170
- | | |acc_norm|0.3049|± |0.0294|
171
- |agieval_gaokao_english | 0|acc |0.6176|± |0.0278|
172
- | | |acc_norm|0.6438|± |0.0274|
173
- |agieval_gaokao_geography | 0|acc |0.3015|± |0.0326|
174
- | | |acc_norm|0.3065|± |0.0328|
175
- |agieval_gaokao_history | 0|acc |0.3106|± |0.0303|
176
- | | |acc_norm|0.3319|± |0.0308|
177
- |agieval_gaokao_mathqa | 0|acc |0.2650|± |0.0236|
178
- | | |acc_norm|0.2707|± |0.0237|
179
- |agieval_gaokao_physics | 0|acc |0.3450|± |0.0337|
180
- | | |acc_norm|0.3550|± |0.0339|
181
- |agieval_logiqa_en | 0|acc |0.2980|± |0.0179|
182
- | | |acc_norm|0.3195|± |0.0183|
183
- |agieval_logiqa_zh | 0|acc |0.2842|± |0.0177|
184
- | | |acc_norm|0.3318|± |0.0185|
185
- |agieval_lsat_ar | 0|acc |0.2000|± |0.0264|
186
- | | |acc_norm|0.2043|± |0.0266|
187
- |agieval_lsat_lr | 0|acc |0.3176|± |0.0206|
188
- | | |acc_norm|0.3275|± |0.0208|
189
- |agieval_lsat_rc | 0|acc |0.4312|± |0.0303|
190
- | | |acc_norm|0.4201|± |0.0301|
191
- |agieval_sat_en | 0|acc |0.6117|± |0.0340|
192
- | | |acc_norm|0.6117|± |0.0340|
193
- |agieval_sat_en_without_passage| 0|acc |0.3398|± |0.0331|
194
- | | |acc_norm|0.3495|± |0.0333|
195
- |agieval_sat_math | 0|acc |0.3182|± |0.0315|
196
- | | |acc_norm|0.2909|± |0.0307|
197
 
198
  ### Training Infrastructure
199
 
 
114
  | Claude 2 | - |RLHF |8.06| 91.36|
115
  | GPT-4 | -| RLHF |8.99| 95.28|
116
 
117
+ ## Other benchmarks:
118
+ | Task | Value |
 
119
  |-----------------------|---------------------------|
120
  | ARC (25-shot) | 47.0 |
121
  | HellaSwag (10-shot) | 74.2 |
 
123
  | TruthfulQA (0-shot) | 46.5 |
124
  | Winogrande (5-shot) | 65.5 |
125
  | GSM8K (5-shot) | 42.3 |
126
+ | BigBench (Avg) | 35.26 |
127
+ | AGI Benchmark (Avg) | 33.23 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
128
 
129
  ### Training Infrastructure
130