DPO fine tunes
Collection
3 items
โข
Updated
This model is a finetune of upstage/SOLAR-10.7B-Instruct-v1.0 using Intel/orca_dpo_pairs
This model follows the chatML chat template.
These scores are the average of 3 iterations.
as compared to the original model:
Model | AGIEval | GPT4All | TruthfulQA | Bigbench | Average |
---|---|---|---|---|---|
SOLAR-10.7b-Instruct-dpo | 47.57 | 74.3 | 72.73 | 45.76 | 60.09 |
Task | Version | Metric | Value | Stderr | |
---|---|---|---|---|---|
agieval_aqua_rat | 0 | acc | 27.56 | ยฑ | 2.81 |
acc_norm | 26.77 | ยฑ | 2.78 | ||
agieval_logiqa_en | 0 | acc | 41.63 | ยฑ | 1.93 |
acc_norm | 41.32 | ยฑ | 1.93 | ||
agieval_lsat_ar | 0 | acc | 25.22 | ยฑ | 2.87 |
acc_norm | 24.35 | ยฑ | 2.84 | ||
agieval_lsat_lr | 0 | acc | 54.12 | ยฑ | 2.21 |
acc_norm | 54.31 | ยฑ | 2.21 | ||
agieval_lsat_rc | 0 | acc | 68.77 | ยฑ | 2.83 |
acc_norm | 69.14 | ยฑ | 2.82 | ||
agieval_sat_en | 0 | acc | 79.13 | ยฑ | 2.84 |
acc_norm | 79.13 | ยฑ | 2.84 | ||
agieval_sat_en_without_passage | 0 | acc | 44.66 | ยฑ | 3.47 |
acc_norm | 44.66 | ยฑ | 3.47 | ||
agieval_sat_math | 0 | acc | 40.45 | ยฑ | 3.32 |
acc_norm | 40.91 | ยฑ | 3.32 |
Average: 47.57%
Task | Version | Metric | Value | Stderr | |
---|---|---|---|---|---|
arc_challenge | 0 | acc | 60.49 | ยฑ | 1.43 |
acc_norm | 63.74 | ยฑ | 1.40 | ||
arc_easy | 0 | acc | 82.07 | ยฑ | 0.79 |
acc_norm | 79.92 | ยฑ | 0.82 | ||
boolq | 1 | acc | 88.56 | ยฑ | 0.56 |
hellaswag | 0 | acc | 68.47 | ยฑ | 0.46 |
acc_norm | 86.06 | ยฑ | 0.35 | ||
openbookqa | 0 | acc | 36.20 | ยฑ | 2.15 |
acc_norm | 46.60 | ยฑ | 2.23 | ||
piqa | 0 | acc | 79.38 | ยฑ | 0.94 |
acc_norm | 79.71 | ยฑ | 0.94 | ||
winogrande | 0 | acc | 75.53 | ยฑ | 1.21 |
Average: 74.3%
Task | Version | Metric | Value | Stderr | |
---|---|---|---|---|---|
truthfulqa_mc | 1 | mc1 | 57.77 | ยฑ | 1.73 |
mc2 | 72.73 | ยฑ | 1.49 |
Average: 72.73%
Task | Version | Metric | Value | Stderr | |
---|---|---|---|---|---|
bigbench_causal_judgement | 0 | multiple_choice_grade | 55.26 | ยฑ | 3.62 |
bigbench_date_understanding | 0 | multiple_choice_grade | 62.87 | ยฑ | 2.52 |
bigbench_disambiguation_qa | 0 | multiple_choice_grade | 46.51 | ยฑ | 3.11 |
bigbench_geometric_shapes | 0 | multiple_choice_grade | 25.63 | ยฑ | 2.31 |
exact_str_match | 0.00 | ยฑ | 0.00 | ||
bigbench_logical_deduction_five_objects | 0 | multiple_choice_grade | 28.00 | ยฑ | 2.01 |
bigbench_logical_deduction_seven_objects | 0 | multiple_choice_grade | 20.57 | ยฑ | 1.53 |
bigbench_logical_deduction_three_objects | 0 | multiple_choice_grade | 46.67 | ยฑ | 2.89 |
bigbench_movie_recommendation | 0 | multiple_choice_grade | 41.80 | ยฑ | 2.21 |
bigbench_navigate | 0 | multiple_choice_grade | 64.00 | ยฑ | 1.52 |
bigbench_reasoning_about_colored_objects | 0 | multiple_choice_grade | 60.00 | ยฑ | 1.10 |
bigbench_ruin_names | 0 | multiple_choice_grade | 39.96 | ยฑ | 2.32 |
bigbench_salient_translation_error_detection | 0 | multiple_choice_grade | 47.90 | ยฑ | 1.58 |
bigbench_snarks | 0 | multiple_choice_grade | 64.09 | ยฑ | 3.58 |
bigbench_sports_understanding | 0 | multiple_choice_grade | 71.10 | ยฑ | 1.44 |
bigbench_temporal_sequences | 0 | multiple_choice_grade | 59.90 | ยฑ | 1.55 |
bigbench_tracking_shuffled_objects_five_objects | 0 | multiple_choice_grade | 24.96 | ยฑ | 1.22 |
bigbench_tracking_shuffled_objects_seven_objects | 0 | multiple_choice_grade | 17.89 | ยฑ | 0.92 |
bigbench_tracking_shuffled_objects_three_objects | 0 | multiple_choice_grade | 46.67 | ยฑ | 2.89 |
Average: 45.76%
Average score: 60.09%
Elapsed time: 02:10:16
Detailed results can be found here
Metric | Value |
---|---|
Avg. | 73.54 |
AI2 Reasoning Challenge (25-Shot) | 71.76 |
HellaSwag (10-Shot) | 88.08 |
MMLU (5-Shot) | 66.06 |
TruthfulQA (0-shot) | 71.98 |
Winogrande (5-shot) | 82.32 |
GSM8k (5-shot) | 61.03 |