winglian commited on
Commit
1daad72
1 Parent(s): 2c21c62

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +49 -0
README.md CHANGED
@@ -27,3 +27,52 @@ DPOpenHermes was trained on a single H100 80GB hosted on RunPod for ~10h for 0.6
27
 
28
  https://wandb.ai/oaaic/openhermes-dpo/reports/DPOpenHermes--Vmlldzo2MTQ3NDg2
29
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
27
 
28
  https://wandb.ai/oaaic/openhermes-dpo/reports/DPOpenHermes--Vmlldzo2MTQ3NDg2
29
 
30
+ # Benchmarks
31
+
32
+ ## AGIEval
33
+
34
+ ```
35
+ | Task |Version| Metric |Value | |Stderr|
36
+ |------------------------------|------:|--------|-----:|---|-----:|
37
+ |agieval_aqua_rat | 0|acc |0.2480|_ |0.0272|
38
+ | | |acc_norm|0.2520|_ |0.0273|
39
+ |agieval_logiqa_en | 0|acc |0.3810|_ |0.0190|
40
+ | | |acc_norm|0.3856|_ |0.0191|
41
+ |agieval_lsat_ar | 0|acc |0.2348|_ |0.0280|
42
+ | | |acc_norm|0.2304|_ |0.0278|
43
+ |agieval_lsat_lr | 0|acc |0.5118|_ |0.0222|
44
+ | | |acc_norm|0.5196|_ |0.0221|
45
+ |agieval_lsat_rc | 0|acc |0.5948|_ |0.0300|
46
+ | | |acc_norm|0.5688|_ |0.0303|
47
+ |agieval_sat_en | 0|acc |0.7427|_ |0.0305|
48
+ | | |acc_norm|0.7427|_ |0.0305|
49
+ |agieval_sat_en_without_passage| 0|acc |0.4563|_ |0.0348|
50
+ | | |acc_norm|0.4515|_ |0.0348|
51
+ |agieval_sat_math | 0|acc |0.3818|_ |0.0328|
52
+ | | |acc_norm|0.3682|_ |0.0326|
53
+ ```
54
+
55
+ Average: 0.4399
56
+
57
+ ## GPT4All
58
+
59
+ ```
60
+ | Task |Version| Metric |Value | |Stderr|
61
+ |-------------|------:|--------|-----:|---|-----:|
62
+ |arc_challenge| 0|acc |0.5930|_ |0.0144|
63
+ | | |acc_norm|0.6323|_ |0.0141|
64
+ |arc_easy | 0|acc |0.8443|_ |0.0074|
65
+ | | |acc_norm|0.8295|_ |0.0077|
66
+ |boolq | 1|acc |0.8599|_ |0.0061|
67
+ |hellaswag | 0|acc |0.6548|_ |0.0047|
68
+ | | |acc_norm|0.8365|_ |0.0037|
69
+ |openbookqa | 0|acc |0.3520|_ |0.0214|
70
+ | | |acc_norm|0.4640|_ |0.0223|
71
+ |piqa | 0|acc |0.8210|_ |0.0089|
72
+ | | |acc_norm|0.8335|_ |0.0087|
73
+ |winogrande | 0|acc |0.7466|_ |0.0122|
74
+ ```
75
+
76
+ Average: 0.7431
77
+
78
+