sudeepdasgupta commited on
Commit
164b2f2
·
verified ·
1 Parent(s): b1cebfc

v3.0.0: all-domain distillation arm (distill4/ck54) replaces v2.0.0 (ck2079); bfcl_all 0.829 [0.802,0.853] vs 0.556 [0.522,0.589] n=840, same session; irrelevance 0.930/0.790 vs 0.010/0.050

Browse files
Files changed (1) hide show
  1. benchmarks/v3.0.0_bfcl_v4.json +111 -0
benchmarks/v3.0.0_bfcl_v4.json ADDED
@@ -0,0 +1,111 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "tag": "astraforge-distill4_ck54_1014T0517",
3
+ "bfcl_irrelevance": {
4
+ "n": 100,
5
+ "acc": 0.93,
6
+ "ci95": [
7
+ 0.863,
8
+ 0.966
9
+ ]
10
+ },
11
+ "bfcl_live_irrelevance": {
12
+ "n": 100,
13
+ "acc": 0.79,
14
+ "ci95": [
15
+ 0.7,
16
+ 0.858
17
+ ]
18
+ },
19
+ "bfcl_live_multiple": {
20
+ "n": 100,
21
+ "acc": 0.75,
22
+ "ci95": [
23
+ 0.657,
24
+ 0.825
25
+ ]
26
+ },
27
+ "bfcl_live_parallel": {
28
+ "n": 16,
29
+ "acc": 0.688,
30
+ "ci95": [
31
+ 0.444,
32
+ 0.858
33
+ ]
34
+ },
35
+ "bfcl_live_parallel_multiple": {
36
+ "n": 24,
37
+ "acc": 0.625,
38
+ "ci95": [
39
+ 0.427,
40
+ 0.788
41
+ ]
42
+ },
43
+ "bfcl_live_simple": {
44
+ "n": 100,
45
+ "acc": 0.69,
46
+ "ci95": [
47
+ 0.594,
48
+ 0.772
49
+ ]
50
+ },
51
+ "bfcl_multiple": {
52
+ "n": 100,
53
+ "acc": 0.9,
54
+ "ci95": [
55
+ 0.826,
56
+ 0.945
57
+ ]
58
+ },
59
+ "bfcl_parallel": {
60
+ "n": 100,
61
+ "acc": 0.85,
62
+ "ci95": [
63
+ 0.767,
64
+ 0.907
65
+ ]
66
+ },
67
+ "bfcl_parallel_multiple": {
68
+ "n": 100,
69
+ "acc": 0.85,
70
+ "ci95": [
71
+ 0.767,
72
+ 0.907
73
+ ]
74
+ },
75
+ "bfcl_simple": {
76
+ "n": 100,
77
+ "acc": 0.94,
78
+ "ci95": [
79
+ 0.875,
80
+ 0.972
81
+ ]
82
+ },
83
+ "multi_domain": {
84
+ "n": 100,
85
+ "acc": 0.73,
86
+ "ci95": [
87
+ 0.636,
88
+ 0.807
89
+ ],
90
+ "tool_recall": 0.878,
91
+ "names_ok": 0.83
92
+ },
93
+ "unseen_family": {
94
+ "n": 100,
95
+ "acc": 1.0,
96
+ "ci95": [
97
+ 0.963,
98
+ 1.0
99
+ ],
100
+ "tool_recall": 1.0,
101
+ "names_ok": 1.0
102
+ },
103
+ "bfcl_all": {
104
+ "n": 840,
105
+ "acc": 0.829,
106
+ "ci95": [
107
+ 0.802,
108
+ 0.853
109
+ ]
110
+ }
111
+ }