lex-eval 0.2.4 → 0.2.5

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
checksums.yaml CHANGED
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  SHA256:
3
- metadata.gz: e519977bbdf83bbb365455d368008b572ec68a46b756e9e33aed4376122549fd
4
- data.tar.gz: c91130345e012f2b585fa86b26f3578b0b111a2d95a76e21ec3c1316065a830c
3
+ metadata.gz: eb201f45138a8db5683e604f829ec83cf83f8dfbb3418d244c77abe84b1ea4b3
4
+ data.tar.gz: a2f596feed36febd77d34a461c6cea22e505c4fcb036eee0fa2c9890c8f40045
5
5
  SHA512:
6
- metadata.gz: 749867fd2b8004bbc96cf29192ad9d33302677e31019eb4c2b8c315320544b0b407b4272edd4a159effeec77c505f3e8fd7d94f4088b6b86695a2bf35d410f86
7
- data.tar.gz: d0579690cba79a88f1b88bee339a571933cc4aa053ca6222ef9f7285ca7cf3b871e9d40c5dec2e56e48b46ef5a04edf4fb412cddc59405f6fb8cf06759065b49
6
+ metadata.gz: 7464a6589b7fe9766218a7b0a7813c71e9771a4b8615a7f4d84e883c2540891849246562e57cdbab714df0bf1b1643fc0db3c5dbbbc933358ce92b9a95e4204c
7
+ data.tar.gz: d384db612df500e5bd54394f53daedc0114e12451a60b9e601910eee50ab25c8b2ede86f58b6c6a90ab737ce4be7a508e539ff467d2b55826ada9dd38ce41fcf
@@ -42,8 +42,29 @@ module Legion
42
42
  review.merge(action: action, escalated: true, priority: priority)
43
43
  end
44
44
 
45
- def review_experiment(**)
46
- { reviewed: false, reason: 'not_yet_implemented' }
45
+ def review_experiment(input:, output_a:, output_b:, review_prompt: nil, **)
46
+ review_a = review_output(input: input, output: output_a, review_prompt: review_prompt)
47
+ review_b = review_output(input: input, output: output_b, review_prompt: review_prompt)
48
+
49
+ conf_a = review_a[:confidence] || 0.0
50
+ conf_b = review_b[:confidence] || 0.0
51
+ delta = (conf_a - conf_b).round(3)
52
+
53
+ winner = if delta.abs < 0.05
54
+ :tie
55
+ elsif conf_a > conf_b
56
+ :a
57
+ else
58
+ :b
59
+ end
60
+
61
+ { reviewed: true,
62
+ winner: winner,
63
+ delta: delta,
64
+ review_a: review_a,
65
+ review_b: review_b }
66
+ rescue StandardError => e
67
+ { reviewed: false, reason: "experiment error: #{e.message}" }
47
68
  end
48
69
 
49
70
  private
@@ -3,7 +3,7 @@
3
3
  module Legion
4
4
  module Extensions
5
5
  module Eval
6
- VERSION = '0.2.4'
6
+ VERSION = '0.2.5'
7
7
  end
8
8
  end
9
9
  end
metadata CHANGED
@@ -1,7 +1,7 @@
1
1
  --- !ruby/object:Gem::Specification
2
2
  name: lex-eval
3
3
  version: !ruby/object:Gem::Version
4
- version: 0.2.4
4
+ version: 0.2.5
5
5
  platform: ruby
6
6
  authors:
7
7
  - Matthew Iverson