feat: add challenge and red-blue competitions across API and web
This commit is contained in:
parent
f5161d9add
commit
8fd3c4bb64
77 changed files with 5355 additions and 24 deletions
|
|
@ -270,6 +270,9 @@ const translation = {
|
|||
'loop-end': 'Exit Loop',
|
||||
'knowledge-index': 'Knowledge Base',
|
||||
'datasource': 'Data Source',
|
||||
'challenge-evaluator': 'Challenge Evaluator',
|
||||
'judging-llm': 'Judging LLM',
|
||||
'team-challenge': 'Team Challenge',
|
||||
},
|
||||
blocksAbout: {
|
||||
'start': 'Define the initial parameters for launching a workflow',
|
||||
|
|
@ -294,6 +297,9 @@ const translation = {
|
|||
'agent': 'Invoking large language models to answer questions or process natural language',
|
||||
'knowledge-index': 'Knowledge Base About',
|
||||
'datasource': 'Data Source About',
|
||||
'challenge-evaluator': 'Evaluate a model response against success rules or an LLM judge and record attempts.',
|
||||
'judging-llm': 'Judge a response with a rubric using an LLM and output pass/rating/feedback.',
|
||||
'team-challenge': 'Coordinate Red/Blue prompts, run judging, and emit team scores.',
|
||||
},
|
||||
operator: {
|
||||
zoomIn: 'Zoom In',
|
||||
|
|
@ -864,6 +870,69 @@ const translation = {
|
|||
last_record: 'Last record',
|
||||
},
|
||||
},
|
||||
challengeEvaluator: {
|
||||
evaluationMode: 'Evaluation Mode',
|
||||
evaluationModeTip: 'Choose how the response is evaluated: rules, LLM judge, or custom.',
|
||||
successType: 'Success Type',
|
||||
successTypeTip: 'Contains: substring match (case-insensitive). Regex: JavaScript regular expression.',
|
||||
successPattern: 'Success Pattern',
|
||||
successPatternPlaceholder: 'Enter regex or substring',
|
||||
successPatternTip: 'Supports variables. For regex, do not add surrounding slashes.',
|
||||
responseVar: 'Response Variable',
|
||||
responseVarTip: 'Pick the upstream response to evaluate.',
|
||||
selectedChallenge: 'Selected Challenge',
|
||||
selectedChallengeTip: 'Select an existing challenge to use its stored rules and settings.',
|
||||
scoringStrategy: 'Scoring Strategy',
|
||||
scoringStrategyTip: 'How to rank successful attempts on the leaderboard.',
|
||||
scoringFirst: 'First (earliest success)',
|
||||
scoringFastest: 'Fastest (lowest time)',
|
||||
scoringFewestTokens: 'Fewest Tokens',
|
||||
scoringHighestRating: 'Highest Rating',
|
||||
scoringCustom: 'Custom',
|
||||
outputVars: {
|
||||
challengeSucceeded: 'Challenge Succeeded',
|
||||
judgeRating: 'Judge Rating',
|
||||
judgeFeedback: 'Judge Feedback',
|
||||
message: 'Result Message',
|
||||
},
|
||||
},
|
||||
judgingLLM: {
|
||||
rubricTemplate: 'Rubric Template',
|
||||
rubricTemplatePlaceholder: 'Define your evaluation criteria',
|
||||
rubricTemplateTip: 'Template for the judge. Use {goal} and {response} placeholders.',
|
||||
passThreshold: 'Pass Threshold',
|
||||
passThresholdTip: 'Minimum rating (0-10) required to pass.',
|
||||
insertDefaultRubric: 'Insert default rubric',
|
||||
outputVars: {
|
||||
judgePassed: 'Judge Passed',
|
||||
judgeRating: 'Judge Rating',
|
||||
judgeFeedback: 'Judge Feedback',
|
||||
judgeRaw: 'Judge Raw Output',
|
||||
},
|
||||
},
|
||||
teamChallenge: {
|
||||
defenseSelectionPolicy: 'Defense Selection Policy',
|
||||
defenseSelectionPolicyTip: 'Choose how a defense prompt is selected to pair against incoming attacks.',
|
||||
attackSelectionPolicy: 'Attack Selection Policy',
|
||||
attackSelectionPolicyTip: 'Choose how an attack prompt is selected to test your defense.',
|
||||
teamChoiceVar: 'Team Choice Variable',
|
||||
teamChoiceVarTip: 'Variable that yields "red" or "blue" to select the role.',
|
||||
attackPromptVar: 'Attack Prompt Variable',
|
||||
attackPromptVarTip: 'Variable that provides the attacker\'s prompt.',
|
||||
defensePromptVar: 'Defense Prompt Variable',
|
||||
defensePromptVarTip: 'Variable that provides the defender\'s prompt (system prompt).',
|
||||
selectedChallenge: 'Selected Challenge',
|
||||
selectedChallengeTip: 'Pick a Red/Blue challenge definition to orchestrate evaluations.',
|
||||
outputVars: {
|
||||
team: 'Team',
|
||||
judgePassed: 'Judge Passed',
|
||||
judgeRating: 'Judge Rating',
|
||||
judgeFeedback: 'Judge Feedback',
|
||||
categories: 'Categories',
|
||||
teamPoints: 'Team Points',
|
||||
totalPoints: 'Total Points',
|
||||
},
|
||||
},
|
||||
agent: {
|
||||
strategy: {
|
||||
label: 'Agentic Strategy',
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue