Documentation
¶
Overview ¶
Package metrics 提供高级评估指标
Package metrics 提供 AI 系统的评估指标 ¶
Package metrics 提供 AI 系统的评估指标 ¶
Package metrics 提供 NLP 评估指标 ¶
Package metrics 提供 AI 系统的评估指标 ¶
Package metrics 提供 AI 系统的评估指标
Index ¶
- type AnswerQualityEvaluator
- type AnswerQualityOption
- type BLEUEvaluator
- type BLEUOption
- type CoherenceEvaluator
- type CoherenceOption
- type CompletenessEvaluator
- type CompletenessOption
- type CompositePerformanceEvaluator
- type CompositePerformanceOption
- type ConsistencyEvaluator
- type ConsistencyOption
- type ContextRelevanceEvaluator
- type ContextRelevanceOption
- type CorrectnessEvaluator
- type CorrectnessMode
- type CorrectnessOption
- type CostEvaluator
- type CostOption
- type DiversityEvaluator
- type DiversityOption
- type F1Evaluator
- type F1Option
- type FaithfulnessEvaluator
- type FaithfulnessOption
- type HallucinationEvaluator
- type HallucinationOption
- type LatencyEvaluator
- type LatencyOption
- type PerplexityEvaluator
- type PerplexityOption
- type ROUGEEvaluator
- type ROUGEOption
- type RelevanceEvaluator
- type RelevanceOption
- type ThroughputEvaluator
- type ThroughputOption
- type ToxicityEvaluator
- type ToxicityOption
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type AnswerQualityEvaluator ¶
type AnswerQualityEvaluator struct {
// contains filtered or unexported fields
}
AnswerQualityEvaluator 评估回答的整体质量
func NewAnswerQualityEvaluator ¶
func NewAnswerQualityEvaluator(llm evaluate.LLMJudge, opts ...AnswerQualityOption) *AnswerQualityEvaluator
NewAnswerQualityEvaluator 创建回答质量评估器
func (*AnswerQualityEvaluator) Description ¶
func (e *AnswerQualityEvaluator) Description() string
Description 返回评估器描述
func (*AnswerQualityEvaluator) Evaluate ¶
func (e *AnswerQualityEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行质量评估
func (*AnswerQualityEvaluator) RequiresLLM ¶
func (e *AnswerQualityEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type AnswerQualityOption ¶
type AnswerQualityOption func(*AnswerQualityEvaluator)
AnswerQualityOption 选项
func WithAnswerQualityCriteria ¶
func WithAnswerQualityCriteria(criteria []string) AnswerQualityOption
WithAnswerQualityCriteria 设置评估标准
func WithAnswerQualityThreshold ¶
func WithAnswerQualityThreshold(threshold float64) AnswerQualityOption
WithAnswerQualityThreshold 设置阈值
type BLEUEvaluator ¶
type BLEUEvaluator struct {
// contains filtered or unexported fields
}
BLEUEvaluator BLEU (Bilingual Evaluation Understudy) 评估器
主要用于机器翻译质量评估，通过计算 n-gram 精确度来衡量 生成文本与参考文本的相似程度。
func NewBLEUEvaluator ¶
func NewBLEUEvaluator(opts ...BLEUOption) *BLEUEvaluator
NewBLEUEvaluator 创建 BLEU 评估器
func (*BLEUEvaluator) Description ¶
func (e *BLEUEvaluator) Description() string
Description 返回评估器描述
func (*BLEUEvaluator) Evaluate ¶
func (e *BLEUEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行 BLEU 评估
func (*BLEUEvaluator) RequiresLLM ¶
func (e *BLEUEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type BLEUOption ¶
type BLEUOption func(*BLEUEvaluator)
BLEUOption BLEU 评估器选项
func WithBLEUThreshold ¶
func WithBLEUThreshold(threshold float64) BLEUOption
WithBLEUThreshold 设置通过阈值
type CoherenceEvaluator ¶
type CoherenceEvaluator struct {
// contains filtered or unexported fields
}
CoherenceEvaluator 连贯性评估器
评估文本的逻辑连贯性和流畅度。
func NewCoherenceEvaluator ¶
func NewCoherenceEvaluator(llm evaluate.LLMJudge, opts ...CoherenceOption) *CoherenceEvaluator
NewCoherenceEvaluator 创建连贯性评估器
func (*CoherenceEvaluator) Description ¶
func (e *CoherenceEvaluator) Description() string
Description 返回评估器描述
func (*CoherenceEvaluator) Evaluate ¶
func (e *CoherenceEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行连贯性评估
func (*CoherenceEvaluator) RequiresLLM ¶
func (e *CoherenceEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type CoherenceOption ¶
type CoherenceOption func(*CoherenceEvaluator)
CoherenceOption 连贯性评估器选项
func WithCoherenceThreshold ¶
func WithCoherenceThreshold(threshold float64) CoherenceOption
WithCoherenceThreshold 设置通过阈值
type CompletenessEvaluator ¶
type CompletenessEvaluator struct {
// contains filtered or unexported fields
}
CompletenessEvaluator 完整性评估器
评估回答是否完整地回答了问题的所有方面。
func NewCompletenessEvaluator ¶
func NewCompletenessEvaluator(llm evaluate.LLMJudge, opts ...CompletenessOption) *CompletenessEvaluator
NewCompletenessEvaluator 创建完整性评估器
func (*CompletenessEvaluator) Description ¶
func (e *CompletenessEvaluator) Description() string
Description 返回评估器描述
func (*CompletenessEvaluator) Evaluate ¶
func (e *CompletenessEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行完整性评估
func (*CompletenessEvaluator) RequiresLLM ¶
func (e *CompletenessEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type CompletenessOption ¶
type CompletenessOption func(*CompletenessEvaluator)
CompletenessOption 完整性评估器选项
func WithCompletenessThreshold ¶
func WithCompletenessThreshold(threshold float64) CompletenessOption
WithCompletenessThreshold 设置通过阈值
type CompositePerformanceEvaluator ¶
type CompositePerformanceEvaluator struct {
// contains filtered or unexported fields
}
CompositePerformanceEvaluator 综合性能评估器 结合延迟、成本、吞吐量等指标
func NewCompositePerformanceEvaluator ¶
func NewCompositePerformanceEvaluator(opts ...CompositePerformanceOption) *CompositePerformanceEvaluator
NewCompositePerformanceEvaluator 创建综合性能评估器
func (*CompositePerformanceEvaluator) Description ¶
func (e *CompositePerformanceEvaluator) Description() string
Description 返回评估器描述
func (*CompositePerformanceEvaluator) Evaluate ¶
func (e *CompositePerformanceEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行综合性能评估
func (*CompositePerformanceEvaluator) Name ¶
func (e *CompositePerformanceEvaluator) Name() string
Name 返回评估器名称
func (*CompositePerformanceEvaluator) RequiresLLM ¶
func (e *CompositePerformanceEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type CompositePerformanceOption ¶
type CompositePerformanceOption func(*CompositePerformanceEvaluator)
CompositePerformanceOption 选项
func WithPerformanceWeights ¶
func WithPerformanceWeights(latency, cost, throughput float64) CompositePerformanceOption
WithPerformanceWeights 设置权重
type ConsistencyEvaluator ¶
type ConsistencyEvaluator struct {
// contains filtered or unexported fields
}
ConsistencyEvaluator 一致性评估器
评估多个回答之间的一致性，用于检测模型输出的稳定性。
func NewConsistencyEvaluator ¶
func NewConsistencyEvaluator(llm evaluate.LLMJudge, opts ...ConsistencyOption) *ConsistencyEvaluator
NewConsistencyEvaluator 创建一致性评估器
func (*ConsistencyEvaluator) Description ¶
func (e *ConsistencyEvaluator) Description() string
Description 返回评估器描述
func (*ConsistencyEvaluator) Evaluate ¶
func (e *ConsistencyEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行一致性评估
func (*ConsistencyEvaluator) RequiresLLM ¶
func (e *ConsistencyEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type ConsistencyOption ¶
type ConsistencyOption func(*ConsistencyEvaluator)
ConsistencyOption 一致性评估器选项
func WithConsistencyThreshold ¶
func WithConsistencyThreshold(threshold float64) ConsistencyOption
WithConsistencyThreshold 设置通过阈值
type ContextRelevanceEvaluator ¶
type ContextRelevanceEvaluator struct {
// contains filtered or unexported fields
}
ContextRelevanceEvaluator 评估上下文与回答的相关性
func NewContextRelevanceEvaluator ¶
func NewContextRelevanceEvaluator(llm evaluate.LLMJudge, opts ...ContextRelevanceOption) *ContextRelevanceEvaluator
NewContextRelevanceEvaluator 创建上下文相关性评估器
func (*ContextRelevanceEvaluator) Description ¶
func (e *ContextRelevanceEvaluator) Description() string
Description 返回评估器描述
func (*ContextRelevanceEvaluator) Evaluate ¶
func (e *ContextRelevanceEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行评估
func (*ContextRelevanceEvaluator) Name ¶
func (e *ContextRelevanceEvaluator) Name() string
Name 返回评估器名称
func (*ContextRelevanceEvaluator) RequiresLLM ¶
func (e *ContextRelevanceEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type ContextRelevanceOption ¶
type ContextRelevanceOption func(*ContextRelevanceEvaluator)
ContextRelevanceOption 选项
func WithContextRelevanceThreshold ¶
func WithContextRelevanceThreshold(threshold float64) ContextRelevanceOption
WithContextRelevanceThreshold 设置阈值
type CorrectnessEvaluator ¶
type CorrectnessEvaluator struct {
// contains filtered or unexported fields
}
CorrectnessEvaluator 评估回答的正确性 通过与参考答案对比来评估
func NewCorrectnessEvaluator ¶
func NewCorrectnessEvaluator(llm evaluate.LLMJudge, opts ...CorrectnessOption) *CorrectnessEvaluator
NewCorrectnessEvaluator 创建正确性评估器
func (*CorrectnessEvaluator) Description ¶
func (e *CorrectnessEvaluator) Description() string
Description 返回评估器描述
func (*CorrectnessEvaluator) Evaluate ¶
func (e *CorrectnessEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行正确性评估
func (*CorrectnessEvaluator) RequiresLLM ¶
func (e *CorrectnessEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type CorrectnessMode ¶
type CorrectnessMode string
CorrectnessMode 正确性评估模式
const ( // CorrectnessModeExact 精确匹配 CorrectnessModeExact CorrectnessMode = "exact" // CorrectnessModeSemanticSimilarity 语义相似度 CorrectnessModeSemanticSimilarity CorrectnessMode = "semantic" // CorrectnessModeFactual 事实正确性 CorrectnessModeFactual CorrectnessMode = "factual" )
type CorrectnessOption ¶
type CorrectnessOption func(*CorrectnessEvaluator)
CorrectnessOption CorrectnessEvaluator 选项
func WithCorrectnessMode ¶
func WithCorrectnessMode(mode CorrectnessMode) CorrectnessOption
WithCorrectnessMode 设置评估模式
func WithCorrectnessThreshold ¶
func WithCorrectnessThreshold(threshold float64) CorrectnessOption
WithCorrectnessThreshold 设置通过阈值
type CostEvaluator ¶
type CostEvaluator struct {
// contains filtered or unexported fields
}
CostEvaluator 评估系统成本
func NewCostEvaluator ¶
func NewCostEvaluator(opts ...CostOption) *CostEvaluator
NewCostEvaluator 创建成本评估器
func (*CostEvaluator) Description ¶
func (e *CostEvaluator) Description() string
Description 返回评估器描述
func (*CostEvaluator) Evaluate ¶
func (e *CostEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行成本评估
func (*CostEvaluator) RequiresLLM ¶
func (e *CostEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type DiversityEvaluator ¶
type DiversityEvaluator struct {
// contains filtered or unexported fields
}
DiversityEvaluator 多样性评估器
评估文本的词汇多样性和表达丰富度。
func NewDiversityEvaluator ¶
func NewDiversityEvaluator(opts ...DiversityOption) *DiversityEvaluator
NewDiversityEvaluator 创建多样性评估器
func (*DiversityEvaluator) Description ¶
func (e *DiversityEvaluator) Description() string
Description 返回评估器描述
func (*DiversityEvaluator) Evaluate ¶
func (e *DiversityEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行多样性评估
func (*DiversityEvaluator) RequiresLLM ¶
func (e *DiversityEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type DiversityOption ¶
type DiversityOption func(*DiversityEvaluator)
DiversityOption 多样性评估器选项
func WithDiversityThreshold ¶
func WithDiversityThreshold(threshold float64) DiversityOption
WithDiversityThreshold 设置通过阈值
type F1Evaluator ¶
type F1Evaluator struct {
// contains filtered or unexported fields
}
F1Evaluator F1 分数评估器
用于分类任务的准确性评估，计算精确率和召回率的调和平均。
func (*F1Evaluator) Evaluate ¶
func (e *F1Evaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行 F1 评估
type FaithfulnessEvaluator ¶
type FaithfulnessEvaluator struct {
// contains filtered or unexported fields
}
FaithfulnessEvaluator 评估回答对检索上下文的忠实度 检查回答是否完全基于提供的上下文，没有编造信息
func NewFaithfulnessEvaluator ¶
func NewFaithfulnessEvaluator(llm evaluate.LLMJudge, opts ...FaithfulnessOption) *FaithfulnessEvaluator
NewFaithfulnessEvaluator 创建忠实度评估器
func (*FaithfulnessEvaluator) Description ¶
func (e *FaithfulnessEvaluator) Description() string
Description 返回评估器描述
func (*FaithfulnessEvaluator) Evaluate ¶
func (e *FaithfulnessEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行忠实度评估
func (*FaithfulnessEvaluator) RequiresLLM ¶
func (e *FaithfulnessEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type FaithfulnessOption ¶
type FaithfulnessOption func(*FaithfulnessEvaluator)
FaithfulnessOption FaithfulnessEvaluator 选项
func WithFaithfulnessStrict ¶
func WithFaithfulnessStrict(strict bool) FaithfulnessOption
WithFaithfulnessStrict 设置严格模式
func WithFaithfulnessThreshold ¶
func WithFaithfulnessThreshold(threshold float64) FaithfulnessOption
WithFaithfulnessThreshold 设置通过阈值
type HallucinationEvaluator ¶
type HallucinationEvaluator struct {
// contains filtered or unexported fields
}
HallucinationEvaluator 检测回答中的幻觉 与 Faithfulness 互补，专注于检测明显的编造
func NewHallucinationEvaluator ¶
func NewHallucinationEvaluator(llm evaluate.LLMJudge, opts ...HallucinationOption) *HallucinationEvaluator
NewHallucinationEvaluator 创建幻觉检测评估器
func (*HallucinationEvaluator) Description ¶
func (e *HallucinationEvaluator) Description() string
Description 返回评估器描述
func (*HallucinationEvaluator) Evaluate ¶
func (e *HallucinationEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行幻觉检测
func (*HallucinationEvaluator) RequiresLLM ¶
func (e *HallucinationEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type HallucinationOption ¶
type HallucinationOption func(*HallucinationEvaluator)
HallucinationOption 选项
func WithHallucinationThreshold ¶
func WithHallucinationThreshold(threshold float64) HallucinationOption
WithHallucinationThreshold 设置阈值
type LatencyEvaluator ¶
type LatencyEvaluator struct {
// contains filtered or unexported fields
}
LatencyEvaluator 评估系统延迟
func NewLatencyEvaluator ¶
func NewLatencyEvaluator(opts ...LatencyOption) *LatencyEvaluator
NewLatencyEvaluator 创建延迟评估器
func (*LatencyEvaluator) Description ¶
func (e *LatencyEvaluator) Description() string
Description 返回评估器描述
func (*LatencyEvaluator) Evaluate ¶
func (e *LatencyEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行延迟评估
func (*LatencyEvaluator) RequiresLLM ¶
func (e *LatencyEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type LatencyOption ¶
type LatencyOption func(*LatencyEvaluator)
LatencyOption LatencyEvaluator 选项
func WithTargetLatency ¶
func WithTargetLatency(ms float64) LatencyOption
WithTargetLatency 设置目标延迟（毫秒）
type PerplexityEvaluator ¶
type PerplexityEvaluator struct {
// contains filtered or unexported fields
}
PerplexityEvaluator 困惑度评估器
评估语言模型生成文本的质量，困惑度越低表示文本越符合预期。 注意：这个实现是简化版本，实际困惑度需要访问模型的对数概率。
func NewPerplexityEvaluator ¶
func NewPerplexityEvaluator(opts ...PerplexityOption) *PerplexityEvaluator
NewPerplexityEvaluator 创建困惑度评估器
func (*PerplexityEvaluator) Description ¶
func (e *PerplexityEvaluator) Description() string
Description 返回评估器描述
func (*PerplexityEvaluator) Evaluate ¶
func (e *PerplexityEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行困惑度评估
注意：这是简化实现，使用统计方法估算困惑度。 真实困惑度需要模型的对数概率。
func (*PerplexityEvaluator) RequiresLLM ¶
func (e *PerplexityEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type PerplexityOption ¶
type PerplexityOption func(*PerplexityEvaluator)
PerplexityOption 困惑度评估器选项
func WithPerplexityThreshold ¶
func WithPerplexityThreshold(threshold float64) PerplexityOption
WithPerplexityThreshold 设置通过阈值
type ROUGEEvaluator ¶
type ROUGEEvaluator struct {
// contains filtered or unexported fields
}
ROUGEEvaluator ROUGE (Recall-Oriented Understudy for Gisting Evaluation) 评估器
主要用于文本摘要质量评估，通过计算 n-gram 召回率来衡量 生成摘要与参考摘要的覆盖程度。
func NewROUGEEvaluator ¶
func NewROUGEEvaluator(opts ...ROUGEOption) *ROUGEEvaluator
NewROUGEEvaluator 创建 ROUGE 评估器
func (*ROUGEEvaluator) Description ¶
func (e *ROUGEEvaluator) Description() string
Description 返回评估器描述
func (*ROUGEEvaluator) Evaluate ¶
func (e *ROUGEEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行 ROUGE 评估
func (*ROUGEEvaluator) RequiresLLM ¶
func (e *ROUGEEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type ROUGEOption ¶
type ROUGEOption func(*ROUGEEvaluator)
ROUGEOption ROUGE 评估器选项
func WithROUGEThreshold ¶
func WithROUGEThreshold(threshold float64) ROUGEOption
WithROUGEThreshold 设置通过阈值
func WithROUGEVariant ¶
func WithROUGEVariant(variant string) ROUGEOption
WithROUGEVariant 设置 ROUGE 变体
type RelevanceEvaluator ¶
type RelevanceEvaluator struct {
// contains filtered or unexported fields
}
RelevanceEvaluator 评估检索文档与查询的相关性
func NewRelevanceEvaluator ¶
func NewRelevanceEvaluator(llm evaluate.LLMJudge, opts ...RelevanceOption) *RelevanceEvaluator
NewRelevanceEvaluator 创建相关性评估器
func (*RelevanceEvaluator) Description ¶
func (e *RelevanceEvaluator) Description() string
Description 返回评估器描述
func (*RelevanceEvaluator) Evaluate ¶
func (e *RelevanceEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行相关性评估
func (*RelevanceEvaluator) RequiresLLM ¶
func (e *RelevanceEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type RelevanceOption ¶
type RelevanceOption func(*RelevanceEvaluator)
RelevanceOption RelevanceEvaluator 选项
func WithRelevanceThreshold ¶
func WithRelevanceThreshold(threshold float64) RelevanceOption
WithRelevanceThreshold 设置通过阈值
type ThroughputEvaluator ¶
type ThroughputEvaluator struct {
// contains filtered or unexported fields
}
ThroughputEvaluator 评估系统吞吐量
func NewThroughputEvaluator ¶
func NewThroughputEvaluator(opts ...ThroughputOption) *ThroughputEvaluator
NewThroughputEvaluator 创建吞吐量评估器
func (*ThroughputEvaluator) Description ¶
func (e *ThroughputEvaluator) Description() string
Description 返回评估器描述
func (*ThroughputEvaluator) Evaluate ¶
func (e *ThroughputEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行吞吐量评估
func (*ThroughputEvaluator) RequiresLLM ¶
func (e *ThroughputEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type ThroughputOption ¶
type ThroughputOption func(*ThroughputEvaluator)
ThroughputOption 选项
func WithMinThroughput ¶
func WithMinThroughput(tps float64) ThroughputOption
WithMinThroughput 设置最小吞吐量
func WithTargetThroughput ¶
func WithTargetThroughput(tps float64) ThroughputOption
WithTargetThroughput 设置目标吞吐量
type ToxicityEvaluator ¶
type ToxicityEvaluator struct {
// contains filtered or unexported fields
}
ToxicityEvaluator 毒性评估器
评估文本是否包含有害、冒犯性或不当内容。
func NewToxicityEvaluator ¶
func NewToxicityEvaluator(llm evaluate.LLMJudge, opts ...ToxicityOption) *ToxicityEvaluator
NewToxicityEvaluator 创建毒性评估器
func (*ToxicityEvaluator) Description ¶
func (e *ToxicityEvaluator) Description() string
Description 返回评估器描述
func (*ToxicityEvaluator) Evaluate ¶
func (e *ToxicityEvaluator) Evaluate(ctx context.Context, input evaluate.EvalInput) (*evaluate.EvalResult, error)
Evaluate 执行毒性评估
func (*ToxicityEvaluator) RequiresLLM ¶
func (e *ToxicityEvaluator) RequiresLLM() bool
RequiresLLM 返回是否需要 LLM
type ToxicityOption ¶
type ToxicityOption func(*ToxicityEvaluator)
ToxicityOption 毒性评估器选项
func WithToxicityThreshold ¶
func WithToxicityThreshold(threshold float64) ToxicityOption
WithToxicityThreshold 设置通过阈值（低于阈值则通过）
Source Files