diff --git a/internal/coverage/evaluator.go b/internal/coverage/evaluator.go index f278bce..8c5c3fe 100644 --- a/internal/coverage/evaluator.go +++ b/internal/coverage/evaluator.go @@ -14,18 +14,19 @@ var noAnswerPhrases = []string{ "没有找到", "无法确定", "暂无相关", + "暂无已发布的权威知识", "知识库中没有", } // Evaluator classifies a successful turn into a coverage verdict. type Evaluator struct{} -// NewEvaluator creates a v1 Evaluator. +// NewEvaluator creates the current Evaluator. func NewEvaluator() *Evaluator { return &Evaluator{} } -// Version returns the evaluator version, always "v1". +// Version returns the current evaluator version. func (e *Evaluator) Version() string { return EvaluatorVersion } diff --git a/internal/coverage/evaluator_test.go b/internal/coverage/evaluator_test.go index 8134b8b..145be28 100644 --- a/internal/coverage/evaluator_test.go +++ b/internal/coverage/evaluator_test.go @@ -8,8 +8,8 @@ import ( func TestEvaluatorVersion(t *testing.T) { e := NewEvaluator() - if e.Version() != "v1" { - t.Errorf("Evaluator version = %q, want v1", e.Version()) + if e.Version() != "v2" { + t.Errorf("Evaluator version = %q, want v2", e.Version()) } } @@ -228,28 +228,56 @@ func TestEvaluator_NoAnswerPhraseMatchesInsideAnswer(t *testing.T) { } } -func TestEvaluator_DoesNotExpandBeyondApprovedNoAnswerPhrases(t *testing.T) { +func TestEvaluator_ProductionUnavailableAuthorityPhraseTrumpsReferences(t *testing.T) { e := NewEvaluator() turn := core.Turn{ - ID: "test-turn", + ID: "cra-price-turn", Status: "succeeded", - Answer: "没有相关限制,设备可以正常运行。", + Answer: "当前暂无已发布的权威知识明确说明CRA系列机器人的价格信息。若需要获取对应报价,建议联系越疆官方销售渠道或技术支持进行咨询。", References: []core.Reference{ - {ID: "r1", Score: 0.9}, + {ID: "ref-1", Score: 0.0164}, }, } d, err := e.Evaluate(turn) if err != nil { - t.Fatalf("unexpected error: %v", err) + t.Fatalf("Evaluate: %v", err) } - if d.Verdict != VerdictCovered { - t.Fatalf("verdict = %s, want covered", d.Verdict) + if d.Verdict != VerdictUncovered { + t.Fatalf("verdict = %s, want %s", d.Verdict, VerdictUncovered) + } + if d.ReasonCode != ReasonUncoveredExplicitNoAnswer { + t.Fatalf("reason_code = %s, want %s", d.ReasonCode, ReasonUncoveredExplicitNoAnswer) + } +} + +func TestEvaluator_DoesNotExpandBeyondApprovedNoAnswerPhrases(t *testing.T) { + e := NewEvaluator() + answers := []string{ + "没有相关限制,设备可以正常运行。", + "当前暂无已发布价格调整计划,现有价格继续执行。", + } + for _, answer := range answers { + turn := core.Turn{ + ID: "test-turn", + Status: "succeeded", + Answer: answer, + References: []core.Reference{ + {ID: "r1", Score: 0.9}, + }, + } + d, err := e.Evaluate(turn) + if err != nil { + t.Fatalf("answer %q: unexpected error: %v", answer, err) + } + if d.Verdict != VerdictCovered { + t.Fatalf("answer %q: verdict = %s, want covered", answer, d.Verdict) + } } } func TestEvaluator_NoScoreThreshold(t *testing.T) { - // v1 evaluator must never use score threshold to determine covered/uncovered. + // The evaluator must never use score threshold to determine covered/uncovered. // Even with very low score references, if there are references and no explicit // no-answer, it should be covered. e := NewEvaluator() @@ -266,7 +294,7 @@ func TestEvaluator_NoScoreThreshold(t *testing.T) { t.Fatalf("unexpected error: %v", err) } if d.Verdict != VerdictCovered { - t.Errorf("verdict = %s, want covered (v1 never uses score threshold)", d.Verdict) + t.Errorf("verdict = %s, want covered (score threshold is not used)", d.Verdict) } if d.ReasonCode != ReasonCoveredHasReferences { t.Errorf("reason_code = %s, want %s", d.ReasonCode, ReasonCoveredHasReferences) diff --git a/internal/coverage/types.go b/internal/coverage/types.go index 9ac7d89..33d9af1 100644 --- a/internal/coverage/types.go +++ b/internal/coverage/types.go @@ -31,7 +31,7 @@ const ( ) // EvaluatorVersion is the current version of the evaluator. -const EvaluatorVersion = "v1" +const EvaluatorVersion = "v2" var ( ErrTurnNotSucceeded = errors.New("coverage evaluator: turn status must be succeeded")