TraceDogTraceDog
ExperimentsOpen sourceAboutContactGitHub
DashboardRead docs
← Traces / Detail

Trace decision

Short answer — sentence scores vs long passages can look low; treat as a signal.
Needs review

Grounding is borderline and should be reviewed.

Grounding0.55Strong
Risk0.45Medium
Reliability0.62Medium
Recommended actionReview retrieved evidence before shipping
gpt-4o-mini1382ms totalNormans2026-03-26T04:47:09.141293+00:00
Why it scored this way
Why this needs review

Retrieved evidence aligns reasonably well with the response (best hybrid 0.55).

  • Best grounding score: 0.55
  • Sentence match: 0.24
  • Keyword overlap: 0.80
Full scorer narrative

Retrieved evidence aligns reasonably well with the response (best hybrid 0.55).

What we measured

  • Hybrid grounding (per chunk): best 0.55 (strong ≥ 0.52)
  • Mean hybrid across chunks: 0.55
  • Best raw sentence match: 0.24 (short answers often score low vs. whole paragraphs)
  • Lexical overlap with sources: 0.80
  • Blend: short-answer blend (45% sentence + 55% keyword)
  • Weak / strong cutoffs: 0.35 / 0.52
Evidence
Prompt & response
Grounding vs thresholds0.55
weak < 0.35review 0.35–0.52strong ≥ 0.52
Blend contribution
Sentence 0.24 (45%)Keyword 0.80 (55%)
Confidence trend

Series padded from this trace’s score — batch view shows drift.

Failure mix (this trace)

One bar — fleet-wide % needs aggregate metrics from the API.

  • Hallucination risk 100%
  • Low grounding 0%
  • Failures 0%
  • Healthy 0%
Execution runtime

Total 1382ms

Retrieval345ms✓
LLM1037ms✓
  • ✓Retrieval345ms
  • ✓LLM1037ms
Debug