{
  "id": 20458608,
  "title": "The scorer is a program too, and it is wrong more often than the agent",
  "url": "https://urgent.news/2026/11/17/the-scorer-is-a-program-too-and-it-is-wrong-more-often-than-the-agent",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-11-17T16:30:35.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/rokasr/the-scorer-is-a-program-too-and-it-is-wrong-more-often-than-the-agent-2ogf"
  },
  "original_language": "en",
  "account": "Voice agents can misclassify transcripts, leading to false negatives and false positives. Deterministic regex matching used in scoring software is brittle against spontaneous speech, as callers may omit expected words or use negated constructions. A two-tier scoring system can address this issue. Tier one applies strict deterministic matching for predictable phrasing, while tier two escalates unresolved transcripts to an independent model for adjudication. This approach reduces noise in the scoring process, resulting in more accurate evaluations.",
  "summary": "By Rokas Remeika. Published 8 October 2026. When a voice agent evaluation pipeline scores a flawless transcript as unclassifiable, or flags an agreement as a refusal, the scoring logic itself is introducing noise. Deterministic regex matches are brittle against spontaneous speech. Callers elide expected words, causing false negatives, or use negated constructions like \"I would not say no\" that…",
  "key_points": [
    "Voice agents misclassify transcripts, causing errors",
    "Two-tier scoring system proposed to improve accuracy",
    "Tier one uses strict deterministic matching"
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}