Loading page...Загружаем страницу...正在加载页面...
AtlasRepo

I Tracked Every AI Hallucination for a Week — The Numbers Were Worse Than I Thought

说明与用途

作者发现约18%的AI输出错误。他构建了一个验证层,自动检查引用、代码、泄漏和参数,节省令牌和调试时间。

问题

AI agents frequently produce hallucinated outputs, wasting token budget and causing debugging effort.

结果

The verification layer reduced hallucinations, saving nearly 20% of token budget and decreasing manual debugging effort.

工作流

  • Run AI model to generate output
  • Pass output to verification layer
  • Citation checker validates references
  • Code validator checks syntax and logic
  • Safety leak detector identifies leaked prompts
  • Argument verifier checks tool call parameters
  • Coherence scorer compares output to prompt
  • If all checks pass, output is accepted; otherwise, it is flagged for review

证据

Out of 200 outputs across Claude, GPT, and DeepSeek: 36 were confidently wrong (18%)

来源

All runs in under 100ms on CPU. Model-agnostic. Free.

来源

  • 记录日期: 2026-09-06T22:07:18.951Z