LLM-as-a-Verifier for DeepSeek Harness
TaurenMountain/dsh-llm-as-a-verifier · skills
LLM for DeepSeek Harness as a certifier: fine particle size awards, best choice in Probability Hub Championship N and tracking of progress per step as proxy tool.
★ 6GitHub stars
0Forks
2026-09-01Last updated
TypeScriptLanguage
NOASSERTIONLicense
Key features
- Why is it more subtle than "LLM-as-a-Judge"?
- The core idea of the upstream framework is: 1 rating by fine particle size (20-class letter scale)
- 2 Full logprob distribution of token rating expectations (rather than only aragmax)
- 3 Repetition to assess reliability of standard decomposition.
- This plugin is based on its open source for scoring extraction, pairing of hints, tournaments and progress tracking, and access to logprob backend and token measurements, with the Cordis tool plugin encapsulated as DSH.
Install command
dsh plugin --profile web add github:TaurenMountain/dsh-llm-as-a-verifier