Popular repositories Loading

1. Mythica Mythica Public

   Python 3

2. inside-dsh inside-dsh Public

   A reader's map of the DeepSeek Harness source code. Explore, discover, and transform the interior of DeepSeek Harness

3. silent-zero silent-zero Public

   Silent zero: when a reasoning model's thinking eats the output budget, harnesses report success and score 0.000 — controlled cross-implementation evidence.

   Python

4. inspect_ai inspect_ai Public

   Forked from UKGovernmentBEIS/inspect_ai

   Inspect: A framework for large language model evaluations

   Python

5. fin-data-rulers fin-data-rulers Public

   三层尺量金融数据的对错（源 / 口径 / 时点）+ 引用可追溯问答与自动报告；每一步可复算

   Python

6. two-walls-two-rulers two-walls-two-rulers Public

   Two Walls, Two Rulers — what context compaction actually drops in a real agent deployment: 28-page report + 19 measurement devices + aggregate data.

   JavaScript