JimchengChina/dsh-frontier-repro

JimchengChina★ 0JavaScript最后同步: 2026-08-16

在 GitHub 打开

Evidence-first frontier AI radar and reproducibility gate for DeepSeek Harness

README 摘要

dsh-frontier-repro An evidence-first frontier AI radar and reproduction gate for DeepSeek Harness. This is not another generic news reader, arXiv search engine, or paper summarizer. It covers the missing path from a curated primary-source signal to an auditable reproduction decision: exact, scaled, behavioral, blocked, or insufficiently evidenced. Executed attempts are recorded separately and a passed verdict requires commands, artifacts, measured metrics, and a passing saved rubric. 中文说明 What is different Existing DSH plugins already cover literature discovery/full text ( dsh-ai4scholar , dsh-literature ), guided paper reading and notes ( dsh-paper-workshop ), general RSS reading ( dsh-news ), scientific-claim adjudication ( dsh-research-plugins ), and replayable experiment execution ( dsh-science-workbench ). This plugin does not duplicate those surfaces. It adds: - a single timeline across arXiv, first-party lab releases, official model artifacts, verified personal blogs, and opt-in X API timelines; - explicit identity provenance for person sources; - explainable ranking rather than an opaque quality label; - a mode-specific evidence matrix for exact, scaled, and behavioral repr…

在 GitHub 查看完整 README →
Agent/智能体ai-agentsai-researcharxivdeepseek-harnessdsh-pluginreproducibilitysearch

分类