llm-evaluation

Evidence-gated completion and trusted code verification for DeepSeek Harness agents

Paired A/B evaluation infrastructure for DeepSeek Harness (dsh) components: interleaved repeated runs through the real runtime, verifier self-checks, one-variable arms, regression gating, cache- and calendar-aware cost, CLI + web UI

Read-only repository research, durable project memory, and evidence-gated definitions of done for AI coding agents on DeepSeek Harness.

Open evaluation datasets, test cases, and metrics for DSH plugins.

Controlled request-surface replay and regression workbench for DeepSeek Harness

Sovereign, agent-driven LLM evaluation harness forked from deepseek-harness (dsh) — Cordis plugin architecture, EntheAI backends, and multi-model benchmarking

Visual workflows and multi-model evaluation for DeepSeek Harness

A blind, fair, local DSH Web arena: same task, isolated worktrees, shared verification, judge before reveal.

Compare multiple coding models side by side in DeepSeek Harness with isolated Git worktrees, live traces, and adoptable results

Proof-carrying correction learning for DSH: explicit adoption, scoped recall, and reconstructable delivery.

Reproducible paired A/B experiments for DeepSeek Harness plugins, with isolated runners, auditable evidence, and deterministic promotion gates.