Bundle
dsh-arena
Local-first experiment and evaluation workbench for DeepSeek Harness.
Apageoflove4
5 results
Local-first experiment and evaluation workbench for DeepSeek Harness.
Isolated multi-model coding matches with deterministic verification, scoring, and reports
A blind, fair, local Agent arena inside DSH Web: same task, same commit, isolated worktrees, shared verification, judge before you reveal.
Evidence-first multi-model coding comparison workbench for DeepSeek Harness
北极星 (Polaris):通用能力聚合与择优进化层——异构嗅探、双轨竞技场、SLSA 出处、不可变索引与语义调度。