Loading...
Loading...
Rigor Explore兼容的skill slug,用于生成有意义且具备潜在新颖性的深度学习研究候选方案。当研究者已选定任务类别、数据集、基准、评估方法,提供了SOTA参考,且希望在`current_research`基础上仅开展候选方案探索,同时需要可审计的仓库理解、想法筛选、公平对比,以及将受控实验结果写入`explore_outputs/`时,可使用本技能。请勿将其用于以README优先的可信复现、开放式方向探索、仅代码或仅运行的窄范围探索、被动仓库分析、已验证的新颖性声明或隐式实验。
npx skill4agent add lllllllama/rigorpilot-skills ai-research-explorecurrent_researchai-research-explore../../references/agent-operating-principles.md../../references/research-rigor-principles.md../../references/deep-learning-experiment-principles.mdcurrent_researchexplore-codeexplore-runanalyze-projectai-research-reproductioncurrent_researchvariant_specresearch_campaignanalyze-projectexplore-codeexplore-runminimal-run-and-auditrun-trainanalysis_outputs/sources/explore_outputs/SCIENTIFIC_CHANGELOG.mdCOMPARABILITY_REPORT.mdevaluation_sourcesota_referenceresearch_campaigncurrent_researchtask_familydatasetbenchmarkevaluation_sourcesota_referencecompute_budgetcandidate_ideasvariant_specresearch_lookupidea_policyidea_generationsource_constraintsfeasibility_policybaseline_gateexecution_policyreferences/research-campaign-spec.mdreferences/ai-research-explore-policy.mdreferences/research-campaign-spec.md../../references/explore-variant-spec.md../../references/research-rigor-principles.md../../references/deep-learning-experiment-principles.mdscripts/orchestrate_explore.pyscripts/write_outputs.py