使用方法
- 复制安装命令，在自己的智能体环境安装前先检查源码包。
- 从原始文档核对所需工具、凭据和支持的宿主。
- 说明输入材料与期望交付物，并对照来源检查输出。
npx skills add https://github.com/yogsoth-ai/de-anthropocentric-research-engine --skill select-next-pair开始前需了解
- 安装不包含模型订阅、API 额度或访问文件的权限。
- 仓库星数和安装量反映活跃情况，不等于独立质量或安全测试。
仓库原始内容
按源仓库内容呈现，保留标题、案例、代码、表格、链接以及原文引用的演示图片。
select-next-pair
Purpose
Select the next comparison pair to maximize expected information gain while respecting unresolved uncertainty and coverage.
Input contract
yaml
required: [candidate_set, current_ratings, uncertainty_matrix, comparison_budget]
optional: [pair_history, exploration_policy, exclusion_constraints]
constraints: [candidate IDs are stable; expected information gain uses the declared model and available evidence]Procedure
- Enumerate eligible unjudged pairs and compute their current uncertainty and decision relevance.
- Estimate expected information gain under the declared comparison model, including the value of resolving a tie or cycle.
- Exclude pairs blocked by scope or data constraints, then select the highest-value eligible pair.
- Return the selected pair, rationale, alternatives considered, and the budget impact.
Output contract
yaml
produces: [selected_pair, information_gain_estimate, candidate_pair_table, selection_rationale]
delta_fields: [findings, evidence_updates, uncertainties, decisions, recommended_jumps]Quality gates
- The selected pair is eligible, not already resolved under the same model, and has a stated information-gain rationale.
- If all pairs are exhausted or blocked, return an explicit stop reason.
Failure and counterexamples
Do not select a pair solely because it is adjacent in a list. Expected information gain cannot be claimed when uncertainty inputs are absent.
Provenance map
- resolved: pair-selector
