使い方
- インストールコマンドをコピーし、自分の環境へ導入する前にソースを確認します。
- 必要なツール、認証情報、対応環境を原文で確認します。
- 入力と成果物を指定し、出力を資料と照合します。
npx skills add https://github.com/yogsoth-ai/de-anthropocentric-research-engine --skill select-next-pair開始前の確認
- 導入にはモデル契約、API枠、ファイルへのアクセス許可は含まれません。
- スター数や導入数は活動の指標であり、独立した品質・安全性試験ではありません。
リポジトリの原文
見出し、例、コード、表、リンク、参照画像を含む原文を表示しています。
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
