AgentSkillsCN

novelty-matrix

创建新颖性/前期工作对比矩阵，将投稿的贡献与相关研究进行比对（找出重叠与差异）。 **触发条件**：新颖性矩阵、前期工作矩阵、重叠与差异、相关工作对比、新颖性矩阵。 **适用场景**：在同行评审中评估新颖性与定位，需要将投稿的贡献与相关研究逐一对照，并详细阐述差异点的证据。 **若跳过**：如果尚未提取主张（请先运行 `claims-extractor`），或您并不打算进行新颖性定位分析。 **网络环境**：无需联网（除非额外提供相关研究，否则不在本次任务范围内）。 **行为准则**：明确标注重叠与差异；尽量提供可追溯的证据来源（来自稿件、引用或作者陈述）。

SKILL.md
--- frontmatter
name: novelty-matrix
description: |
  Create a novelty/prior-work matrix comparing the submission’s contributions against related work (overlaps vs deltas).
  **Trigger**: novelty matrix, prior-work matrix, overlap/delta, 相关工作对比, 新颖性矩阵.
  **Use when**: peer review 中评估 novelty/positioning，需要把贡献与相关工作逐项对齐并写出差异点证据。
  **Skip if**: 缺少 claims（先跑 `claims-extractor`）或你不打算做新颖性定位分析。
  **Network**: none (retrieval of additional related work is out-of-scope unless provided).
  **Guardrail**: 明确 overlap 与 delta；尽量给出可追溯证据来源（来自稿件/引用/作者陈述）。

Novelty Matrix (overlap vs delta)

Goal: make novelty/positioning concrete by aligning each contribution against the closest prior work.

Inputs

Required:

  • •output/CLAIMS.md

Optional:

  • •A provided list of related work (titles/URLs/DOIs) or the submission’s reference list

Outputs

  • •output/NOVELTY_MATRIX.md

Output format (recommended)

  • •Rows: contributions/claims (from output/CLAIMS.md)
  • •Columns: closest related works (provided or cited by the paper)
  • •For each row/column, record:
    • •overlap: what is the same
    • •delta: what is different/new
    • •evidence: where this is supported (paper statement / citation / method difference)

Workflow

  1. •Extract the contribution list from output/CLAIMS.md.
  2. •Pick ≥5 closest related works (or state explicitly why you cannot).
  3. •For each contribution, compare to each related work:
    • •identify overlap
    • •identify delta
    • •attach evidence (quote/section/citation pointer)
  4. •Summarize:
    • •which contributions look clearly novel
    • •which ones look like incremental variants

Definition of Done

  • • Matrix includes ≥5 related works or explains why unavailable.
  • • Every “delta” entry has a short evidence pointer (not just opinion).

Troubleshooting

Issue: no related works list is available

Fix:

  • •Use the paper’s own citations as the initial related set; if even that is missing, mark needs_related_work_list and stop.

Issue: overlap/delta becomes vague

Fix:

  • •Force each cell to reference a concrete axis (problem setting, method component, training data, evaluation protocol, result).