Ideas Have Genomes: Benchmarking Scientific Lineage Reasoning and Lineage-Grounded Idea Generation

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Yifan Zhou, Qihao Yang, Yan Li, Donggang Li, Xiru Hu, Hokin Deng, Ziyang Gong, Xuanyi Zhou, Huacan Wang, Xiangchao Yan, Wanghan Xu, Wenlong Zhang, Shaofeng Zhang, Yue Zhou, Yifan Yang, Zhihang Zhong, Xue Yang
• Venue: arXiv: cs.AI
• Published: 2026-07-09
• Collected: 2026-07-10
• URL: http://arxiv.org/abs/2607.08758v1
• PDF: https://arxiv.org/pdf/2607.08758v1
• DOI: UNKNOWN

Plain-Language Brief

The text presents IdeaGene-Bench (IG-Bench), a new benchmark for evaluating AI systems' ability to follow the structure of scientific idea inheritance. IG-Bench is based on the IdeaGene framework, which represents each paper or proposal as a set of minimal, typed, and evidence-grounded objects. The benchmark includes 1,961 golden lineage traces, 1,085 curated Idea Genome objects, and 920 pairwise GenomeDiff records across 10 scientific domains.

What the Document Says

• IdeaGene-Bench (IG-Bench) is a benchmark for evaluating AI systems' ability to follow the structure of scientific idea inheritance.
• The benchmark contains 1,961 golden lineage traces, 1,085 curated Idea Genome objects, and 920 pairwise GenomeDiff records across 10 scientific domains.
• IG-Bench is based on the IdeaGene framework that represents papers or proposals as sets of minimal, typed, evidence-grounded Idea Genome objects.
• A GenomeDiff aligns these objects to record inheritance, mutation, loss, external import, and novel insertion under six operational evolutionary dynamics.

Key Concepts Mentioned

• IdeaGene-Bench
• Idea Genome
• Golden lineage traces
• GenomeDiff
• Operational evolutionary dynamics

Claims Made by the Document

• IG-Bench can evaluate AI systems' ability to follow the structure of scientific idea inheritance.
• The strongest system reaches only 27.3% exact accuracy on lineage reasoning.

Evidence Used

• The presentation of IdeaGene-Bench and its purpose.
• Statistics about the number of golden lineage traces, curated Idea Genome objects, and pairwise GenomeDiff records.

Methods / Approach

• Using the IdeaGene framework to represent and evaluate AI systems' ability to follow scientific idea inheritance structure

Limitations Stated by the Authors

• The paper does not provide detailed information about the specific algorithms or models used in the experiments.

Limitations Noted by the Agent

Agent interpretation — not from the source.

• The data provided is for research purposes only. The actual performance of AI systems on real-world problems may vary.
• The benchmark is focused on a specific approach (LLM-based scientists), so its applicability to other types of AI or domains might be limited.

Open Questions

• What are the LLM-based scientists mentioned in the text?
• How does the benchmark handle cases where the inheritance structure is unclear or ambiguous?

Notes for Future Q&A

May include interpretation. Labeled where it does.

• LLM likely stands for Large Language Model.
• The text suggests that the benchmark can handle different scientific domains, but more details about its adaptability would be useful.

Raw Abstract / Excerpt

Scientific ideas rarely start from a blank page. They inherit mechanisms, repair known limitations, and recombine pieces of earlier work, much like biological genomes. Current benchmarks still say little about whether AI systems can follow this inheritance structure. We present IdeaGene-Bench (IG-Bench), a benchmark for scientific lineage reasoning and lineage-grounded idea generation. IG-Bench is organized around the IdeaGene framework: each paper or proposal is represented as a set of minimal, typed, evidence-grounded Idea Genome objects, and a GenomeDiff aligns these objects to record inheritance, mutation, loss, external import, and novel insertion under six operational evolutionary dynamics. The benchmark contains 1,961 golden lineage traces, 1,085 curated Idea Genome objects, and 920 pairwise GenomeDiff records across 10 scientific domains. It supports two evaluations. IG-Exam (42 task types, 1,029 instances) tests closed-form lineage reasoning across Idea Genome abstraction, inheritance tracing, evolutionary reasoning, and lineage verification. IG-Arena evaluates generation with a lineage-conditioned Population-Evolution Score(PES), asking whether a proposal can be inserted as a coherent descendant of a given lineage population: it should inherit the right Idea Genome objects, vary meaningfully from nearby work, and offer selection value for future research. Experiments on 14 LLM-based scientists expose a compositional bottleneck. The strongest system reaches only 27.3% exact accuracy on lineage reasoning, and structured lineage context reshuffles system rankings rather than helping every participant uniformly.