TCA-SIR: Learning Target-Conditioned Abstractions for Scientific Inspiration Retrieval

2026-07-30

AI Summary

Existing scientific inspiration retrieval (SIR) methods rely solely on topical similarity, limiting their ability to uncover cross-domain, transferable problem-solving principles. This work proposes a Target-Conditioned Abstraction (TCA) framework that introduces, for the first time, a goal-guided abstraction mechanism to explicitly model the transferability of source literature with respect to a given target problem and generate interpretable abstract principles. Leveraging deep learning, TCA produces target-conditioned abstract representations to predict the degree of transferability between documents and the target. Experiments on ResearchBench demonstrate that TCA significantly outperforms current SIR approaches and direct large language model baselines, achieving a HitRate@top4% that exceeds MOOSE-Chem by more than 10 percentage points.

Abstract

Scientific hypothesis generation for AI for Science typically involves Scientific Inspiration Retrieval (SIR) followed by hypothesis composition. Existing SIR methods rank papers by topical similarity and do not explicitly represent how a candidate inspiration transfers to a target problem. This is especially limiting for remote inspirations, whose value often lies in reusable problem-solving principles rather than topical overlap. Motivated by how humans abstract transferable aspects of a source and remap them to a new target, we reformulate SIR as target-conditioned abstraction (TCA). The retrieval object is a transferable abstract principle extracted from a candidate specifically for the target. We present TCA-SIR, which learns to generate target-conditioned abstractions and uses their representations to predict transferability. On ResearchBench, TCA-SIR outperforms prior SIR methods and direct LLM retrieval, improving HitRate@top4% over MOOSE-Chem by more than 10 percentage points. Learned abstractions also recover target-relevant mechanisms more clearly than an untrained TCA prompt, yielding both stronger retrieval and an interpretable rationale for scientific inspiration.

Problem

• Scientific Inspiration Retrieval
• target-conditioned abstraction
• transferability
• hypothesis generation
• remote inspiration

Innovation

• Target-Conditioned Abstraction
• Scientific Inspiration Retrieval
• Transferable Principles
• Hypothesis Generation
• Interpretable Retrieval