aiqbench analysis
This research introduces Doc2LoRA, a method that converts scientific papers into LoRA adapters to create a vector space where mixed points represent generatable ideas rather than just static embeddings. By enabling LLMs at any point in this space to answer questions or generate abstracts, the approach bridges the gap between semantic search and creative idea generation for researchers. The technique maintains competitive performance on standard retrieval benchmarks while offering unique capabilities for exploring novel combinations of existing scientific concepts.
Doc2LoRA Provides Decodable Representations of Scientific Ideas
arXiv:2609.38374v1 Announce Type: new Abstract: Representing scientific papers as points in a space lets us search for similar papers and inquire about how fields relate to one another and drive innovation. Beyond search, the vector space of papers invites generation: mixing papers through simple vector operations creates new points, mirroring combinatorial novelty, the recombination of existing ideas into new ones. However, a mixed point often represents an idea no paper has yet realized, with no papers nearby to identify the idea. We propose representing each paper by a LoRA adapter generated by the Doc-to-LoRA hypernetwork. Every point in the space, including mixtures, thus represents a large language model (LLM) open to questions and instructions in natural language. On papers from the American Physical Society (APS), we instruct the LLM at the average of each subfield to name the field in a few words and obtain labels closer to the official names than the labels of five baselines, as judged by word overlap and a panel of five LLM judges. We also ask the LLMs at points between two APS papers to write an abstract and obtain descriptions shifting from one paper to the other in step with the mixing weight. While Doc-to-LoRA is trained for generation, a small invertible transform makes the embeddings competitive for search, on par with SPECTER2 and EmbeddingGemma and close to SBERT. Because the transform is invertible, every point in the transformed space still maps back to an LLM. The embeddings thus serve both search and generation, enabling researchers to question the idea at any point in the space as a starting point for generating new ideas.
Abstract:Representing scientific papers as points in a space lets us search for similar papers and inquire about how fields relate to one another and drive innovation. Beyond search, the vector space of papers invites generation: mixing papers through simple vector operations creates new points, mirroring combinatorial novelty, the recombination of existing ideas into new ones. However, a mixed point often represents an idea no paper has yet realized, with no papers nearby to identify the idea. We propose representing each paper by a LoRA adapter generated by the Doc-to-LoRA hypernetwork. Every point in the space, including mixtures, thus represents a large language model (LLM) open to questions and instructions in natural language. On papers from the American Physical Society (APS), we instruct the LLM at the average of each subfield to name the field in a few words and obtain labels closer to the official names than the labels of five baselines, as judged by word overlap and a panel of five LLM judges. We also ask the LLMs at points between two APS papers to write an abstract and obtain descriptions shifting from one paper to the other in step with the mixing weight. While Doc-to-LoRA is trained for generation, a small invertible transform makes the embeddings competitive for search, on par with SPECTER2 and EmbeddingGemma and close to SBERT. Because the transform is invertible, every point in the transformed space still maps back to an LLM. The embeddings thus serve both search and generation, enabling researchers to question the idea at any point in the space as a starting point for generating new ideas.
32 pages, 4 figures, 12 tables. Code: this https URL
Computation and Language (cs.CL); Digital Libraries (cs.DL); Information Retrieval (cs.IR); Machine Learning (cs.LG); Physics and Society (physics.soc-ph)
arXiv:2609.38374 [cs.CL]
(or arXiv:2609.38374v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2609.38374 arXiv-issued DOI via DataCite (pending registration)
Full text under the publisher's open licence · 308 words
Related coverage
[AINews] Opus 5.5 is good at explainer videos
Latent Space · Sep 29, 2026
[AINews] The Future of Latent Space
Latent Space · Sep 25, 2026
Microsoft’s AI chief says superintelligence is near, but won’t take your job
The Verge · Jun 8, 2026
J++ Lens: Jacobian Filtering Enables More Faithful Workspace Lenses
LessWrong · Oct 8, 2026