AI Paper Atlas

Research map / Probabilistic and scientific modeling / Molecules and proteins

ReLaSH: Reconstructing Joint Latent Spaces for Efficient Generation of Synthetic Hypergraphs with Hyperlink Attributes

Feiyan Ma, Shihao Wu, Gongjun Xu, Ji Zhu

ICLR 2026 · Probabilistic and scientific modeling › Molecules and proteins · approach: Probabilistic embeddings and relational inference

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add degree-sequence calibration (ReLaSHc) at generation (ablation)ReLaSH-(7, 0, 2) · MIMIC-III patient profilesFED: FED 0.532 -> 0.013; Delta Hv 3.260 -> 27.794Mixed
add degree-sequence calibration (ReLaSHc) at generation (ablation)ReLaSH-(2, 2, 2) · MADStat co-citationFED: FED 1.246 -> 5.481; Delta Hv 1.996 -> 3.890Hurt
add degree-sequence calibration (ReLaSHc) at generation (ablation)ReLaSH-(2, 7, 8) · MADStat co-citationFED: FED 1.060 -> 0.947; Delta Hv 1.914 -> 2.772Mixed
add degree-sequence calibration (ReLaSHc) at generation (ablation)ReLaSH-(5, 0, 6) · Recipe generationFED: FED 0.003 -> 0.191; Delta Hv 2.129 -> 3.583Hurt
add degree-sequence calibration (ReLaSHc) at generation (ablation)ReLaSH-(5, 0, 16) · Recipe generationFED: FED 0.766 -> 0.180; Delta Hv 2.355 -> 1.847Mixed
increase hyperlink-only latent dimension k3 from 2 to 16 (ablation)ReLaSH-(7, 0, k3) · MIMIC-III patient profilesFED: FED 0.532 -> 11.738Hurt
increase hyperlink-only latent dimension k3 from 2 to 6 (ablation)ReLaSH-(5, 0, k3) · Recipe generationFED: FED 0.293 -> 0.003Improved
increase hyperlink-only latent dimension k3 from 6 to 16 (ablation)ReLaSH-(5, 0, k3) · Recipe generationFED: FED 0.003 -> 0.766Hurt
increase latent dimension k from 6 to 12 (ablation)joint embedding model (simulation) · simulated hypergraphs, rho=0slope of log error vs log n: -0.4907 -> -0.5462No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy