IMAGIREP: how you represent imagined worlds matters more than RGB
mohitban47 · x · 2026-10-29
- New paper IMAGIREP ("Representing the Unseen") challenges the default in world models: feeding reasoners imagined future RGB views isn't always helpful — even ground-truth oracle endpoint RGB hurts accuracy on farthest-object questions.
- Representing the same imagined state as depth or a top-down map improves performance, so the question isn't just whether to imagine, but how to represent the imagination.
- Follow-up to AVIC (NeurIPS 2026) on adaptive when to imagine; together they frame when-to-imagine plus how-to-represent for spatial reasoning.
Related event: Paper: Depth and Top-Down Views Beat RGB for Spatial Reasoning(3 posts)→
More from Research
- Albert Gu's new H-Net hits 3-5x efficiency gains learning directly from raw bytes — stochasticchasm · 2026-10-29
- DNA motif search tool adds sequence search linked to tripartite graph and ESM2 embeddings — jmschreiber91 · 2026-10-29
- Interactive filters for TF–motif interaction data: family, source, and confidence — jmschreiber91 · 2026-10-29
- Preprint: Human peer reviewers caught only 27% of major errors, all 15 AI models beat them (53-96%) — anshulkundaje · 2026-10-29
- New paper rules out Hamiltonian Phase States as a route to quantum-only Microcrypt — StefanoGogioso · 2026-10-29
- MSR AI for Science team hiring postdoc to generate molecular data for Skala — FrankNoeBerlin · 2026-10-29