AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Harpoon: Generalised Manifold Guidance for Conditional Tabular Diffusion

Aditya Shankar, Yuandou Wang, Rihan Hai, Lydia Y. Chen

ICLR 2026 · Image generation › Fast diffusion sampling · approach: Guidance for conditional diffusion · 0 citations (OpenAlex)

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
use MAE inference-time loss instead of MSE (ablation)Harpoon · Default imputation MAR 0.25accuracy: 71.57 -> 73.49Improved
use MAE inference-time loss instead of MSE (ablation)Harpoon · Adult MSE, MAR 0.25MSE: 1.15 -> 0.99Improved
use MAE inference-time loss instead of MSE (ablation)Harpoon · Default MSE, MAR 0.75MSE: 0.88 -> 0.72Improved
add cross-entropy term to MAE loss (ablation)Harpoon · Adult accuracy MAR 0.25accuracy: 69.44 -> 69.30No clear effect
use constant guidance instead of linear schedule (ablation)Harpoon · Default accuracy MAR 0.75accuracy: 54.64 -> 61.48Improved
use constant guidance instead of linear schedule (ablation)Harpoon · Adult MSE MAR 0.75MSE: 1.37 -> 1.26Improved
use Harpoon manifold guidance instead of DiffPuter on range constraintsDiffPuter · Default range constraintviolation rate (%): 88.53 -> 7.78Improved
use Harpoon instead of DiffPuter on conjunction constraintsDiffPuter · Adult range AND categoryviolation rate (%): 78.88 -> 3.37Improved
use Harpoon instead of DiffPuter on categorical constraintsDiffPuter · Adult category constraintviolation rate (%): 0.00 -> 2.18Hurt
use Harpoon instead of DiffPuter on categorical imputation at 0.25DiffPuter · Default categorical imputation MAR 0.25accuracy: 71.58 -> 73.49Improved
use Harpoon instead of DiffPuter on Gesture-like continuous imputationDiffPuter · Default MSE MAR 0.25MSE: 0.62 -> 0.51Improved
add tangential gradient guidance steps over DiffPuterDiffPuter · Adult runtimeseconds: 2.83 -> 4.43Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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