AI Paper Atlas

Research map / Probabilistic and scientific modeling / Causal discovery

TRACE: Trajectory Recovery for Continuous Mechanism Evolution in Causal Representation Learning

Shicheng Fan, Kun Zhang, Lu Cheng

ICML 2026 · Probabilistic and scientific modeling › Causal discovery · approach: Causal representation identifiability · 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
constrain mixing coefficients to the simplex instead of unconstrained combinations (ablation)TRACE · Synthetic OOD transitionsWeight Corr: 0.971 -> 0.956Hurt
increase observation noise from 0.01 to 0.50 (ablation)TRACE · Synthetic noise sweepMCC: 0.996 -> 0.703Hurt
increase observation noise from 0.01 to 0.50 (ablation)TRACE · Synthetic noise sweepCorr: 0.998 -> 0.854Hurt
increase perturbation magnitude from 0.1 to 0.7 (ablation)TRACE · Synthetic perturbation sweepMCC: 0.934 -> 0.990Improved
increase perturbation magnitude from 0.1 to 0.7 (ablation)TRACE · Synthetic perturbation sweepCorr: 0.972 -> 0.998Improved
add scale calibration (ablation)TRACE · Synthetic noise sweep (sigma=0.10)MAE: 0.254 -> 0.030Improved
increase number of active mechanisms from 2 to 7 (simple) (ablation)TRACE · Synthetic K_total=10alpha recovery corr: .993 -> .776Hurt
increase number of active mechanisms from 2 to 7 (complex) (ablation)TRACE · Synthetic K_total=10alpha recovery corr: .979 -> .459Hurt
increase number of active mechanisms from 2 to 7 (simple) (ablation)TRACE · Synthetic K_total=10W recovery corr: 1.000 -> .995No clear effect
reduce training data from 100% to 10% (ablation)TRACE · Synthetic sample efficiencyMCC: 0.963 -> 0.752 ± 0.163Hurt
reduce training data from 100% to 10% (ablation)TRACE · Synthetic sample efficiency (K=3)Weight Corr: 0.986 -> 0.960 ± 0.013No clear effect
reduce training data from 100% to 1% (ablation)TRACE · Synthetic sample efficiency (K=3)Weight Corr: 0.986 -> 0.688 ± 0.377Hurt

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