Research map / Probabilistic and scientific modeling / Molecules and proteins
Learning Discrete Energy-based Models via Auxiliary-variable Local Exploration
Hanjun Dai, Rishabh Singh, Bo Dai, Charles A. Sutton, Dale Schuurmans
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Add local edits (<=16) to fully factorized q0 (ablation) | ALOE-fac · Synthetic 2spirals | MMD-hamming: 51.24 -> 32.6 | Improved |
| Add local edits to fully factorized q0 (ablation) | ALOE-fac · Synthetic 8gaussians | MMD-hamming: 91.2 -> 3 | Improved |
| Add local edits to fully factorized q0 (ablation) | ALOE-fac · Synthetic swissroll | MMD-hamming: 15 -> 0.44 | Improved |
| Variational power iteration objective instead of REINFORCE objective (ADE), factorized q0 (ablation) | ALOE-fac-edit · Synthetic 2spirals | MMD-hamming: 236.6 -> 32.6 | Improved |
| Variational power iteration objective instead of REINFORCE objective (ADE), factorized q0 (ablation) | ALOE-fac-edit · Synthetic circles | MMD-hamming: 261.7 -> -1.5 | Improved |
| ALOE vs PCD-10 baseline | ALOE · Synthetic 2spirals | MMD-hamming: 34.73 -> 30.37 | Improved |
| ALOE vs PCD-10 baseline | ALOE · Synthetic circles | MMD-hamming: -0.3 -> -0.83 | Improved |
| ALOE vs ADE baseline | ALOE · Synthetic pinwheel | MMD-hamming: 7.64 -> -0.64 | Improved |
| ALOE vs ADE on gradient variance | ALOE · pinwheel | gradient variance and EBM NLL | Improved |
| ADE vs PCD | ADE · Synthetic datasets | MMD-hamming | Mixed |
| ALOE sampler vs autoregressive model | ALOE · Generative fuzzing | coverage | Improved |
| Combine LibFuzzer and ALOE inputs | ALOE · Generative fuzzing | coverage | Improved |
Closest papers by meaning
- Energy-Based Modelling for Discrete and Mixed Data via Heat Equations on Structured Spaces (NeurIPS 2024 · similarity 0.81)
- Differentiable Structure Learning and Causal Discovery for General Binary Data (NeurIPS 2025 · similarity 0.81)
- SOInter: A Novel Deep Energy-Based Interpretation Method for Explaining Structured Output Models (ICLR 2024 · similarity 0.78)
- Autoregressive Language Models are Secretly Energy-Based Models: Insights into the Lookahead Capabilities of Next-Token Prediction (ICML 2026 · similarity 0.78)
- Learning Discrete Latent Variable Structures with Tensor Rank Conditions (NeurIPS 2024 · similarity 0.78)
- Differentiable Structure Learning with Partial Orders (NeurIPS 2024 · similarity 0.78)
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.
