Research map / Probabilistic and scientific modeling / Molecules and proteins
Statistical analysis of ML practice
Applies survival analysis, additive models and variable selection to understand reproducibility and importance.
122 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Molecules and proteins, keyword cluster “decision · machine · statistical”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
38 papers in 2023–24 and 39 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 7.8% to 4.2%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · Copula Ordinal Regression for Joint Estimation of Facial Action Unit Intensity (CVPR 2016 · 49 citations)
- 2017 · Automatic Discovery of the Statistical Types of Variables in a Dataset (ICML 2017 · 27 citations)
- 2018 · Early Prediction of Diabetes Complications from Electronic Health Records: A Multi-Task Survival Analysis Approach (AAAI 2018 · 37 citations)
- 2019 · Minimum Divergence vs. Maximum Margin: an Empirical Comparison on Seq2Seq Models (ICLR 2019 · 9 citations)
- 2020 · Dynamic Time Lag Regression: Predicting What & When (ICLR 2020 · 6 citations)
- 2021 · Neural Symbolic Regression that scales (ICML 2021 · 26 citations)
- 2022 · Predicting Cellular Responses to Novel Drug Perturbations at a Single-Cell Resolution (NeurIPS 2022 · 44 citations)
- 2023 · Transformer-based Planning for Symbolic Regression (NeurIPS 2023 · 16 citations)
- 2024 · Racing Control Variable Genetic Programming for Symbolic Regression (AAAI 2024 · 8 citations)
- 2025 · Whole Genome Transformer for Gene Interaction Effects in Microbiome Habitat Specificity (AAAI 2025 · 3 citations)
- 2026 · PluRel: Synthetic Data unlocks Scaling Laws for Relational Foundation Models (ICML 2026 · 1 citation)
Earliest papers here
- Copula Ordinal Regression for Joint Estimation of Facial Action Unit Intensity (CVPR 2016 · 49 citations)
- Launch and Iterate: Reducing Prediction Churn (NeurIPS 2016 · 23 citations)
- Finding significant combinations of features in the presence of categorical covariates (NeurIPS 2016 · 21 citations)
What has been tried
93 papers here report 868 measured changes (a component added, removed or swapped, compared with a baseline): helped 66%, no clear effect 10%, hurt 18%, mixed 6%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use a better experiment schedule rather than CVGP's default schedule | Trigonometric (4,4,6), 10 expressions · NMSE | Helped | Racing Control Variable Genetic Programming for Symbolic Regression (AAAI 2024 · 8 citations) | |
| Replace gradient ascent with DENs | Jurkat, K562, and THP1 · cell-type-specificity (DE) | Hurt | Designing Cell-Type-Specific Promoter Sequences Using Conservative Model-Based Optimization (NeurIPS 2024 · 6 citations) | |
| halve feedforward layer size from 2048 to 1024 | ProGenomes v3 test set · accuracy | 71.2 -> 66.2 | Hurt | Whole Genome Transformer for Gene Interaction Effects in Microbiome Habitat Specificity (AAAI 2025 · 3 citations) |
| Add QP-Net | Pre-School · MAE | 8.129 -> 7.949 | Helped | Q-PART: Quasi-Periodic Adaptive Regression with Test-time Training for Pediatric Left Ventricular Ejection Fraction Regression (CVPR 2025 · 3 citations) |
| Add MMF | Disease progression identification · F1 | 0.460 -> 0.466 | Helped | Multimodal Disease Progression Modeling via Spatiotemporal Disentanglement and Multiscale Alignment (NeurIPS 2025 · 3 citations) |
| Replace the Utah chilling function with a learned MLP chilling function | Japan — temporal · MAE (days) | 3.17 \textpm 0.07 -> 2.17 \textpm 0.03 | Helped | Hybrid Phenology Modeling for Predicting Temperature Effects on Tree Dormancy (AAAI 2025 · 2 citations) |
| Use service and feature representations instead of omitting representations | CARES of NY service assignment test set · accuracy | 0.717 -> 0.832 | Helped | Predictive Modeling of Homeless Service Assignment: A Representation Learning Approach (AAAI 2025 · 2 citations) |
| Vary clustGeo parameters α and λ across parameterizations | 31 bird species · AUC | Mixed | Spatial Clustering of Citizen Science Data Improves Downstream Species Distribution Models (AAAI 2025 · 2 citations) | |
| Add range text rather than providing no metadata | IUCN, zero-shot · MAP | 0.05 -> 0.52 | Helped | Feedforward Few-shot Species Range Estimation (ICML 2025 · 2 citations) |
| Add a prevalence constraint using known disease prevalence | Synthetic data, Bernoulli-sigmoid models · parameter inference precision and accuracy | Helped | Domain constraints improve risk prediction when outcome data is missing (ICLR 2024 · 2 citations) | |
| Tokenize numerical metrics using uniform quantization | Avg. Test (Base) · MSE | 96.99 -> 37.72 | Helped | ARECHO: Autoregressive Evaluation via Chain-Based Hypothesis Optimization for Speech Multi-Metric Estimation (NeurIPS 2025 · 1 citation) |
| Replace one foundation model with an ensemble of 5 to 7 models | Study by [cite], 50 samples · variance | approximately 6% to nearly 12%; around 16% | Helped | Efficient Randomized Experiments Using Foundation Models (NeurIPS 2025 · 1 citation) |
What reviewers will expect
Datasets and baselines named in the 28 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Baselines
- LSTM · 2 of 28 papers
- Transformer · 2 of 28 papers
- Difference-in-means · 2 of 28 papers
- PPI++ · 2 of 28 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 77 papers; from OpenAlex and the papers' affiliations).
- Cornell University · 4 papers
- Microsoft · 3 papers
- Duke University · 3 papers
- Inria · 3 papers
- Google · 3 papers
- Carnegie Mellon University · 3 papers
- Imperial College London · 3 papers
- Oregon State University · 2 papers
- Harvard T.H. Chan School of Public Health · 2 papers
- University of Cambridge · 2 papers
Use it
Download BibTeX of 40 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Prediction-powered and conformal inference (129 papers)
- Machine learning for physical simulation (96 papers)
- Counterfactual and statistical explanations (81 papers)
- Sparse concepts and hypothesis testing (78 papers)
- Cold-start recommendation models (80 papers)
Elsewhere in the topic
- Generative protein structure design (142 papers)
- Neuro-symbolic probabilistic layers (120 papers)
- Structure-based 3D molecule generation (107 papers)
- Sparse graphical and tensor models (97 papers)
- Flow matching generative modeling (97 papers)
- Multimodal trajectory prediction (95 papers)
- Probabilistic embeddings and relational inference (93 papers)
- Equivariant and disentangled latent representations (90 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).