Research map / Optimization / Optimizers
Zeroth-order training methods
Trains models using gradient estimates from function queries, with convergence and learning-dynamics analysis.
23 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2019–2026) · an approach in Optimizers, keyword cluster “variational · sampling · inference”
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.
3 papers in 2023–24 and 14 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 1.2% to 4.8%.
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
- 2019 · ZO-AdaMM: Zeroth-Order Adaptive Momentum Method for Black-Box Optimization (NeurIPS 2019 · 33 citations)
- 2020 · Learning to Learn by Zeroth-Order Oracle (ICLR 2020 · 9 citations)
- 2021 · On the Convergence of Prior-Guided Zeroth-Order Optimization Algorithms (NeurIPS 2021 · 1 citation)
- 2022 · Improve Single-Point Zeroth-Order Optimization Using High-Pass and Low-Pass Filters (ICML 2022 · 4 citations)
- 2023 · Zeroth-Order Optimization with Trajectory-Informed Derivative Estimation (ICLR 2023)
- 2024 · ReLIZO: Sample Reusable Linear Interpolation-based Zeroth-order Optimization (NeurIPS 2024 · 0 citations)
- 2025 · PseuZO: Pseudo-Zeroth-Order Algorithm for Training Deep Neural Networks (NeurIPS 2025 · 3 citations)
- 2026 · Learning Dynamics of Zeroth-Order Optimization: A Kernel Perspective (ICML 2026 · 0 citations)
Earliest papers here
- ZO-AdaMM: Zeroth-Order Adaptive Momentum Method for Black-Box Optimization (NeurIPS 2019 · 33 citations)
- Learning to Learn by Zeroth-Order Oracle (ICLR 2020 · 9 citations)
- Min-Max Optimization without Gradients: Convergence and Applications to Black-Box Evasion and Poisoning Attacks (ICML 2020)
What has been tried
12 papers here report 109 measured changes (a component added, removed or swapped, compared with a baseline): helped 65%, no clear effect 12%, hurt 18%, mixed 5%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Replace the standard second-moment estimate with the refined estimate | Quadratic function · relative error of second-moment estimate | Helped | Refining Adaptive Zeroth-Order Optimization at Ease (ICML 2025 · 0 citations) | |
| Set beta to 1, removing the unbiased perturbation contribution | MNIST, 2-layer MLP · training loss | Hurt | Improving the Straight-Through Estimator with Zeroth-Order Information (NeurIPS 2025 · 0 citations) | |
| Reduce the estimator from P4 to P3 using random selection between estimator pathways | Variance analysis · variance | Hurt | On the Optimal Construction of Unbiased Gradient Estimators for Zeroth-Order Optimization (NeurIPS 2025 · 0 citations) | |
| add sampling momentum | CIFAR10 test · test accuracy | 92.45 -> 93.41 | Helped | ReLIZO: Sample Reusable Linear Interpolation-based Zeroth-order Optimization (NeurIPS 2024 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 8 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- SST-2 · 3 of 8 papers
- MNIST · 2 of 8 papers
Baselines
- ZO-SGD · 2 of 8 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 17 papers; from OpenAlex and the papers' affiliations).
- National University of Singapore · 2 papers
- Peking University · 2 papers
Use it
Download BibTeX of 23 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
- Stochastic gradient MCMC (142 papers)
- Natural gradient and Fisher approximations (77 papers)
- Stein variational gradient descent (47 papers)
Elsewhere in the topic
- Surrogate losses and robust solvers (119 papers)
- Differentiating through embedded optimization (117 papers)
- Constrained and non-backprop training (97 papers)
- Memory-efficient LLM optimizers (79 papers)
- Adam convergence and adaptive learning rates (78 papers)
- Randomized gradient estimation (77 papers)
- SGD momentum and step-size schedules (72 papers)
- Second-order and structured preconditioning (70 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).