Research map / Architectures and efficiency / RNNs and normalization
Dataset condensation and flow estimation
Condenses large image datasets and improves optical flow and ResNet training strategies.
126 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in RNNs and normalization, keyword cluster “memory · scaling · cost”
State History Tried 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.
36 papers in 2023–24 and 42 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 5.5% to 4.5%.
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 · D3: Deep Dual-Domain Based Fast Restoration of JPEG-Compressed Images (CVPR 2016 · 229 citations)
- 2017 · InterpoNet, a Brain Inspired Neural Network for Optical Flow Dense Interpolation (CVPR 2017 · 49 citations)
- 2018 · Towards Image Understanding from Deep Compression Without Decoding (ICLR 2018 · 110 citations)
- 2019 · In Defense of Pre-Trained ImageNet Architectures for Real-Time Semantic Segmentation of Road-Driving Images (CVPR 2019 · 438 citations)
- 2020 · Learned Image Compression With Discretized Gaussian Mixture Likelihoods and Attention Modules (CVPR 2020 · 1,117 citations)
- 2021 · Revisiting ResNets: Improved Training and Scaling Strategies (NeurIPS 2021 · 205 citations)
- 2022 · ELIC: Efficient Learned Image Compression With Unevenly Grouped Space-Channel Contextual Adaptive Coding (CVPR 2022 · 443 citations)
- 2023 · FlowFormer++: Masked Cost Volume Autoencoding for Pretraining Optical Flow Estimation (CVPR 2023 · 113 citations)
- 2024 · Selective-Stereo: Adaptive Frequency Information Selection for Stereo Matching (CVPR 2024 · 103 citations)
- 2025 · SuperLightNet: Lightweight Parameter Aggregation Network for Multimodal Brain Tumor Segmentation (CVPR 2025 · 22 citations)
- 2026 · WAFT: Warping-Alone Field Transforms for Optical Flow (ICLR 2026 · 1 citation)
Earliest papers here
- D3: Deep Dual-Domain Based Fast Restoration of JPEG-Compressed Images (CVPR 2016 · 229 citations)
- PatchBatch: A Batch Augmented Loss for Optical Flow (CVPR 2016 · 90 citations)
- Efficient Learning of Timeseries Shapelets (AAAI 2016 · 75 citations)
What has been tried
105 papers here report 1,152 measured changes (a component added, removed or swapped, compared with a baseline): helped 71%, no clear effect 6%, hurt 20%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add multi-kernel SRUs in place of RAFT-Stereo GRUs | Scene Flow test · EPE (px) | 0.53 -> 0.50 | Helped | Selective-Stereo: Adaptive Frequency Information Selection for Stereo Matching (CVPR 2024 · 103 citations) |
| Apply frequency-selective masking with α=0.5 | Gen1 val · mAP | S4D-legS: 46.93 -> 47.33; S5-legS: 48.33 -> 48.48 | Mixed | State Space Models for Event Cameras (CVPR 2024 · 44 citations) |
| Add EKF-based landmark solver | EuRoC · APE RMSE | 0.107 -> 0.075 | Helped | SchurVINS: Schur Complement-Based Lightweight Visual Inertial Navigation System (CVPR 2024 · 36 citations) |
| increase iterations per scale from (2, 2, 2) to (4, 4, 4) | MSCOCO · MACE | 0.031 -> 0.023 | Helped | MCNet: Rethinking the Core Ingredients for Accurate and Efficient Homography Estimation (CVPR 2024 · 27 citations) |
| Add Sparse Feature Pulling | nuScenes val, vehicle, batch size 1 · memory usage | 4.57 -> 3.50 | Helped | PointBeV: A Sparse Approach for BeV Predictions (CVPR 2024 · 23 citations) |
| add bidirectional scanning to THPA | BraTS2019 ablation · DSC | 78.84 -> 77.58 | Hurt | SuperLightNet: Lightweight Parameter Aggregation Network for Multimodal Brain Tumor Segmentation (CVPR 2025 · 22 citations) |
| replace the PKConv encoder with ResNet | Sintel clean · EPE | 1.40 -> 1.28 | Helped | Recurrent Partial Kernel Network for Efficient Optical Flow Estimation (AAAI 2024 · 22 citations) |
| Add DD, GBM, and GSM with generalized matching | ImageNet-1k, IPC 10 · Top-1 accuracy | 21.3% -> 31.4% | Helped | Generalized Large-Scale Data Condensation via Various Backbone and Statistical Matching (CVPR 2024 · 19 citations) |
| add MLE to CycleGAN | LASED, clipped signals (τ=6g) · CSREτ (g) | 1.803 -> 0.474 | Helped | HEROS-GAN: Honed-Energy Regularized and Optimal Supervised GAN for Enhancing Accuracy and Range of Low-Cost Accelerometers (AAAI 2025 · 17 citations) |
| Add the ZoomOut consistency loss | SHREC19 feature functions after training on SCAPE · feature smoothness | Helped | Memory-Scalable and Simplified Functional Map Learning (CVPR 2024 · 13 citations) | |
| Replace the proposed initialization with random initialization | Truck, RE bound 0.05 · compression ratio | 6.42% -> 8.17% | Hurt | SVDinsTN: A Tensor Network Paradigm for Efficient Structure Search from Regularized Modeling Perspective (CVPR 2024 · 9 citations) |
| Add the centrality layer | OASIS-1 test · Centrality | 16125 ± 11494 -> 12.0 ± 39.48 | Helped | MultiMorph: On-demand Atlas Construction (CVPR 2025 · 8 citations) |
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 78 papers; from OpenAlex and the papers' affiliations).
- Tsinghua University · 4 papers
- Zhejiang University · 3 papers
- Harbin Institute of Technology · 2 papers
- Shanghai Jiao Tong University · 2 papers
- Huawei · 2 papers
- Alibaba Group · 2 papers
- Hong Kong University of Science and Technology · 2 papers
- University of Science and Technology of China · 2 papers
- Carnegie Mellon University · 2 papers
- National University of Singapore · 2 papers
Use it
Download BibTeX of 39 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
- Efficient network compression and training (229 papers)
- Dataset distillation and data-efficient training (158 papers)
- Neural scaling laws (155 papers)
- Distributed LLM inference systems (147 papers)
- Scalable approximation and equivariant compute (120 papers)
Elsewhere in the topic
- Theory of depth and width (223 papers)
- Representation geometry and collapse (169 papers)
- Layer-wise analysis of language models (145 papers)
- Differentiable optimization layers (125 papers)
- Non-autoregressive translation (102 papers)
- RNN training dynamics and gradients (102 papers)
- Neural-symbolic reasoning (98 papers)
- Sparse dictionary representations (98 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).