Research map / Architectures and efficiency / RNNs and normalization
Backprop-free credit assignment
Scales biologically plausible credit assignment that avoids weight symmetry and exact backpropagation.
44 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in RNNs and normalization, keyword cluster “brain · biological · feedback”
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.
9 papers in 2023–24 and 8 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 1.4% to 0.9%.
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 · Direct Feedback Alignment Provides Learning in Deep Neural Networks (NeurIPS 2016 · 294 citations)
- 2017 · Feedback Networks (CVPR 2017 · 124 citations)
- 2018 · Assessing the Scalability of Biologically-Motivated Deep Learning Algorithms and Architectures (NeurIPS 2018 · 102 citations)
- 2019 · Deep Learning without Weight Transport (NeurIPS 2019 · 70 citations)
- 2020 · Can the Brain Do Backpropagation? --- Exact Implementation of Backpropagation in Predictive Coding Networks (NeurIPS 2020 · 58 citations)
- 2021 · Credit Assignment in Neural Networks through Deep Feedback Control (NeurIPS 2021 · 14 citations)
- 2022 · Single-phase deep learning in cortico-cortical networks (NeurIPS 2022 · 24 citations)
- 2023 · Hebbian Deep Learning Without Feedback (ICLR 2023 · 17 citations)
- 2024 · Understanding and Improving Optimization in Predictive Coding Networks (AAAI 2024 · 5 citations)
- 2025 · $\mu$PC: Scaling Predictive Coding to 100+ Layer Networks (NeurIPS 2025 · 1 citation)
- 2026 · On the Infinite Width and Depth Limits of Predictive Coding Networks (ICML 2026 · 0 citations)
Earliest papers here
- Direct Feedback Alignment Provides Learning in Deep Neural Networks (NeurIPS 2016 · 294 citations)
- Bottom-Up and Top-Down Reasoning With Hierarchical Rectified Gaussians (CVPR 2016 · 95 citations)
- How Important Is Weight Symmetry in Backpropagation? (AAAI 2016 · 92 citations)
What has been tried
36 papers here report 366 measured changes (a component added, removed or swapped, compared with a baseline): helped 52%, no clear effect 15%, hurt 29%, mixed 4%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| add the MQ optimizer to sequential inference | SVHN, CIFAR-10, and Tiny ImageNet classification · test accuracy | Helped | Understanding and Improving Optimization in Predictive Coding Networks (AAAI 2024 · 5 citations) | |
| Replace the standard feedforward structure with a reversible structure | CIFAR-10 · test accuracy | 56.9±0.49 -> 42.8±0.18 | Hurt | Efficient Target Propagation by Deriving Analytical Solution (AAAI 2024 · 2 citations) |
| Use a single inference step instead of as many inference steps as hidden layers | MNIST · performance | Hurt | $\mu$PC: Scaling Predictive Coding to 100+ Layer Networks (NeurIPS 2025 · 1 citation) | |
| Use RFLO with feedback control during learning (RFLO+c) instead of RFLO without control | Visuomotor rotation task · MSE | Helped | Feedback control guides credit assignment in recurrent neural networks (NeurIPS 2024 · 1 citation) | |
| Update Θ rather than keep the backward weights fixed and random | MNIST · accuracy | 97.37% -> 97.57% | Helped | Dendritic Localized Learning: Toward Biologically Plausible Algorithm (ICML 2025 · 0 citations) |
| Set the feedback network learning rate to zero (fixed random feedback weights) | classification tasks shown in the asymmetric-learning-rate ablation · test accuracy | No clear effect | Counter-Current Learning: A Biologically Plausible Dual Network Approach for Deep Learning (NeurIPS 2024 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 5 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- CIFAR-10 · 4 of 5 papers
- MNIST · 2 of 5 papers
- FashionMNIST · 2 of 5 papers
- SVHN · 2 of 5 papers
- CIFAR-100 · 2 of 5 papers
- STL-10 · 2 of 5 papers
Baselines
- backpropagation · 3 of 5 papers
- Feedback Alignment · 2 of 5 papers
- Predictive Coding · 2 of 5 papers
- Equilibrium Propagation · 2 of 5 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).
- Tokyo Institute of Technology · 2 papers
- Denso IT Laboratory · 2 papers
- Google · 2 papers
- Imperial College London · 2 papers
Use it
Download BibTeX of 33 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
- Brain-inspired network comparison (57 papers)
- Hebbian and predictive coding (52 papers)
Elsewhere in the topic
- Efficient network compression and training (229 papers)
- Theory of depth and width (223 papers)
- Representation geometry and collapse (169 papers)
- Dataset distillation and data-efficient training (158 papers)
- Neural scaling laws (155 papers)
- Distributed LLM inference systems (147 papers)
- Layer-wise analysis of language models (145 papers)
- Dataset condensation and flow estimation (126 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).