Research map / Probabilistic and scientific modeling / Dynamical systems
Latent models of neural activity
Learns identifiable, behavior-aligned latent models of neural population activity.
70 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Dynamical systems, keyword cluster “brain · activity · neurons”
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.
20 papers in 2023–24 and 18 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 8.9% to 4.7%.
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 · Structure Inference Machines: Recurrent Neural Networks for Analyzing Relations in Group Activity Recognition (CVPR 2016 · 267 citations)
- 2018 · Transferring Decomposed Tensors for Scalable Energy Breakdown Across Regions (AAAI 2018 · 24 citations)
- 2019 · Universality and individuality in neural dynamics across large populations of recurrent networks (NeurIPS 2019 · 38 citations)
- 2020 · Learning identifiable and interpretable latent models of high-dimensional neural activity using pi-VAE (NeurIPS 2020 · 49 citations)
- 2021 · Neural Latents Benchmark ‘21: Evaluating latent variable models of neural population activity (NeurIPS 2021 · 44 citations)
- 2022 · No Free Lunch from Deep Learning in Neuroscience: A Case Study through Models of the Entorhinal-Hippocampal Circuit (NeurIPS 2022 · 56 citations)
- 2023 · Cognitive Model Discovery via Disentangled RNNs (NeurIPS 2023 · 27 citations)
- 2024 · Partial observation can induce mechanistic mismatches in data-constrained models of neural dynamics (NeurIPS 2024 · 7 citations)
- 2025 · NOBLE - Neural Operator with Biologically-informed Latent Embeddings to Capture Experimental Variability in Biological Neuron Models (NeurIPS 2025 · 1 citation)
- 2026 · Learning Biophysical Models of Large-Scale Multineuronal Data To Enable Precise Neurostimulation (ICML 2026 · 0 citations)
Earliest papers here
- Structure Inference Machines: Recurrent Neural Networks for Analyzing Relations in Group Activity Recognition (CVPR 2016 · 267 citations)
- Transferring Decomposed Tensors for Scalable Energy Breakdown Across Regions (AAAI 2018 · 24 citations)
- Model-based targeted dimensionality reduction for neuronal population data (NeurIPS 2018 · 22 citations)
What has been tried
58 papers here report 500 measured changes (a component added, removed or swapped, compared with a baseline): helped 61%, no clear effect 15%, 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 |
|---|---|---|---|---|
| Add automatic AlignPost synapse merging | 30-area spiking neural network · memory usage | Helped | A differentiable brain simulator bridging brain simulation and brain-inspired computing (ICLR 2024 · 4 citations) | |
| Add accelerometer input | WildPPG · MAE | 8.62±0.06 -> 8.22±0.09 | Helped | WildPPG: A Real-World PPG Dataset of Long Continuous Recordings (NeurIPS 2024 · 4 citations) |
| Increase the smoothing window from w=1 to w=3 | CMU-CTIS, California, Texas, New York, and Pennsylvania (accum=7) · MAE | California: 0.0292 -> 0.0811; Texas: 0.0270 -> 0.0782; New York: 0.0299 -> 0.0886; Pennsylvania: 0.0263 -> 0.0691 | Hurt | Nowcasting Temporal Trends Using Indirect Surveys (AAAI 2024 · 3 citations) |
| Add an L1 penalty to HG | Synthetic dataset · LL | -5.49 -> 9.14 | Helped | One-hot Generalized Linear Model for Switching Brain State Discovery (ICLR 2024 · 2 citations) |
| Add both I_thr and s_thr biologically informed embeddings | PVALB test set, HoF train models · relative L2 error (voltage) | 2.73% -> 1.92% (I_thr alone); 2.83% -> 1.92% (s_thr alone) | Helped | NOBLE - Neural Operator with Biologically-informed Latent Embeddings to Capture Experimental Variability in Biological Neuron Models (NeurIPS 2025 · 1 citation) |
| Use chirp responses only; omit direction selectivity and soma size | Mouse RGC type classification, test · accuracy | 79% to 49% | Hurt | Most discriminative stimuli for functional cell type clustering (ICLR 2024 · 1 citation) |
| Increase number of discrete states K from 4 to 5 (vs 6) for reaching | Behavioral-state variability across conditions · variability of recovered states | 0.098 (K=5) vs 0.384 (K=4), 0.282 (K=6) | Helped | Inference of Neural Dynamics Using Switching Recurrent Neural Networks (NeurIPS 2024 · 1 citation) |
| Use a diagonal covariance approximation instead of the full covariance | Single-branch neuron · conductance recovery | Hurt | Identifying multi-compartment Hodgkin-Huxley models with high-density extracellular voltage recordings (NeurIPS 2025 · 0 citations) | |
| Add pre-stimulation state information to the uStim pattern-only response model | Monkey PFC uStim, 6 sessions · post-uStim state prediction performance | p<0.05 for all four models | Helped | OMiSO: Adaptive optimization of state-dependent brain stimulation to shape neural population states (NeurIPS 2025 · 0 citations) |
| Predict latent feature vectors from cortical positions instead of freely learning them | SENSORIUM mice 1–5, conditioned on half of neurons · Log-likelihood | -0.34 -> -0.69 | Hurt | Modeling Dynamic Neural Activity by combining Naturalistic Video Stimuli and Stimulus-independent Latent Factors (NeurIPS 2025 · 0 citations) |
| Silence the kick-neurons | Two-state HMM · state transitions | Hurt | Mechanistic Interpretability of RNNs emulating Hidden Markov Models (NeurIPS 2025 · 0 citations) | |
| Add negative samples to the contrastive loss | Natural scenes, Mouse 1 · decoding score (%) | 21.85\(±\)1.86 -> 50.86\(±\)0.81 | Helped | Time-Evolving Dynamical System for Learning Latent Representations of Mouse Visual Neural Activity (NeurIPS 2025 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 14 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Baselines
- CEBRA · 3 of 14 papers
- LSTM · 2 of 14 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 38 papers; from OpenAlex and the papers' affiliations).
- University College London · 4 papers
- ETHZ - ETH Zurich · 4 papers
- Baylor College of Medicine · 3 papers
- Stanford University · 3 papers
- University of Cambridge · 3 papers
- Yale University · 2 papers
- University of Oxford · 2 papers
- University of Tuebingen · 2 papers
- Georg-August Universität Göttingen · 2 papers
- Peking University · 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
- Recurrent circuit models of inference (70 papers)
- Low-rank and diffusion neural dynamics (53 papers)
- Functional connectivity modeling (50 papers)
- Synaptic plasticity learning rules (72 papers)
Elsewhere in the topic
- Physics-informed dynamics learning (127 papers)
- Latent dynamics models (111 papers)
- State-space inference for stochastic systems (104 papers)
- Switching and law-discovery dynamics (81 papers)
- Latent-space control and planning (78 papers)
- Data assimilation from sparse observations (58 papers)
- Stochastic population dynamics (53 papers)
- Stable contractive dynamics (42 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).