Research map / Probabilistic and scientific modeling / Molecules and proteins
Neuro-symbolic probabilistic layers
Combines probabilistic structured prediction and exact inference with neural networks.
120 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Molecules and proteins, keyword cluster “latent · space · representations”
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.
28 papers in 2023–24 and 30 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 5.7% to 3.2%.
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 · Collaborative Recurrent Autoencoder: Recommend while Learning to Fill in the Blanks (NeurIPS 2016 · 80 citations)
- 2017 · Hierarchical Multiscale Recurrent Neural Networks (ICLR 2017 · 235 citations)
- 2018 · Fast Decoding in Sequence Models Using Discrete Latent Variables (ICML 2018 · 169 citations)
- 2019 · BridgeNet: A Continuity-Aware Probabilistic Network for Age Estimation (CVPR 2019 · 121 citations)
- 2020 · Data-dependent Gaussian Prior Objective for Language Generation (ICLR 2020 · 53 citations)
- 2021 · DeepWriteSYN: On-Line Handwriting Synthesis via Deep Short-Term Representations (AAAI 2021 · 41 citations)
- 2022 · LasUIE: Unifying Information Extraction with Latent Adaptive Structure-aware Generative Language Model (NeurIPS 2022 · 69 citations)
- 2023 · Neuron Structure Modeling for Generalizable Remote Physiological Measurement (CVPR 2023 · 64 citations)
- 2024 · Neural Markov Random Field for Stereo Matching (CVPR 2024 · 44 citations)
- 2025 · Speech Watermarking with Discrete Intermediate Representations (AAAI 2025 · 4 citations)
- 2026 · Out of the Shadows: Exploring a Latent Space for Neural Network Verification (ICLR 2026 · 1 citation)
Earliest papers here
- Collaborative Recurrent Autoencoder: Recommend while Learning to Fill in the Blanks (NeurIPS 2016 · 80 citations)
- Learning a Hybrid Architecture for Sequence Regression and Annotation (AAAI 2016 · 2 citations)
- Convex Two-Layer Modeling with Latent Structure (NeurIPS 2016 · 1 citation)
What has been tried
94 papers here report 917 measured changes (a component added, removed or swapped, compared with a baseline): helped 66%, no clear effect 8%, hurt 21%, 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 adaptive positional bias | SceneFlow test · Bad1.0 [%] | 5.96 -> 5.80 | Helped | Neural Markov Random Field for Stereo Matching (CVPR 2024 · 44 citations) |
| Local (per-stroke) time modelling instead of global time modelling | sketch reconstruction (qualitative) · visual smoothness | Hurt | SketchINR: A First Look into Sketches as Implicit Neural Representations (CVPR 2024 · 11 citations) | |
| Use signal-CAV instead of regression-based CAV optimizers | Controlled bias experiments · accuracy on biased test set | Helped | From Hope to Safety: Unlearning Biases of Deep Models via Gradient Penalization in Latent Space (AAAI 2024 · 9 citations) | |
| Use transformer-based DR instead of mean DR | EMOPIA · PR | 50.58±11.09 -> 53.47±11.50 | Helped | MusER: Musical Element-Based Regularization for Generating Symbolic Music with Emotion (AAAI 2024 · 8 citations) |
| Add the cascaded CasCast framework with EarthFormer as deterministic backbone | SEVIR, CSI-219 POOL16 · CSI | +91.83% | Helped | CasCast: Skillful High-resolution Precipitation Nowcasting via Cascaded Modelling (ICML 2024 · 7 citations) |
| increase chunk size from 5 to 10 | Marker Writing · MSE | 1.18 -> 0.57 | Helped | Hierarchical State Space Models for Continuous Sequence-to-Sequence Modeling (ICML 2024 · 5 citations) |
| Use discrete rather than continuous intermediate representations | Information hiding · Mean BER (%) | 2.39 -> 0.28 | Helped | Speech Watermarking with Discrete Intermediate Representations (AAAI 2025 · 4 citations) |
| Replace [sin(·), arcsin(·)] metrics with p-norm metrics [x2, x3] | Kodak · PSNR | 41.57 -> 40.63 | Hurt | MetricGrids: Arbitrary Nonlinear Approximation with Elementary Metric Grids based Implicit Neural Representation (CVPR 2025 · 4 citations) |
| replace FSQ with VQ | LIBERO-90 and LIBERO few-shot · success rate | 88.6 ± 0.4 -> 81.2 ± 0.6; 68.8 ± 1.7 -> 62.5 ± 2.0 | Hurt | QueST: Self-Supervised Skill Abstractions for Learning Continuous Control (NeurIPS 2024 · 4 citations) |
| Increase VAE latent dimension | VAE reconstruction across variables at 1.40625° · reconstruction error | 0.96 -> 0.80 -> 0.71 (T2m); 0.65 -> 0.51 -> 0.43 (U10); 0.62 -> 0.48 -> 0.40 (V10); 48.72 -> 35.42 -> 27.34 (Z500); 0.77 | Helped | OmniCast: A Masked Latent Diffusion Model for Weather Forecasting Across Time Scales (NeurIPS 2025 · 1 citation) |
| Replace affine coupling with Mix-1D mixture coupling | OpenWebText · NELBO | Helped | Flexible Language Modeling in Continuous Space with Transformer-based Autoregressive Flows (NeurIPS 2025 · 0 citations) | |
| Replace Energy Distance with Root Mean Squared Error | LibriSpeech-PC test-clean, 3s Prefix as Prompt · WER-C (%) | 1.59 -> 40.60 | Hurt | Efficient Speech Language Modeling via Energy Distance in Continuous Latent Space (NeurIPS 2025 · 0 citations) |
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 58 papers; from OpenAlex and the papers' affiliations).
- University of California, Los Angeles · 3 papers
- Xi’an Jiaotong University · 2 papers
- University of Edinburgh · 2 papers
- Microsoft · 2 papers
- University of Cambridge · 2 papers
- Carnegie Mellon University · 2 papers
- Shanghai Jiao Tong University · 2 papers
- Georgia Institute of Technology · 2 papers
- Harvard 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
- Sparse graphical and tensor models (97 papers)
- Probabilistic embeddings and relational inference (93 papers)
- Equivariant and disentangled latent representations (90 papers)
- Generative latent models for single-cell data (41 papers)
Elsewhere in the topic
- Generative protein structure design (142 papers)
- Prediction-powered and conformal inference (129 papers)
- Statistical analysis of ML practice (122 papers)
- Structure-based 3D molecule generation (107 papers)
- Flow matching generative modeling (97 papers)
- Machine learning for physical simulation (96 papers)
- Multimodal trajectory prediction (95 papers)
- Counterfactual and statistical explanations (81 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).