OSCR

Pretraining of Embodied Recurrent Networks Bridges the Gap Between Artificial and Cortical Neural Activities.

Code ↔ Paper

8 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 8 matches · 3 of them tie a paragraph to a whole file, not to given lines: weak matches, whose lines are not tinted
  1. [1] § 3. Results › 3.2. The SR Model Shows Enhanced Alignment with Biological Neural Geometry and Dynamics ↔ analysis/run_model_comparison.py, the whole file · a weak match · score 0.77 · Kruskal Wallis, SR Dense, model comparison, DR Multi, CO Single, Friedman
  2. [2] § 3. Results › 3.2. The SR Model Shows Enhanced Alignment with Biological Neural Geometry and Dynamics ↔ analysis/run_stats_dsa.py, the whole file · a weak match · score 0.63 · SR Dense, DR Multi, CO Single, pairwise, Moderate, ranking
  3. [3] § 3. Results › 3.2. The SR Model Shows Enhanced Alignment with Biological Neural Geometry and Dynamics ↔ figures/fig7.py, lines 148–187 · score 0.60 · CO Multi model, SR model, CO Single, dense, CC, AUC
  4. [4] § 2. Materials and Methods › 2.1. Model Architecture › 2.1.2. Musculoskeletal Arm Model ↔ training/train_dr.py, lines 177–252 · score 0.60 · RigidTendonHillMuscle, RigidTendonArm26, nn, proprioceptive, class, effector
  5. [5] § 2. Materials and Methods › 2.2. Task Representations and Training Procedure › 2.2.4. Training Procedure ↔ tasks/grid_reach.py, lines 75–161 · score 0.59 · 100–300 ms, catch trial, uniformly, GO, 100 ms, joint
  6. [6] § 3. Results › 3.1. Behavioral Performance and Neural Activity of the Single-Reach Model ↔ analysis/eval_BasicModel.py, lines 20–54 · score 0.59 · 0.05–0.35, reaching distance, joint space, Cartesian, workspace, intervals
  7. [7] § 2. Materials and Methods › 2.1. Model Architecture › 2.1.2. Musculoskeletal Arm Model ↔ scripts/pre_data.py, the whole file · a weak match · score 0.58 · RigidTendonHillMuscle, RigidTendonArm26, pre, angles, class, effector
  8. [8] § 3. Results › 3.1. Behavioral Performance and Neural Activity of the Single-Reach Model ↔ model/build.py, lines 83–132 · score 0.57 · angle intervals, reaching distance, joint space, Cartesian, workspace, model

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

Python · 40 lines · 1.5 KB · no license · 1 match

  1. import numpy as np
  2. from itertools import combinations
  3. from scipy.stats import friedmanchisquare, wilcoxon, kruskal
  6. models = ["SR_Dense", "SR_Moderate", "SR_Sparse",
  7. "CO_Single", "CO_Multi", "DR_Single", "DR_Multi"]
  9. model_idx = {label: i for i, label in enumerate(
  10. ["MC_agg", "MM_agg", "SR_Dense", "SR_Moderate", "SR_Sparse",
  11. "CO_Single", "CO_Multi", "DR_Single", "DR_Multi"])}
  12. file_name = "Data/CO/"
  13. for fname, monkey in [("DSA_C.npy", "Monkey C"), ("DSA_M.npy", "Monkey M")]:
  14. arr = np.load(file_name+fname)
  15. print(f"\n{monkey}")
  17. data = np.array([arr[model_idx[m]] for m in models])
  18. stat, p = friedmanchisquare(*data)
  19. print(f"Friedman: chi2={stat:.3f}, p={p:.6f}")
  21. pairs = []
  22. for a, b in combinations(range(7), 2):
  23. w, p_w = wilcoxon(data[a], data[b])
  24. pairs.append({"m1": models[a], "m2": models[b],
  25. "diff": float(np.mean(data[a] - data[b])), "p": p_w})
  26. pairs.sort(key=lambda x: x["p"])
  27. for r, pr in enumerate(pairs):
  28. pr["adj"] = 0.05 / (len(pairs) - r)
  29. pr["reject"] = pr["p"] < pr["adj"]
  31. sig = [p for p in pairs if p["reject"]]
  32. print(f" Sig: {len(sig)}/{len(pairs)}")
  33. for pr in sig:
  34. print(f" {pr['m1']:15s} vs {pr['m2']:15s} diff={pr['diff']:+8.6f} p={pr['p']:.6f}")
  36. sr = data[[0, 1, 2]].ravel()
  37. co = data[[3, 4]].ravel()
  38. dr = data[[5, 6]].ravel()
  39. h, p_k = kruskal(sr, co, dr)
  40. print(f" Kruskal-Wallis (SR vs CO vs DR): H={h:.3f}, p={p_k:.6f}")

run_model_comparison.py at commit 7a85393, no license · at the source

Overview

Authors: Xiangdong Bu1, Hongru Jiang1, Tianruo Guo2, Heng Li3, Yao Chen1
ORCID iDs: Hongru Jiang, Heng Li, Yao Chen
  1. School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China; (X.B.); (H.J.)
  2. Graduate School of Biomedical Engineering, University of New South Wales, Sydney, NSW 2052, Australia
  3. Department of Physical Education, Shanghai Jiao Tong University, Shanghai 200240, China
Institutions: Shanghai Jiao Tong University (China); UNSW Sydney (Australia)
Journal: Biomimetics (Basel, Switzerland), volume 11, issue 8, article 569
Dates: received 8 June 2026; accepted 3 August 2026; published online 9 August 2026
Type: Research article · Language: English
License: CC BY
Identifiers: DOI 10.3390/biomimetics11080569 · PMID 42645286 · PMCID PMC13510150 · OpenAlex W7202077701
Open access: gold, a free copy (OpenAlex)
Status: code verified
Categories: systems (subfield)
Methods: Connectivity, Statistics, Smoothing, state filtering, decompositions, Machine learning, Single-unit activity, calcium imaging
Keywords: motor control, motor primitive, recurrent neural network, musculoskeletal arm model, neural population dynamics
Topic: Motor Control and Adaptation (Cognitive Neuroscience, Neuroscience), according to OpenAlex
Funding: STI 2030-Major Projects (2022ZD0208604); National Natural Science Foundation of China (62176151); Natural Science Foundation of Shanghai (25ZR1401181)
Citations: not cited yet (Europe PMC); 99 references in the paper

Abstract

Task-driven recurrent neural networks (RNNs) have been widely employed as tools for investigating neural dynamics in neural motor control research by modeling the motor cortex. RNNs are often implicitly assumed to learn the underlying computational mechanisms in accordance with biological neural circuits. However, the brain network has a highly structured and specific network connectivity and during individual development the motor cortex has acquired a rich repertoire of behavioral primitives via continuous learning of body control. Considering that the task-driven RNNs are often initialized randomly and trained directly on the specific task, how much these models can truly reveal about the motor cortex is still a crucial question awaiting further research. In this study, we propose a method for modeling the motor cortex pretrained on single reaching skills. Specifically, we use an RNN, receiving sensory feedback and task inputs, as the controller to produce motor commands that drive a musculoskeletal arm model. This model can perform reaching movements along a mini-jerk trajectory between arbitrary points in the workspace, prior to training on specific tasks. The model pretrained on single-reach task has more similarity with real neural data both on a neural geometry and neural dynamics level in center-out (CO) and random target touch (RTT) tasks than models directly trained on these tasks. Surprisingly, we observed the opposite pattern in a double-reach (DR) task, in which two targets appeared simultaneously, rather than presenting the next target after the completion of the prior movement as in the RTT task. This suggests that sequential movements are planned as an integrated unit, and this capability may be implemented at the level of motor cortical circuits. In summary, our results suggest that endowing the network with capabilities beyond the immediate task demands—through more systematic training or other methods—can help better understand the dynamics of biological neural circuits.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 8 matches between paragraphs and lines of code.

RevuBu/Pretrained_RNN

License: none: the authors keep all their rights
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Commit: 7a853932e22cc30a1f38185ec562f5f2058d0a49, 31 July 2026
Languages: Python (34)
Size: 128 files, 34 scripts
Software Heritage: not archived
Found in: “Data Availability Statement”
Holds: README, environment (requirements.txt, setup.py)
Not found: license file, CITATION.cff, tests, continuous integration, documentation
Tools: NumPy (25 files), PyTorch (18 files), Matplotlib (9 files), scikit-learn (7 files), SciPy (6 files)
Availability: 1 check, the latest on 27 September 2026: the link answers
  • 27 September 2026: the link answers
35 files
At the source: github.com/RevuBu/Pretrained_RNN

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

  • 1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 34 scripts, each with its path and the digest of its content;
  • 8 matches between paragraphs of the paper and lines of the code (method lexical-v1);
  • neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

Datasets cited

Data Availability Statement

The original contributions presented in this study are included in the article. The custom code used for this study will be made publicly available on GitHub (https://github.com/RevuBu/Pretrained_RNN (accessed on 31 July 2026)) upon publication.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 5 authors, 5 keywords, 3 funders, 92 references.

Cite

This paper

Bu, X., Jiang, H., Guo, T., Li, H., & Chen, Y. (2026). Pretraining of Embodied Recurrent Networks Bridges the Gap Between Artificial and Cortical Neural Activities. Biomimetics (Basel, Switzerland), 11(8), 569. https://doi.org/10.3390/biomimetics11080569

BibTeX

@article{bu2026pretraining,
author = {Bu, Xiangdong and Jiang, Hongru and Guo, Tianruo and Li, Heng and Chen, Yao},
title = {{Pretraining of Embodied Recurrent Networks Bridges the Gap Between Artificial and Cortical Neural Activities}},
journal = {Biomimetics (Basel, Switzerland)},
year = {2026},
month = aug,
volume = {11},
number = {8},
pages = {569},
publisher = {Multidisciplinary Digital Publishing Institute (MDPI)},
issn = {2313-7673},
doi = {10.3390/biomimetics11080569},
url = {https://doi.org/10.3390/biomimetics11080569},
pmid = {42645286},
pmcid = {PMC13510150}
}

RIS

TY - JOUR
AU - Bu, Xiangdong
AU - Jiang, Hongru
AU - Guo, Tianruo
AU - Li, Heng
AU - Chen, Yao
TI - Pretraining of Embodied Recurrent Networks Bridges the Gap Between Artificial and Cortical Neural Activities
T2 - Biomimetics (Basel, Switzerland)
J2 - Biomimetics (Basel)
PY - 2026
DA - 2026/08/09
VL - 11
IS - 8
SP - 569
SN - 2313-7673
PB - Multidisciplinary Digital Publishing Institute (MDPI)
DO - 10.3390/biomimetics11080569
UR - https://doi.org/10.3390/biomimetics11080569
LA - en
ER -

CSL-JSON

{
"id": "10.3390/biomimetics11080569",
"type": "article-journal",
"title": "Pretraining of Embodied Recurrent Networks Bridges the Gap Between Artificial and Cortical Neural Activities",
"container-title": "Biomimetics (Basel, Switzerland)",
"author": [
{
"family": "Bu",
"given": "Xiangdong"
},
{
"family": "Jiang",
"given": "Hongru"
},
{
"family": "Guo",
"given": "Tianruo"
},
{
"family": "Li",
"given": "Heng"
},
{
"family": "Chen",
"given": "Yao"
}
],
"container-title-short": "Biomimetics (Basel)",
"volume": "11",
"issue": "8",
"page": "569",
"DOI": "10.3390/biomimetics11080569",
"PMID": "42645286",
"PMCID": "PMC13510150",
"ISSN": "2313-7673",
"publisher": "Multidisciplinary Digital Publishing Institute (MDPI)",
"URL": "https://doi.org/10.3390/biomimetics11080569",
"language": "en",
"issued": {
"date-parts": [
[
2026,
8,
9
]
]
}
}

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1016/j.neuron.2026.07.016 [code]
Inferring brain-wide interactions using data-constrained recurrent neural network models.
Journal: Neuron
In common: Matplotlib, NumPy, systems, 10 references
[2] doi:10.1038/s41467-026-75924-7 [code]
Data-driven reduced modeling of neural dynamics.
Journal: Nature communications
In common: scikit-learn, SciPy, Matplotlib, 1 other tool, 9 references
[3] doi:10.1371/journal.pcbi.1014162 [code]
Exploring neural manifolds across a wide range of intrinsic dimensions.
Journal: PLoS computational biology
In common: scikit-learn, SciPy, Matplotlib, 1 other tool, 8 references
[4] doi:10.7554/elife.107423 [code]
A context-free model of savings in motor learning.
Journal: eLife
In common: PyTorch, scikit-learn, SciPy, 2 other tools, 6 references
[5] doi:10.7554/elife.111876 [code]
Distinct sensorimotor encoding in tuft dendrites and somata associated with action, correction, and learning.
Journal: eLife
In common: PyTorch, scikit-learn, SciPy, 2 other tools, 5 references
[6] doi:10.1371/journal.pbio.3003831 [code]
Disinhibitory signaling enables flexible coding of top-down information in cortical networks.
Journal: PLoS biology
In common: PyTorch, scikit-learn, SciPy, 2 other tools, systems, 5 references
[7] doi:10.1038/s41467-026-75455-1 [code]
Shared latent representations of speech production for cross-patient speech decoding.
Journal: Nature communications
In common: PyTorch, scikit-learn, SciPy, 2 other tools, 5 references
[8] doi:10.1371/journal.pcbi.1014585 [code]
Decoding behavior with minimal and interpretable agent models.
Journal: PLoS computational biology
In common: PyTorch, SciPy, Matplotlib, 1 other tool, 5 references
[9] doi:10.1038/s41586-026-10528-1 [code]
A critical initialization for biological neural networks.
Journal: Nature
In common: PyTorch, scikit-learn, SciPy, 2 other tools, 4 references
[10] doi:10.1016/j.isci.2026.117492 [code]
Neural subspace reorganization reflects value-based decision-making.
Journal: iScience
In common: scikit-learn, SciPy, Matplotlib, 1 other tool, 5 references

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.