Giving Simulated Cells a Voice: Evolving Prompt-to-Intervention Models for Cellular Control

Josh Bongard; Michael Levin; Nam H. Le; Patrick Erikson; Yanbo Zhang

REVIEW 4 major objections 6 minor 29 references

Giving Simulated Cells a Voice: Evolving Prompt-to-Intervention Models for Cellular Control

T0 review · 4 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read An evolvable neural network maps plain-language prompts to vector fields that steer simulated cells into clustering or scattering behaviors, verified by a vision-language model acting as an oracle.

desk verdict Honest workshop proof-of-concept: evolved language-to-vector-field control works in simulation, but the VLM is both fitness and judge, so the central claim relies on an unvalidated oracle. read the letter →

arxiv 2505.02766 v1 pith:7MF3HNEF submitted 2025-05-05 cs.AI cs.NEcs.ROq-bio.TO

Nam H. Le , Patrick Erikson , Yanbo Zhang , Michael Levin , Josh Bongard This is my paper · ORCID

classification cs.AIcs.NEcs.ROq-bio.TO

keywords naturallanguagecontrolevolutionarystrategiesgeneticalgorithmsvision-languagemodelcellularcollectivebehaviorvectorfieldinterventionclusteringandscatteringprompt-to-intervention

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper sets out to show that natural language can serve as a control interface for cellular collectives: a user says "cluster" or "scatter," and an evolved neural network turns that word into a spatial vector field that makes simulated cells behave accordingly. The authors build a complete pipeline, ZapGPT, in which a Prompt-to-Intervention (P2I) network generates the intervention, a simulated environment responds, and a vision-language model (D2R) judges whether the outcome matches the prompt. They then use evolutionary strategies to optimize the P2I weights against this language-based fitness signal. The central claim is a feasibility proof: even with a constrained vocabulary and simplified cells, evolved P2I networks can align cellular dynamics with user-defined goals expressed in plain language. A sympathetic reader would care because this offers a blueprint for top-down, language-driven control of complex biological or synthetic systems without hand-coded reward functions.

What carries the argument

The central object is the Prompt-to-Intervention (P2I) network: a feed-forward neural network that takes a frozen BERT embedding of the prompt (768 dimensions) and outputs a flattened vector field, reshaped to an (n, n, 2) grid of force vectors. It is paired with a Dynamics-to-Response (D2R) model—a lightweight vision-language model that reads plots of the simulation and returns a one-word label. The load-bearing mechanism is the evolutionary optimization loop: the P2I weights are mutated and selected according to the average binary agreement between the D2R verdict and the prompt over 30 simulation epochs, with a combined reward that balances the distance-trend plot and the final-position plot. This loop lets the system discover interventions without gradients or hand-crafted rewards, treating the VLM as a differentiable-free fitness oracle.

What would settle it

Compute a confusion matrix for the vision-language oracle against a ground-truth label derived from an independent quantitative measure (e.g., average pairwise distance compared to initial spread, or the number of connected clusters) on held-out simulations that include edge cases like multiple small clusters or partial scattering; if the oracle's accuracy is near chance or systematically biased, the fitness signal is too noisy to validate the central claim. Alternatively, evolve a P2I network to maximize the oracle's score and then check whether the resulting vector field achieves the goal under the independent metric.

Watch

Extended reading notes

Core claim

The paper demonstrates, in a 500-step 2D simulation with 100 cells, that a small feed-forward network can learn to translate the one-word prompts "cluster" and "scatter" into grid vector fields (resolutions 2×2 to 10×10) that produce the requested collective behavior. The fitness signal is not a hand-written metric but the agreement between the prompt and a one-word verdict from a pretrained vision-language model that inspects two plots: the time series of average pairwise distance and the final cell-position overlay. Using a (1+1) evolution strategy for simpler grids and a genetic algorithm with arithmetic crossover for larger grids, the authors report statistically significant improvements in fitness over generations, with the GA notably improving the final-position score for 5×5 and 10×10 fields. The discovery is that this closed loop—language to intervention to behavior to language—can be optimized end-to-end using only the language model as a judge, establishing a proof-of-concept for natural-language-driven cellular control.

Load-bearing premise

The entire optimization rests on the assumption that the vision-language model's one-word classification of the two plots is a faithful and reliable oracle for whether cells are actually clustering or scattering, and that evolutionary search will not exploit quirks or artifacts of that classifier rather than producing genuinely correct behavior.

Editorial extensions

If this is right

If the central claim holds, language-driven control of collective behavior becomes feasible without explicit reward engineering: any goal expressible as a one-word verdict from a vision-language model could, in principle, be optimized in simulation.
The same P2I-D2R loop could be transferred to real biological systems where the vision-language model reads microscopy images instead of synthetic plots, enabling automated discovery of bioelectric or chemical intervention schedules.
The finding that genetic algorithms outperform (1+1)-ES on larger vector fields suggests that crossover helps balance local attraction and global cohesion, informing how to scale such language-guided evolution to higher-dimensional interventions.
The demonstration that linguistic similarity need not track behavioral similarity (e.g., "scattering slowly" vs. "scattering quickly") warns that naive embedding-based evaluation can mislead optimization, motivating richer prompt-conditioned architectures.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural extension, not tested in the paper, is to replace the one-word oracle with a free-form captioner and a separate similarity model; the authors mention this, but a reader might infer that success would hinge on defining a behavioral, rather than purely semantic, similarity metric to avoid the 'form a cluster' vs. 'cells spreading out' failure mode they identify.
If the VLM oracle is faithful, the approach suggests a general recipe for 'steering by description' in any simulated complex system, from robot swarms to morphogenesis models, where the evaluator is a perception model trained on natural images or plots.
The paper's reliance on a future validation link for the VLM oracle raises a testable implication: if the oracle is inaccurate on adversarial or out-of-distribution plots, the evolved policies may exploit classifier artifacts, so one should verify evolved vector fields against an independent clustering metric like the number of connected components.
A reader might speculate that multi-word prompts (e.g., 'cluster slowly') could be learned with prompt-conditioned layers, but the current single-word, single-behavior training may not transfer without explicitly addressing interference, as the paper notes.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The useful thing to know: this is a thin but honest proof-of-concept, not a hoax. The pipeline maps a BERT embedding of "cluster" or "scatter" through a small evolved neural network to a grid vector field, runs 100 cells for 500 steps, then asks Moondream2 to classify the distance plot and final positions as clustering or scattering. The reward is simply whether the VLM's one-word answer matches the prompt. The authors optimize that reward with a (1+1)-ES and a GA and show fitness improves across seeds. That is the whole empirical content.

What is genuinely new is the composition: LLM embeddings plus an evolvable controller plus VLM-based evaluation, all in one closed language-to-intervention-to-behavior loop. I did not find that in the cited prior work, and for a workshop paper that is a fair contribution. The paper also earns credit for candor. Section 5.2 explicitly says the reward is engineered, that semantic similarity may not track behavioral success, and that better evaluators are needed. The average pairwise distance metric in Equation 1 gives some quantitative grounding, and the authors do not pretend the simulator dynamics are derived from the VLM.

The soft spot is the one the stress test names, and it lands. Moondream2 is simultaneously the fitness signal and the success evaluator, so the reported improvements in Figures 5–7 are improvements in VLM agreement, not independently confirmed clustering. The footnote says the authors validated Moondream2 on synthetic plots, but no confusion matrix or code appears in the paper, and no independent ground truth is reported in the main text. Evolutionary search is an adversarial optimizer; it can exploit VLM biases such as axis scaling or line thickness to raise the score without producing a real cluster. The statistical tests only show the optimizer moves the VLM score. There is also no trivial baseline (random or hand-coded vector fields), no sensitivity analysis for alpha and beta, and no shipped code or data. These are fixable, and they do not destroy the proof-of-concept value, but they do mean the central feasibility claim is plausible rather than sealed.

Who is this for? People working on language-conditioned control, bioelectric interfaces, swarm robotics, or neuroevolution will find a compact demonstration and a useful discussion of the semantic-vs-behavioral gap. It deserves a serious referee at GECCO workshop level, with the clear request to validate the oracle, add a baseline, and release code. I would not cite it in my own work until those are in place, but I would not dismiss it either.

Referee Report

4 major / 6 minor

Summary. The paper proposes ZapGPT, a pipeline that translates a natural language prompt (e.g., "cluster") into a spatial vector field via a Prompt-to-Intervention (P2I) neural network; the field is applied to 100 cells in a 2D simulation, and after 500 steps a Dynamics-to-Response (D2R) model (Moondream2, a vision-language model) reads a time-series distance plot and a final-position overlay, returning a one-word label ("clustering" or "scattering"). A binary reward indicates whether the VLM label matches the prompt, and evolutionary algorithms ((1+1)-ES and a genetic algorithm) optimize the P2I weights to maximize this reward over 30 epochs. Experiments across vector-field resolutions n=2,3,5,10 show fitness improvements with reported Wilcoxon significance, and the authors claim that the evolved P2I networks successfully align cellular dynamics with plain-language goals. The paper is presented as a proof-of-concept for language-guided evolutionary control of collective cellular behavior.

Significance. If the central claim is validated, the framework provides a novel and potentially useful language-mediated control loop for simulated collective systems, combining BERT embeddings, an evolvable feed-forward controller, and a vision-language evaluator. The paper has several strengths: the pipeline is clearly described and simple to reproduce; the use of two complementary visual criteria (temporal trend and final spatial configuration) is sensible; optimization progress is reported across seeds with statistical tests; and the authors explicitly discuss limitations and future directions. The main weakness, which is load-bearing, is that the only success measure is the same VLM that provides the fitness signal, and its reliability is not independently established in the main text. Because evolutionary search can exploit systematic biases in a learned classifier, the reported gains currently demonstrate only that evolution can increase agreement with Moondream2's labels, not that the cells genuinely cluster or scatter.

major comments (4)

[§3.5, Eq. (2) and Eq. (3); Figures 5–7] The fitness function used for evolution is identical to the evidence of success: R_i is defined by whether Moondream2's one-word response matches the prompt, and R_combined (Eq. 3) is the only objective optimized. The manuscript does not provide an independent validation of Moondream2's classifications in the main text; the §3.5 footnote defers this to a future GitHub release. Since evolutionary search is an adversarial optimizer, it can exploit systematic VLM biases (e.g., line thickness, axis scaling, legend text, or a learned prior that a downward distance curve means "clustering") to raise R without producing actual clustering. To support the central claim, the authors should report a confusion matrix or agreement rate for Moondream2 on held-out plots with ground-truth labels, and should show that evolved vector fields change quantitative spatial statistics (e.g., final average pairwise distance from Eq. (1), local density, or nearest-neighbor distances) relative to random or initial P2I weights.
[§4.1–4.3, Wilcoxon tests] All reported statistical tests compare the VLM-based fitness at generation 0 with the final generation. They establish that the optimizer increases the reward, not that the reward reflects genuine behavioral success. The p-values in §4.3 (<0.0001) refer to fitness series only, and therefore do not validate the behavioral outcome. The authors should either (a) provide independent ground-truth evaluation of the evolved behaviors, or (b) reframe the claims throughout the abstract and conclusions as "evolution improves agreement with Moondream2's classifications" rather than "cells cluster/scatter." This distinction is essential because the paper's stated contribution depends on the former interpretation.
[§4.3, GA advantage claim] The claim that the genetic algorithm outperforms (1+1)-ES is based on comparing separate experiments with different vector-field sizes and reward criteria, without a matched statistical comparison under the same evaluation budget (e.g., identical number of fitness evaluations, seeds, and reward settings). The sentence "crossover mechanisms here contribute to a more effective search process" is thus not directly supported by the reported data. A matched comparison or an explicit disavowal of the cross-algorithm claim would remove this unsupported implication.
[§5.2, internal concession] The discussion explicitly states that the current strategy "depends on constrained language and engineered reward functions" and that "ensuring that semantic similarity reflects behavioral success" remains an open problem. These concessions are exactly the unsecured link between the reward and the claimed behavior. The abstract's assertion that "evolved P2I networks can successfully align cellular dynamics with user-defined goals expressed in plain language" is therefore stronger than the evidence presented. The authors should either add the missing validation or temper the abstract and introduction to claim only that the pipeline optimizes a VLM-based alignment score in a simulated environment.

minor comments (6)

[Throughout] The terminology for the D2R model is inconsistent: §3.1 uses "Dynamic-to-Response" while §3.5 and elsewhere use "Dynamics-to-Response." Please unify.
[§1] There are several typos in the introduction: "articular" should be "articulate," "desied" should be "desired," and "dynamamics" should be "dynamics."
[§3.5.3] The word "consistint" in the sentence describing an epoch is a typo for "consisting."
[Figures 5–7] The y-axis labels in Figures 5–7 are not defined in the text; please state explicitly that the plotted quantity is R_average or R_combined from Eq. (2)–(3), and describe the error bars in Figure 6.
[Footnote 1 (page 5)] The validation of Moondream2 is deferred to a GitHub repository that is not yet available; the manuscript should state when this code and the evaluation results will be released, or include a supplementary file with the confusion matrix and example classifications.
[References] Reference [28] for Moondream2 lists only the author as "vik" and does not include the full model name or publication venue; please provide a complete citation.

Circularity Check

1 steps flagged · score 6.0 of 10

The reported 'successful alignment' is the evolved fitness score itself: Moondream2's one-word match is both the objective and the evidence, with no independent behavioral metric reported.

self definitional [Section 3.5.3 (Eqs. 2-3) and Section 4.1]
"For each epoch, we assign a binary scoreRi = 1 where: • Ri = 1 if the D2R response matches the input prompt, and • Ri = 0 otherwise. ... Rcombined =αRdistance +βRposition, (3) whereα = 0.5 andβ = 0.5. This combined reward function captures both the temporal dynamics and the final spatial configuration, providing a robust metric for evaluating the alignment between human intent and system behavior. ... These results indicate that, when evaluated solely on the distance plot, (1+1)-ES can effectively optimize the clustering behavior in our simulated environment."

The P2I network is evolved to maximize R, which is defined entirely by whether Moondream2's one-word label matches the prompt (Eqs. 2-3). The paper then reports improved fitness as evidence that evolution 'optimize[s] the clustering behavior.' Because the only optimized objective is the D2R-match score, the success claim reduces by construction to the fitness function itself. No independent measure of clustering (e.g., the Eq. 1 average pairwise distance values) is reported for the evolved policies, and the external validation of Moondream2 is deferred to a future GitHub link in the Section 3.5 footnote. Thus the central demonstration that language goals are 'successfully align[ed]' with cellular dynamics is the optimized score, not an independently verified behavioral outcome.

full rationale

The paper is an empirical proof-of-concept, not a formal derivation, and it does not rely on load-bearing self-citation: references to the authors' earlier work appear only as related background. However, there is a genuine self-referential element in the experimental loop. The fitness function that drives evolution (Section 3.5.3) is identical to the evidence used to claim success: Moondream2's one-word verdict matching the prompt. Figures 5-7 plot this fitness value over generations, so the statistically significant improvements show only that the optimizer increased its own objective. The average pairwise distance metric in Eq. 1 is defined as a quantitative measure of clustering/scattering, but it is fed into Moondream2 as a plot rather than being reported directly as the outcome measure for evolved policies. The authors themselves concede in Section 5.2 that the reward is 'engineered' and that 'ensuring that semantic similarity reflects behavioral success' remains an open problem, which is precisely the unvalidated link. Because the core feasibility claim reduces to the optimized fitness score, the circularity score is 6 rather than lower; the simulation dynamics and the use of an external pretrained VLM provide some independent content, but the main reported evidence is not independent of the optimization objective.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

Most of the paper's machinery is standard software components, but the semantic evaluation rests on an assumed behavioral oracle (Moondream2 labels) and an assumed simplification of biology into 2D vector-field dynamics. The hidden fitted quantities are the P2I network weights and a few hand-set reward and evolutionary algorithm constants. No new physical entities are introduced; P2I and D2R are software modules, and 'bioelectric-like' is a metaphorical label for a vector field.

free parameters (4)

P2I neural network weights = not reported (evolved)
The entire optimization target: feedforward weights mapping BERT embeddings to vector fields. Values are not disclosed.
Combined reward weights alpha, beta = 0.5, 0.5
Hand-chosen weights in Eq. (3) that balance the distance-plot reward and the final-position reward; they shape the evolved behavior.
Evolutionary algorithm hyperparameters = GA: pop_size=20, tournament=8, sigma=0.1, generations=50; ES: p_target=0.2, window_size=5
User-chosen algorithm settings. Results depend on them, but they are not fitted to external data.
Vector-field grid resolution n = 2, 3, 5, 10
Experimental condition varied by the authors; performance is reported separately for each resolution and the higher resolutions motivate the switch from ES to GA.

assumptions (4)

domain assumption Moondream2 one-word classification is a faithful behavioral oracle for clustering and scattering.
The pipeline uses Moondream2 both as fitness signal and as success evidence; validation is deferred to a future GitHub link (§3.5 footnote).
domain assumption The simplified 2D simulation (500x500 box, 100 circular cells, local repulsion, reflective boundaries) adequately represents the relevant collective cell dynamics.
No real cells, bioelectricity, or 3D mechanics are modeled; the authors call it a proof of concept in §3.2.
domain assumption Pretrained BERT embeddings capture the semantics of the one-word prompts well enough for this task.
The P2I model starts from BERT embeddings, but no analysis of embedding quality for this domain is provided in the main text.
domain assumption Evolutionary search over P2I weights reliably optimizes the VLM-based reward.
Statistical tests compare generation 0 with the final generation only; there is no guarantee of convergence or global optimality.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Giving Simulated Cells a Voice: Evolving Prompt-to-Intervention Models for Cellular Control." pith.science (2026). https://pith.science/paper/7MF3HNEF

@misc{pith2026250502766,
  author       = {Pith},
  title        = {Pith review of: Giving Simulated Cells a Voice: Evolving Prompt-to-Intervention Models for Cellular Control},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7MF3HNEF}},
  note         = {Machine review of arXiv:2505.02766}
}

read the original abstract

Guiding biological systems toward desired states, such as morphogenetic outcomes, remains a fundamental challenge with far-reaching implications for medicine and synthetic biology. While large language models (LLMs) have enabled natural language as an interface for interpretable control in AI systems, their use as mediators for steering biological or cellular dynamics remains largely unexplored. In this work, we present a functional pipeline that translates natural language prompts into spatial vector fields capable of directing simulated cellular collectives. Our approach combines a large language model with an evolvable neural controller (Prompt-to-Intervention, or P2I), optimized via evolutionary strategies to generate behaviors such as clustering or scattering in a simulated 2D environment. We demonstrate that even with constrained vocabulary and simplified cell models, evolved P2I networks can successfully align cellular dynamics with user-defined goals expressed in plain language. This work offers a complete loop from language input to simulated bioelectric-like intervention to behavioral output, providing a foundation for future systems capable of natural language-driven cellular control.

Figures

Figures reproduced from arXiv: 2505.02766 by the authors.

**Figure 2.** The Prompt-to-Intervention (P2I) model architecture. By directly generating spatial interventions from natural language, the P2I model establishes a bridge between symbolic instructions and cellular dynamics, forming a foundational component of the ZapGPT framework. 4 [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** D2R input-output example using the average distance plot. The prompt directs the VLM to choose between “clustering” or “scattering” based on the observed trend [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: D2R input-output example using the final cell position plot. The VLM determines whether the cells exhibit clustering or scattering behavior. 3. The intervention is applied to the environment, and the simulation is run for 500 steps, allowing the cellular dynamics to ev…

**Figure 5.** Figure 5: Best fitness scores across generations for different vector field sizes, using (1+1)-ES and average distance metric. 4.2 Evaluating (1+1)-ES with Combined Rewards To better capture both the progression of cellular behavior and the final spatial arrangement, we extended…

**Figure 6.** Figure 6: Scores over generations for different vector field sizes using (1+1)-ES with combined rewards. configuration (position score) remains suboptimal. This suggests that while (1+1)-ES can reduce pairwise distances effectively, it struggles to enforce global cohesion in mor…

**Figure 7.** Figure 7: Fitness over generations using GAs for different vector field sizes Statistical tests comparing the initial and final generations yielded p-values < 0.0001, indicating a significant improvement in both metrics. Although these results are preliminary, they highlight th…

**Figure 8.** Figure 8: Cosine similarity matrix of BERT embeddings for different prompts. Linguistic similarity does not imply behavioral similarity. Training a single P2I model on multiple behaviors—especially those with conflicting goals—can lead to interference, where learning one behavio…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 23 canonical work pages

[1]

Resilient machines through continuous self-modeling

Josh Bongard, Victor Zykov, and Hod Lipson. Resilient machines through continuous self-modeling. Science, 314(5802):1118–1121, 2006

work page 2006
[2]

Evolving cell models for systems and synthetic biology

Hongqing Cao, Francisco J Romero-Campero, Stephan Heeb, Miguel Cámara, and Natalio Krasnogor. Evolving cell models for systems and synthetic biology. Systems and synthetic biology, 4:55–84, 2010

work page 2010
[3]

Morphological coordination: a common ancestral function unifying neural and non-neural signaling

Chris Fields, Johanna Bischof, and Michael Levin. Morphological coordination: a common ancestral function unifying neural and non-neural signaling. Physiology, 35(1):16–30, 2020

work page 2020
[4]

Exploring the behavior of bioelectric circuits using evolution heuristic search

Hananel Hazan and Michael Levin. Exploring the behavior of bioelectric circuits using evolution heuristic search. Bioelectricity, 4(4):207–227, 2022

work page 2022
[5]

Emergence of informative higher scales in biological systems: a computational toolkit for optimal prediction and control

Erik Hoel and Michael Levin. Emergence of informative higher scales in biological systems: a computational toolkit for optimal prediction and control. Communicative & Integrative Biology, 13(1):108–118, 2020

work page 2020
[6]

Can the macro beat the micro? integrated information across spatiotemporal scales

Erik P Hoel, Larissa Albantakis, William Marshall, and Giulio Tononi. Can the macro beat the micro? integrated information across spatiotemporal scales. Neuroscience of Consciousness, 2016(1):niw012, 2016

work page 2016
[7]

Genetic algorithms

John H Holland. Genetic algorithms. Scientific american, 267(1):66–73, 1992

1992
[8]

Prediction of biomedical signals using deep learning techniques

K Kalaivani, Pravin R Kshirsagarr, J Sirisha Devi, Surekha Reddy Bandela, Ilhami Colak, J Nageswara Rao, and A Rajaram. Prediction of biomedical signals using deep learning techniques. Journal of Intelligent & Fuzzy Systems, 44(6):9769–9782, 2023

work page 2023

Show all 29 references

[9]

Chemical communication in artificial cells: basic concepts, design and challenges

Hedi Karoui, Pankaj Singh Patwal, BVVS Pavan Kumar, and Nicolas Martin. Chemical communication in artificial cells: basic concepts, design and challenges. Frontiers in Molecular Biosciences, 9:880525, 2022

2022
[10]

Le, Richard Watson, Mike Levin, and Chrys Buckley

Nam H. Le, Richard Watson, Mike Levin, and Chrys Buckley. Emergent collective reproduction via evolving neuronal flocks, 2024

2024
[11]

Reprogramming cells and tissue patterning via bioelectrical pathways: molecular mechanisms and biomedical opportunities

Michael Levin. Reprogramming cells and tissue patterning via bioelectrical pathways: molecular mechanisms and biomedical opportunities. Wiley Interdisciplinary Reviews: Systems Biology and Medicine, 5(6):657–676, 2013

2013
[12]

The computational boundary of a “self”: developmental bioelectricity drives multicellularity and scale-free cognition

Michael Levin. The computational boundary of a “self”: developmental bioelectricity drives multicellularity and scale-free cognition. Frontiers in psychology, 10:2688, 2019

2019
[13]

Bioelectric signaling: Reprogrammable circuits underlying embryogenesis, regeneration, and cancer

Michael Levin. Bioelectric signaling: Reprogrammable circuits underlying embryogenesis, regeneration, and cancer. Cell, 184(8):1971–1989, 2021

1971
[14]

Collective intelligence of morphogenesis as a teleonomic process

Michael Levin. Collective intelligence of morphogenesis as a teleonomic process. 2022

2022
[15]

Llava-next: Improved reasoning, ocr, and world knowledge, January 2024

Haotian Liu, Chunyuan Li, Yuheng Li, Bo Li, Yuanhan Zhang, Sheng Shen, and Yong Jae Lee. Llava-next: Improved reasoning, ocr, and world knowledge, January 2024

2024
[16]

Deep learning for cellular image analysis

Erick Moen, Dylan Bannon, Takamasa Kudo, William Graf, Markus Covert, and David Van Valen. Deep learning for cellular image analysis. Nature methods, 16(12):1233–1246, 2019

2019
[17]

Gpt-4 technical report, 2024

OpenAI. Gpt-4 technical report, 2024

2024
[18]

A derandomized approach to self-adaptation of evolution strategies

Andreas Ostermeier, Andreas Gawelczyk, and Nikolaus Hansen. A derandomized approach to self-adaptation of evolution strategies. Evolutionary Computation, 2(4):369–380, December 1994

1994
[19]

Top-down models in biology: explanation and control of complex living systems above the molecular level

Giovanni Pezzulo and Michael Levin. Top-down models in biology: explanation and control of complex living systems above the molecular level. Journal of The Royal Society Interface, 13(124):20160555, 2016

2016
[20]

Learning transferable visual models from natural language supervision

Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning, pa...

2021
[21]

An evolutionary approach to synthetic biology: Zen and the art of creating life

Thomas S Ray. An evolutionary approach to synthetic biology: Zen and the art of creating life. Artificial Life, 1(1_2):179–209, 1993

1993
[22]

Evolution strategies as a scalable alternative to reinforcement learning

Tim Salimans, Jonathan Ho, Xi Chen, Szymon Sidor, and Ilya Sutskever. Evolution strategies as a scalable alternative to reinforcement learning. arXiv preprint arXiv:1703.03864, 2017

2017 arXiv
[23]

Swarm robotic behaviors and current applications

Melanie Schranz, Martina Umlauft, Micha Sende, and Wilfried Elmenreich. Swarm robotic behaviors and current applications. Frontiers in Robotics and AI, 7:36, 2020. 13 A PREPRINT - M AY 6, 2025

2020
[24]

Designing neural networks through neuroevolution

Kenneth O Stanley, Jeff Clune, Joel Lehman, and Risto Miikkulainen. Designing neural networks through neuroevolution. Nature Machine Intelligence, 1(1):24–35, 2019

2019
[25]

Ultrasonic bioreactor as a platform for studying cellular response

Anuradha Subramanian, Joseph A Turner, Gaurav Budhiraja, Sanjukta Guha Thakurta, Nicholas P Whitney, and Sai Siddhartha Nudurupati. Ultrasonic bioreactor as a platform for studying cellular response. Tissue Engineering Part C: Methods, 19(3):244–255, 2013

2013
[26]

Surpassing cosine similarity for multidimensional comparisons: Dimension insensitive euclidean metric (diem)

Federico Tessari and Neville Hogan. Surpassing cosine similarity for multidimensional comparisons: Dimension insensitive euclidean metric (diem). arXiv preprint arXiv:2407.08623, 2024

2024 arXiv
[27]

Evolving aggregation behaviors in a swarm of robots

Vito Trianni, Roderich Groß, Thomas H Labella, Erol ¸ Sahin, and Marco Dorigo. Evolving aggregation behaviors in a swarm of robots. In Advances in Artificial Life: 7th European Conference, ECAL 2003, Dortmund, Germany, September 14-17, 2003. Proceedings 7, pages 865–874. Sprin...

2003
[28]

moondream2 (revision 92d3d73), 2024

vik. moondream2 (revision 92d3d73), 2024

2024
[29]

Distance-based protein folding powered by deep learning

Jinbo Xu. Distance-based protein folding powered by deep learning. Proceedings of the National Academy of Sciences, 116(34):16856–16865, 2019. 14

2019

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [1]

Resilient machines through continuous self-modeling

Josh Bongard, Victor Zykov, and Hod Lipson. Resilient machines through continuous self-modeling. Science, 314(5802):1118–1121, 2006

work page 2006

[2] [2]

Evolving cell models for systems and synthetic biology

Hongqing Cao, Francisco J Romero-Campero, Stephan Heeb, Miguel Cámara, and Natalio Krasnogor. Evolving cell models for systems and synthetic biology. Systems and synthetic biology, 4:55–84, 2010

work page 2010

[3] [3]

Morphological coordination: a common ancestral function unifying neural and non-neural signaling

Chris Fields, Johanna Bischof, and Michael Levin. Morphological coordination: a common ancestral function unifying neural and non-neural signaling. Physiology, 35(1):16–30, 2020

work page 2020

[4] [4]

Exploring the behavior of bioelectric circuits using evolution heuristic search

Hananel Hazan and Michael Levin. Exploring the behavior of bioelectric circuits using evolution heuristic search. Bioelectricity, 4(4):207–227, 2022

work page 2022

[5] [5]

Emergence of informative higher scales in biological systems: a computational toolkit for optimal prediction and control

Erik Hoel and Michael Levin. Emergence of informative higher scales in biological systems: a computational toolkit for optimal prediction and control. Communicative & Integrative Biology, 13(1):108–118, 2020

work page 2020

[6] [6]

Can the macro beat the micro? integrated information across spatiotemporal scales

Erik P Hoel, Larissa Albantakis, William Marshall, and Giulio Tononi. Can the macro beat the micro? integrated information across spatiotemporal scales. Neuroscience of Consciousness, 2016(1):niw012, 2016

work page 2016

[7] [7]

Genetic algorithms

John H Holland. Genetic algorithms. Scientific american, 267(1):66–73, 1992

1992

[8] [8]

Prediction of biomedical signals using deep learning techniques

K Kalaivani, Pravin R Kshirsagarr, J Sirisha Devi, Surekha Reddy Bandela, Ilhami Colak, J Nageswara Rao, and A Rajaram. Prediction of biomedical signals using deep learning techniques. Journal of Intelligent & Fuzzy Systems, 44(6):9769–9782, 2023

work page 2023