pith. sign in

arxiv: 2503.14333 · v4 · submitted 2025-03-18 · 💻 cs.LG · cs.AI· q-bio.NC

Characterizing higher-order representations through generative diffusion models explains human decoded neurofeedback performance

Hojjat Azimi Asrari , Megan A. K. Peters This is my paper

Pith reviewed 2026-05-22 23:45 UTC · model grok-4.3

classification 💻 cs.LG cs.AIq-bio.NC
keywords decoded neurofeedbackhigher-order representationsdiffusion modelsreinforcement learningfMRIrepresentational uncertaintyneural representationsindividual differences
0
0 comments X

The pith

A reinforcement-learned diffusion model captures how humans minimize representational uncertainty in decoded neurofeedback.

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

The paper introduces the NERD model to test whether people succeed in decoded neurofeedback by learning to reduce their own uncertainty about neural states. NERD trains denoising diffusion models with reinforcement learning to estimate noise distributions in fMRI data from the task. It outperforms backpropagation-trained models, with added power from clustering those noise distributions. The results also link individual differences in expected uncertainty to task success.

Core claim

Participants accomplish the decoded neurofeedback task by learning about and then minimizing their own representational uncertainty. The NERD model trains denoising diffusion models via reinforcement learning to infer distributions of noise in fMRI data and mirrors brain-like unsupervised learning. This approach outperforms backpropagation-trained control models, with explanatory power increased by clustering the learned noise distributions, and it identifies individual differences in expected-uncertainty representations that predict task success.

What carries the argument

NERD (Noise Estimation through Reinforcement-based Diffusion), a framework that trains denoising diffusion models via reinforcement learning to infer noise distributions in fMRI data from neurofeedback tasks.

If this is right

  • NERD captures human performance in the neurofeedback task better than backpropagation-trained control models.
  • Clustering the learned noise distributions increases the explanatory power of the model.
  • Individual differences in expected-uncertainty representations predict success in the decoded neurofeedback task.
  • The model provides a tool for probing higher-order neural representations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar reinforcement-based diffusion training might model uncertainty monitoring in other cognitive tasks.
  • The results suggest generative models can approximate unsupervised mechanisms the brain uses for metacognition.
  • This framework could be tested on whether it predicts performance in adaptive behaviors outside the lab.

Load-bearing premise

Participants accomplish the decoded neurofeedback task by learning about and minimizing their own representational uncertainty, and the NERD model mirrors brain-like unsupervised learning enough to explain observed performance.

What would settle it

A backpropagation-trained model matching or exceeding NERD in explaining human performance, or no observed correlation between individual expected-uncertainty representations and task success.

Figures

Figures reproduced from arXiv: 2503.14333 by Hojjat Azimi Asrari, Megan A. K. Peters.

Figure 1
Figure 1. Figure 1: Model learning progress. (A) NERD and (B) control-diffusion models show decreased loss (top) and increased reward (bottom) across training epochs. the human subjects. Thus, we determined the best-fitting model for each participant by minimizing Negative Log Likelihood (NLL, Eq. 10) between model-predicted and human voxel patterns during DecNef induction trials (see Methods). Both model families achieved si… view at source ↗
Figure 2
Figure 2. Figure 2: Model fitting results. (A) Both NERD and control-diffusion models predicted humans’ multivoxel patterns equally well: min(NLL) distributions show no difference between models (t(23) = 0.63, p = 0.537). (B) NERD models reached min(NLL) at later epochs (e ∗ NERD = 145.6 ± 93.2; e ∗ control-diffusion = 47.0 ± 28.0). (C,D) Linear models revealed that NERD models (R 2 = 0.782) out-performed than control-diffusi… view at source ↗
Figure 3
Figure 3. Figure 3: Reward trajectories. (A) NERD models show gradual reward increase as denoising progresses. (B) Control-diffusion models achieve maximal reward nearly immediately. Solid lines show mean reward; shaded areas show standard deviation. We then used pairwise pattern representational similarity analysis (Pearson correlation, Eq. 12) to re￾veal further model differences ( [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Pairwise pattern similarities for denoising steps and trials. (A,C) NERD; (B,D) control-diffusion. (A,B) Stepwise RDMs for 5 representative participants (8 representative trials each) show gradual (A) vs. abrupt (B) denoising for the NERD and control-diffusion models, respectively. (C,D) Trial-pair RDMs show heterogeneity in both model families, converging only at the end of denoising. However, only contro… view at source ↗
Figure 5
Figure 5. Figure 5: Multidimensional scaling visualization of activity patterns across denoising. (A) NERD models show gradual convergence in similarity space. (B) Control-diffusion models form a tight cluster immediately after initial denoising steps. 8 [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Noise distributions and state-space analysis. (A) [µ, σ] estimates (raw: left column; normalized: middle column; clustered: right column) for three representative participants. NERD models show heterogeneous, non-monotonic µ and decreasing σ; control-diffusion models show monotonic µ and mixed σ. (B) PCA trajectories reveal participant clusters through noise-distribution higher order representation space w… view at source ↗
Figure 7
Figure 7. Figure 7: Policy network architecture and closed-loop training. (A) The network has an input layer (state dimension + 1), a 128-node hidden layer, and an output layer (2 × state size) estimating µ and σ for p(µ, σ|t, xt). (B) Brain state (TR) is input, denoised, and passed to the decoder (Eq. 1) for feedback, leading to updates to the parameters θ of the policy network. 4.4 Evaluation Metrics 4.4.1 Model Learning As… view at source ↗
read the original abstract

Brains construct not only "first-order" representations of the environment but also "higher-order" representations about those representations -- including higher-order uncertainty estimates that guide learning and adaptive behavior. Higher-order expectations about representational uncertainty -- i.e., learned through experience -- may play a key role in guiding behavior and learning, but their characterization remains empirically and theoretically challenging. Here, we introduce the Noise Estimation through Reinforcement-based Diffusion (NERD) model, a novel computational framework that trains denoising diffusion models via reinforcement learning to infer distributions of noise in functional MRI data from a decoded neurofeedback task, where healthy human participants learn to achieve target neural states. We hypothesize that participants accomplish this task by learning about and then minimizing their own representational uncertainty. We test this hypothesis with NERD, which mirrors brain-like unsupervised learning. Our results show that NERD outperforms backpropagation-trained control models in capturing human performance with explanatory power enhanced by clustering learned noise distributions. Importantly, our results also reveal individual differences in expected-uncertainty representations that predict task success, demonstrating NERD's utility as a powerful tool for probing higher-order neural representations.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript introduces the Noise Estimation through Reinforcement-based Diffusion (NERD) model, which trains denoising diffusion models via reinforcement learning to infer distributions of noise in fMRI data from a decoded neurofeedback task. It hypothesizes that participants accomplish the task by learning about and minimizing their own representational uncertainty, with NERD mirroring brain-like unsupervised learning. The central claims are that NERD outperforms backpropagation-trained control models in capturing human performance (with explanatory power enhanced by clustering learned noise distributions) and that individual differences in expected-uncertainty representations predict task success.

Significance. If the empirical results and derivations hold, the work would provide a novel generative-modeling framework for probing higher-order neural representations and uncertainty estimates in decoded neurofeedback, with potential to link unsupervised learning mechanisms to behavioral outcomes. The absence of any quantitative results, error bars, data exclusion criteria, or derivation details prevents assessment of whether these advantages are demonstrated.

major comments (2)
  1. [Abstract] Abstract: the abstract states performance advantages and predictive links but supplies no quantitative results, error bars, data exclusion criteria, or derivation details. Full methods and results sections are required to evaluate whether the data support the claims as stated.
  2. [Abstract] Abstract: the abstract contains no equations or fitting procedures, so it is impossible to determine whether any reported 'predictions' reduce to fitted quantities by construction or rest on independent grounding.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their comments on the abstract. We address each point below, noting that the full manuscript contains the requested quantitative details, methods, results, and derivations as described in the paper.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the abstract states performance advantages and predictive links but supplies no quantitative results, error bars, data exclusion criteria, or derivation details. Full methods and results sections are required to evaluate whether the data support the claims as stated.

    Authors: The abstract is a concise summary of the work, consistent with standard practice in the field. The full manuscript provides detailed methods and results sections, including quantitative performance comparisons (NERD vs. backpropagation controls), error bars on key metrics, explicit data exclusion criteria, and full derivations of the reinforcement learning objective for the diffusion model. These sections directly support the claims regarding outperformance and predictive links to neurofeedback success. We can expand the abstract with select quantitative highlights if requested. revision: partial

  2. Referee: [Abstract] Abstract: the abstract contains no equations or fitting procedures, so it is impossible to determine whether any reported 'predictions' reduce to fitted quantities by construction or rest on independent grounding.

    Authors: The abstract omits equations for accessibility and brevity. The full manuscript details the NERD architecture, the RL-based training procedure for inferring noise distributions from fMRI, the clustering of learned representations, and the independent validation of predictions against held-out behavioral performance data. The individual-difference predictions are tested via cross-validation and compared against control models, establishing that they reflect genuine explanatory power rather than fitting artifacts. revision: no

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The provided abstract introduces the NERD model and a hypothesis about uncertainty minimization but contains no equations, fitting procedures, self-citations, or derivation steps that could be inspected for reduction to inputs by construction. Without the full manuscript's technical details on training, controls, or clustering, no load-bearing circular steps (self-definitional, fitted-input-as-prediction, or otherwise) can be identified or quoted. The central claim therefore remains self-contained against external benchmarks as far as the available text allows.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract supplies insufficient technical detail to enumerate free parameters, axioms, or invented entities. No specific numbers fitted to data, background mathematical assumptions, or new postulated entities are described.

pith-pipeline@v0.9.0 · 5735 in / 1185 out tokens · 32755 ms · 2026-05-22T23:45:48.415674+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

117 extracted references · 117 canonical work pages · 6 internal anchors

  1. [1]

    & Kording, K

    Baker, B., Lansdell, B. & Kording, K. P. Three aspects of representation in neuroscience. Trends in cognitive sciences 26, 942–958 (2022)

    work page 2022

  2. [2]

    Tarr, M. J. & Vuong, Q. C. Visual object recognition. Steven’s handbook of experimental psychology 1, 287–314 (2002)

    work page 2002

  3. [3]

    Squire, L. R. & Zola-Morgan, S. The medial temporal lobe memory system. Science 253, 1380–1386 (1991)

    work page 1991

  4. [4]

    Miller, E. K. & Cohen, J. D. An integrative theory of prefrontal cortex function. Annu. Rev. Neurosci. 24, 167–202 (2001)

    work page 2001

  5. [5]

    & Pasquali, A

    Cleeremans, A., Timmermans, B. & Pasquali, A. Conscious access to first-order and higher-order representations. Trends Cogn. Sci. 11, 465–472 (2007)

    work page 2007

  6. [6]

    & LeDoux, J

    Brown, R., Lau, H. & LeDoux, J. E. Understanding the higher-order approach to consciousness. Trends in Cognitive Sciences 23, 754–768 (2019). URL https://pubmed.ncbi.nlm.nih.gov/31375408/

    work page arXiv 2019

  7. [7]

    Fleming, S. M. Awareness as inference in a higher-order state space. Neuroscience of consciousness 2020, niz020 (2020)

    work page 2020

  8. [8]

    Consciousness, metacognition, & perceptual reality monitoring

    Lau, H. Consciousness, metacognition, & perceptual reality monitoring. PsyArxiv (2019)

    work page 2019

  9. [9]

    The Perceptual Reality Monitoring Theory (1st edition)

    Michel, M. The Perceptual Reality Monitoring Theory (1st edition). In Herzog, M., Schurger, A. & Doerig, A. (eds.) Scientific Theories of Consciousness: The Grand Tour (Cambridge University Press, 2024)

    work page 2024

  10. [10]

    & Dehaene, S

    Meyniel, F., Schlunegger, D. & Dehaene, S. The sense of confidence during probabilistic learning: A normative account. PLoS Computational Biology 11, e1004305 (2015). URL https://journals. plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1004305

    work page doi:10.1371/journal.pcbi.1004305 2015

  11. [11]

    & Mainen, Z

    Meyniel, F., Sigman, M. & Mainen, Z. F. Confidence as a metacognitive source of learning speed. Nat. Rev. Neurosci. 16, 721–729 (2015)

    work page 2015

  12. [12]

    Response-based outcome predictions and confidence regulate feedback processing and learning

    Fr¨ omer, R.et al. Response-based outcome predictions and confidence regulate feedback processing and learning. eLife 10, e62825 (2021). URL https://elifesciences.org/articles/62825

    work page 2021

  13. [13]

    & de Gardelle, V

    Hainguerlot, M., Vergnaud, J.-C. & de Gardelle, V. Metacognitive ability predicts learning cue- stimulus associations in the absence of external feedback. Scientific Reports 8, 5602 (2018). URL https://www.nature.com/articles/s41598-018-23936-9 . 21

    work page 2018

  14. [14]

    & Dehaene, S

    Meyniel, F. & Dehaene, S. Brain networks for confidence weighting and hierarchical inference during probabilistic learning. Proceedings of the National Academy of Sciences 114, E3859–E3868 (2017). URL https://www.pnas.org/doi/10.1073/pnas.1615773114

    work page doi:10.1073/pnas.1615773114 2017

  15. [15]

    Guggenmos, M., Wilbertz, G., Hebart, M. N. & Sterzer, P. Mesolimbic confidence signals guide perceptual learning in the absence of external feedback. eLife 5, e13388 (2016)

    work page 2016

  16. [16]

    Reverse engineering of metacognition

    Guggenmos, M. Reverse engineering of metacognition. eLife 11, e75420 (2022). URL https:// elifesciences.org/articles/75420

    work page 2022

  17. [17]

    Peters, M. A. Introspective psychophysics for the study of subjective experience. Cerebral Cortex 35, 49–57 (2025)

    work page 2025

  18. [18]

    & Rahnev, D

    Shekhar, M. & Rahnev, D. How do humans give confidence? a comprehensive comparison of process models of perceptual metacognition. Journal of Experimental Psychology: General 153, 656 (2024)

    work page 2024

  19. [19]

    M., Ziemba, C

    Boundy-Singer, Z. M., Ziemba, C. M. & Goris, R. L. T. Confidence reflects a noisy decision reliability estimate. Nature Human Behaviour 7, 142–154 (2023). URL https://www.nature.com/articles/ s41562-022-01464-x

    work page 2023

  20. [20]

    Confidence forced-choice and other metaperceptual tasks

    Mamassian, P. Confidence forced-choice and other metaperceptual tasks. Perception 47, 1023–1035 (2018)

    work page 2018

  21. [21]

    & de Gardelle, V

    Mamassian, P. & de Gardelle, V. Modeling perceptual confidence and the confidence forced-choice paradigm. Psychol. Rev. 129, 976–998 (2022)

    work page 2022

  22. [22]

    & de Gardelle, V

    Mamassian, P. & de Gardelle, V. The confidence-noise confidence-boost (cncb) model of confi- dence rating data. bioRxiv (2024). URL https://www.biorxiv.org/content/10.1101/2024.09. 04.611165v2

    work page doi:10.1101/2024.09 2024

  23. [23]

    Peters, M. A. et al. Perceptual confidence neglects decision-incongruent evidence in the brain. Nature human behaviour 1, 0139 (2017)

    work page 2017

  24. [24]

    Peters, M. A. K. & Azimi Asrari, H. How brains build higher order representations of uncertainty. arXiv preprint arXiv:2506.19057 (2025)

    work page arXiv 2025

  25. [25]

    Winter, C. J. & Peters, M. A. Variance misperception under skewed empirical noise statistics explains overconfidence in the visual periphery. Attention, Perception, & Psychophysics 84, 161–178 (2022). 22

    work page 2022

  26. [26]

    Knill, D. C. & Richards, W. (eds.) Perception as Bayesian Inference (Cambridge University Press, 1996). URL https://www.cambridge.org/core/books/perception-as-bayesian-inference/ 0442F577F5E4CD874FA6819978574C8F

    work page 1996

  27. [27]

    & Fleming, S

    Bang, D. & Fleming, S. M. Distinct encoding of decision confidence in human medial prefrontal cortex. Proceedings of the National Academy of Sciences 115, 6082–6087 (2018). URL https://doi.org/10. 1073/pnas.1800795115

    work page 2018

  28. [28]

    & Kawato, M

    Cortese, A., Amano, K., Koizumi, A., Lau, H. & Kawato, M. Decoded fmri neurofeedback can induce bidirectional confidence changes within single participants. NeuroImage 149, 323–337 (2017)

    work page 2017

  29. [29]

    & Shadlen, M

    Kiani, R. & Shadlen, M. N. Representation of confidence associated with a decision by neurons in the parietal cortex. Science (2009)

    work page 2009

  30. [30]

    Odegaard, B. et al. Superior colliculus neuronal ensemble activity signals optimal rather than subjective confidence. Proceedings of the National Academy of Sciences 115, E1588–E1597 (2018). URL https: //www.pnas.org/doi/abs/10.1073/pnas.1711628115

    work page doi:10.1073/pnas.1711628115 2018

  31. [31]

    Walker, E. Y. et al. Studying the neural representations of uncertainty. Nat. Neurosci. 26, 1857–1867 (2023)

    work page 2023

  32. [32]

    Peters, M. A. K. Towards characterizing the canonical computations generating phenomenal experi- ence. Neurosci. Biobehav. Rev. 142, 104903 (2022)

    work page 2022

  33. [33]

    J., Beck, J

    Ma, W. J., Beck, J. M., Latham, P. E. & Pouget, A. Bayesian inference with probabilistic population codes. Nat. Neurosci. 9, 1432–1438 (2006)

    work page 2006

  34. [34]

    Ma, W. J. & Pouget, A. Population codes, correlations, and coding. In Gazzaniga, M. S. (ed.) The Cognitive Neurosciences, 135–144 (MIT Press, 2009), 4th edn

    work page 2009

  35. [35]

    van Bergen, R. S. & Jehee, J. F. M. Tafkap: An improved method for probabilistic decoding of cortical activity. bioRxiv (2021)

    work page 2021

  36. [36]

    Kay, K. N. et al. Disentangling signal and noise in neural responses through generative modeling. bioRxiv (2024). URL https://pubmed.ncbi.nlm.nih.gov/38712051/

    work page arXiv 2024

  37. [37]

    Prince, J. S. et al. Improving the accuracy of single-trial fmri response estimates using glmsingle. Elife 11, e77599 (2022)

    work page 2022

  38. [38]

    LaConte, S. M. Decoding fmri brain states in real-time. Neuroimage 56, 440–454 (2011). 23

    work page 2011

  39. [39]

    & Kawato, M

    Watanabe, T., Sasaki, Y., Shibata, K. & Kawato, M. Advances in fmri real-time neurofeedback. Trends in cognitive sciences 21, 997–1010 (2017)

    work page 2017

  40. [40]

    & Baranes, A

    Gottlieb, J., Oudeyer, P.-Y., Lopes, M. & Baranes, A. Information-seeking, curiosity, and attention: computational and neural mechanisms. Trends in Cognitive Sciences 17, 585–593 (2013)

    work page 2013

  41. [41]

    & Oudeyer, P.-Y

    Gottlieb, J. & Oudeyer, P.-Y. Towards a neuroscience of active sampling and curiosity. Nature Reviews Neuroscience 19, 758–770 (2018)

    work page 2018

  42. [42]

    & Kawato, M

    Cortese, A., Lau, H. & Kawato, M. Unconscious reinforcement learning of hidden brain states sup- ported by confidence. Nature Communications 11, 4429 (2020). URL https://www.nature.com/ articles/s41467-020-17828-8

    work page 2020

  43. [43]

    & O’Doherty, J

    Cross, L., Cockburn, J., Yue, Y. & O’Doherty, J. Using deep reinforcement learning to reveal how the brain encodes abstract state-space representations in high-dimensional environments. Neuron 109, 724–738.e7 (2021)

    work page 2021

  44. [44]

    & Huth, A

    LeBel, A., Jain, S. & Huth, A. G. Voxelwise encoding models show that cerebellar language represen- tations are highly conceptual. Journal of Neuroscience 41, 10341–10355 (2021)

    work page 2021

  45. [45]

    O., Huth, A

    Nunez-Elizalde, A. O., Huth, A. G. & Gallant, J. L. Voxelwise encoding models with non-spherical multivariate normal priors. NeuroImage 197, 482–492 (2019)

    work page 2019

  46. [46]

    Nishimoto, S. et al. Reconstructing visual experiences from brain activity evoked by natural movies. Current Biology 21, 1641–1646 (2011)

    work page 2011

  47. [47]

    Yamins, D. L. K. et al. Performance-optimized hierarchical models predict neural responses in higher visual cortex. Proceedings of the National Academy of Sciences 111, 8619–8624 (2014). URL https: //www.pnas.org/doi/10.1073/pnas.1403112111

    work page doi:10.1073/pnas.1403112111 2014

  48. [48]

    Cortese, A. et al. The decnef collection, fmri data from closed-loop decoded neurofeedback experiments. Scientific Data 8, 69 (2021)

    work page 2021

  49. [49]

    & Peters, M

    Azimi Azrari, H. & Peters, M. A. Diffusion models and reinforcement learning: Novel pathways to modeling decoded fmri neurofeedback. Proceedings of the Cognitive Computational Neuroscience Meeting (2024)

    work page 2024

  50. [50]

    & Peters, M

    Azimi Asrari, H. & Peters, M. A. K. Diffusion models and reinforcement learning: Novel pathways to modeling decoded fmri neurofeedback. Conference on Cognitive Computational Neuroscience (CCN) (2024). 24

    work page 2024

  51. [51]

    & other authors

    Wang, C., Chen, J., Jiang, H. & other authors. Diffusion-driven policy optimization. In International Conference on Learning Representations (2023)

    work page 2023

  52. [52]

    Training Diffusion Models with Reinforcement Learning

    Black, K., Janner, M., Du, Y., Kostrikov, I. & Levine, S. Training diffusion models with reinforcement learning. arXiv preprint arXiv:2305.13301 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  53. [53]

    & Sasaki, Y

    Shibata, K., Watanabe, T., Kawato, M. & Sasaki, Y. Differential activation patterns in the same brain region led to opposite emotional states. PLOS Biology (2016)

    work page 2016

  54. [54]

    Bishop, C. M. Pattern Recognition and Machine Learning (Springer, 2006)

    work page 2006

  55. [55]

    Cunningham, J. P. & Yu, B. M. Dimensionality reduction for large-scale neural recordings. Nature neuroscience 17, 1500–1509 (2014)

    work page 2014

  56. [56]

    & Ostojic, S

    Jazayeri, M. & Ostojic, S. Interpreting neural computations by examining intrinsic and embedding dimensionality of neural activity. Current opinion in neurobiology 70, 113–120 (2021)

    work page 2021

  57. [57]

    Pang, R., Lansdell, B. J. & Fairhall, A. L. Dimensionality reduction in neuroscience. Current Biology 26, R656–R660 (2016)

    work page 2016

  58. [58]

    On lines and planes of closest fit to systems of points in space

    Pearson, K. On lines and planes of closest fit to systems of points in space. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science 2, 559–572 (1901)

    work page 1901

  59. [59]

    Knill, D. C. & Pouget, A. The bayesian brain: the role of uncertainty in neural coding and computation. Trends in Neurosciences 27, 712–719 (2004). URL https://pubmed.ncbi.nlm.nih.gov/15541511/

    work page arXiv 2004

  60. [60]

    Cockburn, J., Man, V., Cunningham, W. A. & O’Doherty, J. P. Novelty and uncertainty regulate the balance between exploration and exploitation through distinct mechanisms in the human brain. Neuron 110, 2691–2702 (2022). URL https://pubmed.ncbi.nlm.nih.gov/35809575/

    work page arXiv 2022

  61. [61]

    Attention, learning, and the value of information

    Gottlieb, J. Attention, learning, and the value of information. Neuron 76, 281–295 (2012). URL https://pubmed.ncbi.nlm.nih.gov/23083732/

    work page arXiv 2012

  62. [62]

    & Gottlieb, J

    Jiwa, M., Rothkopf, C. & Gottlieb, J. Generating saccades for reducing uncertainty: Cognitive and sensorimotor trade-offs. Journal of Vision 24 (2024). URL https://doi.org/10.1167/jov.24.10. 908

    work page doi:10.1167/jov.24.10 2024

  63. [63]

    C., Blumstein, D

    Mobbs, D., Trimmer, P. C., Blumstein, D. T. & Dayan, P. Foraging for foundations in decision neuroscience: insights from ethology. Nature Reviews Neuroscience 19, 419–427 (2018). URL https: //www.nature.com/articles/s41583-018-0010-7 . 25

    work page 2018

  64. [64]

    The central role of uncertainty reduction in determining behaviour

    Inglis, I. The central role of uncertainty reduction in determining behaviour. Behaviour 137, 1567–1599 (2000). URL https://brill.com/view/journals/beh/137/12/article-p1567_1.xml

    work page 2000

  65. [65]

    Shibata, K. et al. Toward a comprehensive understanding of the neural mechanisms of decoded neu- rofeedback. Neuroscience & Biobehavioral Reviews (2021)

    work page 2021

  66. [66]

    Taschereau-Dumouchel, V., Cushing, C. A. & Lau, H. Real-time functional mri in the treatment of mental health disorders. Annual review of clinical psychology 18, 125–154 (2022)

    work page 2022

  67. [67]

    Taschereau-Dumouchel, V. et al. Towards an unconscious neural reinforcement intervention for com- mon fears. Proceedings of the National Academy of Sciences 115, 3470–3475 (2018)

    work page 2018

  68. [68]

    Pospisil, D. A. & Pillow, J. W. Revisiting the high-dimensional geometry of population responses in visual cortex. bioRxiv (2024). URL https://doi.org/10.1101/2024.02.16.580726

    work page doi:10.1101/2024.02.16.580726 2024

  69. [69]

    & Harris, K

    Stringer, C., Pachitariu, M., Steinmetz, N., Carandini, M. & Harris, K. D. High-dimensional geometry of population responses in visual cortex. Nature 571, 361–365 (2019). URL https://www.nature. com/articles/s41586-019-1346-5

    work page 2019

  70. [70]

    Williams, A. H. & Linderman, S. W. Statistical neuroscience in the single trial limit. Current Opinion in Neurobiology 70, 193–205 (2021). URL https://pubmed.ncbi.nlm.nih.gov/34861596/

    work page arXiv 2021

  71. [71]

    S., Ma, W

    van Bergen, R. S., Ma, W. J., Pratte, M. S. & Jehee, J. F. M. Sensory uncertainty decoded from visual cortex predicts behavior. Nature Neuroscience 18, 1728–1730 (2015). URL https://www.nature. com/articles/nn.4150

    work page 2015

  72. [72]

    Deep Unsupervised Learning using Nonequilibrium Thermodynamics

    Sohl-Dickstein, J., Weiss, E., Maheswaranathan, N. & Ganguli, S. Deep unsupervised learning using nonequilibrium thermodynamics. arXiv preprint arXiv:1503.03585 (2015)

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  73. [73]

    Proximal Policy Optimization Algorithms

    Schulman, J., Wolski, F., Dhariwal, P., Radford, A. & Klimov, O. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347 (2017)

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  74. [74]

    Williams, R. J. Simple statistical gradient-following algorithms for connectionist reinforcement learn- ing. Machine learning 8, 229–256 (1992)

    work page 1992

  75. [75]

    Schneider, S., Lee, J. H. & Mathis, M. W. Learnable latent embeddings for joint behavioural and neural analysis. Nature 617, 360–368 (2023). URL https://www.nature.com/articles/ s41586-023-06031-6 . 26

    work page 2023

  76. [76]

    Steinmetz, N. A. et al. Neuropixels 2.0: A miniaturized high-density probe for stable, long-term brain recordings. Science 372, eabf4588 (2021). URL https://www.science.org/doi/abs/10.1126/ science.abf4588

    work page 2021

  77. [77]

    & G´ orriz, J

    V´ azquez-Garc´ ıa, C., Mart´ ınez-Murcia, F., Segovia Rom´ an, F. & G´ orriz, J. M. A review of latent representation models in neuroimaging. arXiv preprint arXiv:2412.19844 (2024). URL https:// arxiv.org/abs/2412.19844

    work page arXiv 2024

  78. [78]

    Favela, L. H. & Machery, E. Investigating the concept of representation in the neural and psychological sciences. Frontiers in Psychology 14, 1165622 (2023)

    work page 2023

  79. [79]

    Favela, L. H. & Machery, E. Contextualizing, eliminating, or glossing: What to do with unclear scientific concepts like representation. Mind & Language (2025)

    work page 2025

  80. [80]

    Neural representations: A normative account

    Machery, E. Neural representations: A normative account. Mind & Language (2025). URL https: //onlinelibrary.wiley.com/doi/10.1111/mila.12531

    work page doi:10.1111/mila.12531 2025

Showing first 80 references.