Nicholas Menghi
@nichome.bsky.social
Postdoc @mpicbs.bsky.social, interested in Generalization, Transfer Learning, Cats and Pirates
Pinned
Nicholas Menghi
@nichome.bsky.social
· Dec 2
The effects of task similarity during representation learning in brains and neural networks
Nature Communications - Here, the authors show learning tasks with similar structures can initially cause interference and slow down learning, but both the brain and artificial networks gradually...
rdcu.be
Our new paper, now published in @natcomms.nature.com , asks a simple question: when two tasks share a common structure, does the brain learn them more efficiently? Surprisingly, this was not the case. Thread below (1/7)
rdcu.be/eSwvU
rdcu.be/eSwvU
Reposted by Nicholas Menghi
Final paper of my PhD 🤗
www.nature.com/articles/s44...
There is growing interest in how cognitive control may improve value-based decision making.
However, we find that a recent paper overestimated the role of control in their task, leading to erroneous interpretations of dACC recordings.
www.nature.com/articles/s44...
There is growing interest in how cognitive control may improve value-based decision making.
However, we find that a recent paper overestimated the role of control in their task, leading to erroneous interpretations of dACC recordings.
Misspecified models create the appearance of adaptive control during value-based choice - Communications Psychology
In a new computational analysis of previous work, this study shows that a control-free mechanism better accounts for value-based decisions than an account that assumes top-down control invigorating th...
www.nature.com
February 3, 2026 at 10:59 PM
Final paper of my PhD 🤗
www.nature.com/articles/s44...
There is growing interest in how cognitive control may improve value-based decision making.
However, we find that a recent paper overestimated the role of control in their task, leading to erroneous interpretations of dACC recordings.
www.nature.com/articles/s44...
There is growing interest in how cognitive control may improve value-based decision making.
However, we find that a recent paper overestimated the role of control in their task, leading to erroneous interpretations of dACC recordings.
Reposted by Nicholas Menghi
Finally: the fantastic #registeredreport from bsky-less Roni Tibon is out: www.nature.com/articles/s41... showing less difference between #episodic vs. #semantic #memory than one might have thought.
Proud to have contributed a tiny part to this great paper.
Proud to have contributed a tiny part to this great paper.
Neural activations and representations during episodic versus semantic memory retrieval - Nature Human Behaviour
In this Stage 2 Registered Report, Tibon et al. showed using fMRI that neural activity associated with successful memory retrieval did not differ between semantic and episodic memory, using a task wit...
www.nature.com
January 27, 2026 at 11:44 AM
Finally: the fantastic #registeredreport from bsky-less Roni Tibon is out: www.nature.com/articles/s41... showing less difference between #episodic vs. #semantic #memory than one might have thought.
Proud to have contributed a tiny part to this great paper.
Proud to have contributed a tiny part to this great paper.
Reposted by Nicholas Menghi
🚨 New paper out in Science Advances 🚨
With @suryagayet.bsky.social and @peelen.bsky.social, in two fMRI studies we investigate mental object rotations that are driven by the scene context, rather than purely by cognitive operations. 🧵 www.science.org/doi/10.1126/...
With @suryagayet.bsky.social and @peelen.bsky.social, in two fMRI studies we investigate mental object rotations that are driven by the scene context, rather than purely by cognitive operations. 🧵 www.science.org/doi/10.1126/...
January 23, 2026 at 3:16 PM
🚨 New paper out in Science Advances 🚨
With @suryagayet.bsky.social and @peelen.bsky.social, in two fMRI studies we investigate mental object rotations that are driven by the scene context, rather than purely by cognitive operations. 🧵 www.science.org/doi/10.1126/...
With @suryagayet.bsky.social and @peelen.bsky.social, in two fMRI studies we investigate mental object rotations that are driven by the scene context, rather than purely by cognitive operations. 🧵 www.science.org/doi/10.1126/...
Reposted by Nicholas Menghi
How do we achieve few-shot generalization? New work led by @fabianrenz.bsky.social dives into the role of replay in learning and using structure to generalize reward. Dream team effort with Shany Grossman @nathanieldaw.bsky.social Peter Dayan & @doellerlab.bsky.social
www.biorxiv.org/content/10.6...
www.biorxiv.org/content/10.6...
www.biorxiv.org
January 18, 2026 at 3:33 PM
How do we achieve few-shot generalization? New work led by @fabianrenz.bsky.social dives into the role of replay in learning and using structure to generalize reward. Dream team effort with Shany Grossman @nathanieldaw.bsky.social Peter Dayan & @doellerlab.bsky.social
www.biorxiv.org/content/10.6...
www.biorxiv.org/content/10.6...
Reposted by Nicholas Menghi
Our new paper, now published in Cell Reports, asks how the brain adaptively shapes its representations according to the statistical structure of the environment to overcome the limits of working memory capacity.
www.cell.com/cell-reports...
www.cell.com/cell-reports...
Efficient coding in working memory is adapted to the structure of the environment
Huang et al. show that the brain optimizes working memory by compressing information
when environmental regularities exist. MEG reveals distinct neural systems for abstract
structure and item details,...
www.cell.com
January 16, 2026 at 1:47 PM
Our new paper, now published in Cell Reports, asks how the brain adaptively shapes its representations according to the statistical structure of the environment to overcome the limits of working memory capacity.
www.cell.com/cell-reports...
www.cell.com/cell-reports...
Reposted by Nicholas Menghi
Efficient coding in working memory is adapted to the structure of the environment
doi.org/10.1016/j.ce...
#neuroscience
doi.org/10.1016/j.ce...
#neuroscience
Redirecting
doi.org
January 15, 2026 at 8:16 PM
Efficient coding in working memory is adapted to the structure of the environment
doi.org/10.1016/j.ce...
#neuroscience
doi.org/10.1016/j.ce...
#neuroscience
Reposted by Nicholas Menghi
super excited to share my preprint with @meganakpeters.bsky.social stimulus familiarity shapes hierarchical structure learning and metacognitive dynamics🚀😊!!
osf.io/preprints/ps...
osf.io/preprints/ps...
January 15, 2026 at 3:07 AM
super excited to share my preprint with @meganakpeters.bsky.social stimulus familiarity shapes hierarchical structure learning and metacognitive dynamics🚀😊!!
osf.io/preprints/ps...
osf.io/preprints/ps...
Reposted by Nicholas Menghi
1/ Excited to share our new work published in Transactions on Machine Learning Research (TMLR), Stacking Variational Bayesian Monte Carlo (S-VBMC)!
January 14, 2026 at 2:31 PM
1/ Excited to share our new work published in Transactions on Machine Learning Research (TMLR), Stacking Variational Bayesian Monte Carlo (S-VBMC)!
Reposted by Nicholas Menghi
New preprint: Inference over hidden contexts shapes the geometry of conceptual knowledge for flexible behaviour.
In this pre-reg study, our core claim was that we don’t just learn stimulus-reward. We infer hidden context and that inference re-wires attention and neural state space on the fly.
1/8
In this pre-reg study, our core claim was that we don’t just learn stimulus-reward. We infer hidden context and that inference re-wires attention and neural state space on the fly.
1/8
January 8, 2026 at 7:46 AM
New preprint: Inference over hidden contexts shapes the geometry of conceptual knowledge for flexible behaviour.
In this pre-reg study, our core claim was that we don’t just learn stimulus-reward. We infer hidden context and that inference re-wires attention and neural state space on the fly.
1/8
In this pre-reg study, our core claim was that we don’t just learn stimulus-reward. We infer hidden context and that inference re-wires attention and neural state space on the fly.
1/8
Reposted by Nicholas Menghi
Our holiday celebrations are especially happy this year, with a new paper out: "Global and Local Deviance Effects in the Processing of Temporal Patterns", about how we detect and keep track of temporal regularities.
Congrats to @duniagiomo.bsky.social and the whole team!🎉
doi.org/10.1111/nyas...
Congrats to @duniagiomo.bsky.social and the whole team!🎉
doi.org/10.1111/nyas...
NYAS Publications
Our experience of the world is inherently structured by temporal patterns. Yet a full understanding of how we process such patterns is still lacking. Across three finger-tapping experiments employing...
nyaspubs.onlinelibrary.wiley.com
December 26, 2025 at 5:42 PM
Our holiday celebrations are especially happy this year, with a new paper out: "Global and Local Deviance Effects in the Processing of Temporal Patterns", about how we detect and keep track of temporal regularities.
Congrats to @duniagiomo.bsky.social and the whole team!🎉
doi.org/10.1111/nyas...
Congrats to @duniagiomo.bsky.social and the whole team!🎉
doi.org/10.1111/nyas...
Reposted by Nicholas Menghi
We’re very happy to share that our work on 3D spatial memory was published in PNAS just before the end of the year! 🎉
Link: www.pnas.org/doi/10.1073/...
(1/8)
Link: www.pnas.org/doi/10.1073/...
(1/8)
PNAS
Proceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans...
www.pnas.org
December 22, 2025 at 10:43 PM
We’re very happy to share that our work on 3D spatial memory was published in PNAS just before the end of the year! 🎉
Link: www.pnas.org/doi/10.1073/...
(1/8)
Link: www.pnas.org/doi/10.1073/...
(1/8)
Reposted by Nicholas Menghi
New preprint!
www.biorxiv.org/content/10.6...
As you are silently reading this, you may experience a little voice in your head. How is it represented in the brain, and what purpose does it serve? Our new study answers the questions.
Together with @adriendoerig.bsky.social and Radek Cichy.(1/8)
www.biorxiv.org/content/10.6...
As you are silently reading this, you may experience a little voice in your head. How is it represented in the brain, and what purpose does it serve? Our new study answers the questions.
Together with @adriendoerig.bsky.social and Radek Cichy.(1/8)
Auditory representations of words during silent visual reading
Silent visual reading is accompanied by the phenomenological experience of an inner voice. However, the temporal dynamics and functional role of the underlying neural representations remain unclear. H...
www.biorxiv.org
December 15, 2025 at 3:10 PM
New preprint!
www.biorxiv.org/content/10.6...
As you are silently reading this, you may experience a little voice in your head. How is it represented in the brain, and what purpose does it serve? Our new study answers the questions.
Together with @adriendoerig.bsky.social and Radek Cichy.(1/8)
www.biorxiv.org/content/10.6...
As you are silently reading this, you may experience a little voice in your head. How is it represented in the brain, and what purpose does it serve? Our new study answers the questions.
Together with @adriendoerig.bsky.social and Radek Cichy.(1/8)
Reposted by Nicholas Menghi
Our new preprint on promises and caveats of the extraction of M/EEG activity from ROIs is out (see more details in the reposted thread): www.biorxiv.org/content/10.6...
Huge thanks to @studenova.bsky.social, Ruben Eguinoa, Guido Nolte, @sparsity.bsky.social, Arno Villringer, and Vadim Nikulin!
Huge thanks to @studenova.bsky.social, Ruben Eguinoa, Guido Nolte, @sparsity.bsky.social, Arno Villringer, and Vadim Nikulin!
December 12, 2025 at 2:10 PM
Our new preprint on promises and caveats of the extraction of M/EEG activity from ROIs is out (see more details in the reposted thread): www.biorxiv.org/content/10.6...
Huge thanks to @studenova.bsky.social, Ruben Eguinoa, Guido Nolte, @sparsity.bsky.social, Arno Villringer, and Vadim Nikulin!
Huge thanks to @studenova.bsky.social, Ruben Eguinoa, Guido Nolte, @sparsity.bsky.social, Arno Villringer, and Vadim Nikulin!
Reposted by Nicholas Menghi
🧠 New paper alert (the 1st one from our new lab)!
Led by 1st author & VR wizard @jaquent.bsky.social
@natcomms.nature.com
How do our brains distinguish novel from familiar places as we explore our environments, e.g., a new city?
🔗 doi.org/10.1038/s414...
🧵 Thread below with key findings ⬇️
Led by 1st author & VR wizard @jaquent.bsky.social
@natcomms.nature.com
How do our brains distinguish novel from familiar places as we explore our environments, e.g., a new city?
🔗 doi.org/10.1038/s414...
🧵 Thread below with key findings ⬇️
December 9, 2025 at 9:49 AM
🧠 New paper alert (the 1st one from our new lab)!
Led by 1st author & VR wizard @jaquent.bsky.social
@natcomms.nature.com
How do our brains distinguish novel from familiar places as we explore our environments, e.g., a new city?
🔗 doi.org/10.1038/s414...
🧵 Thread below with key findings ⬇️
Led by 1st author & VR wizard @jaquent.bsky.social
@natcomms.nature.com
How do our brains distinguish novel from familiar places as we explore our environments, e.g., a new city?
🔗 doi.org/10.1038/s414...
🧵 Thread below with key findings ⬇️
Reposted by Nicholas Menghi
New article by @jaquent.bsky.social and co:
Graded encoding of spatial novelty scales in the human brain
www.nature.com/articles/s41...
Graded encoding of spatial novelty scales in the human brain
www.nature.com/articles/s41...
December 8, 2025 at 12:18 PM
New article by @jaquent.bsky.social and co:
Graded encoding of spatial novelty scales in the human brain
www.nature.com/articles/s41...
Graded encoding of spatial novelty scales in the human brain
www.nature.com/articles/s41...
Reposted by Nicholas Menghi
New preprint! Have you ever wondered, what are these fuzzy simplicial sets, the theoretical framework behind e.g. UMAP? Here we show that you may simply see them as marginal distributions over simplicial sets. This provides a generative model for UMAP. (1/2)
arxiv.org/abs/2512.03899
arxiv.org/abs/2512.03899
Probabilistic Foundations of Fuzzy Simplicial Sets for Nonlinear Dimensionality Reduction
Fuzzy simplicial sets have become an object of interest in dimensionality reduction and manifold learning, most prominently through their role in UMAP. However, their definition through tools from alg...
arxiv.org
December 4, 2025 at 12:31 PM
New preprint! Have you ever wondered, what are these fuzzy simplicial sets, the theoretical framework behind e.g. UMAP? Here we show that you may simply see them as marginal distributions over simplicial sets. This provides a generative model for UMAP. (1/2)
arxiv.org/abs/2512.03899
arxiv.org/abs/2512.03899
Thank you, Jacob! Appreciate it!
December 3, 2025 at 8:08 AM
Thank you, Jacob! Appreciate it!
😂 Yeah, I was having a hard time too, squeezing everything every time. I'll send you an email!
December 2, 2025 at 5:17 PM
😂 Yeah, I was having a hard time too, squeezing everything every time. I'll send you an email!
For the neural network, the situation is different: it has no spatial priors, so the key variable is the feature similarity between tasks. By manipulating that similarity, we make the tasks more or less similar, which mirrors the similarity manipulation in the human experiment.
December 2, 2025 at 4:46 PM
For the neural network, the situation is different: it has no spatial priors, so the key variable is the feature similarity between tasks. By manipulating that similarity, we make the tasks more or less similar, which mirrors the similarity manipulation in the human experiment.
If the concern is that the effect might reflect only a spatial bias, I don’t think that’s the case, because the rule itself is nonlinear.
December 2, 2025 at 4:46 PM
If the concern is that the effect might reflect only a spatial bias, I don’t think that’s the case, because the rule itself is nonlinear.
What we find is that transfer and interference depend jointly on rule similarity and training regime. Even when tasks share the same structure, this can lead to either transfer or interference depending on whether training is blocked or interleaved.
December 2, 2025 at 4:46 PM
What we find is that transfer and interference depend jointly on rule similarity and training regime. Even when tasks share the same structure, this can lead to either transfer or interference depending on whether training is blocked or interleaved.
If you mean rotating the screen 90°, that effectively changes the rule between the two tasks, this is one of the conditions we test. We have two groups: one where the tasks share the same mapping rule and one where they have different rules, while the stimulus positions remain identical across tasks
December 2, 2025 at 4:46 PM
If you mean rotating the screen 90°, that effectively changes the rule between the two tasks, this is one of the conditions we test. We have two groups: one where the tasks share the same mapping rule and one where they have different rules, while the stimulus positions remain identical across tasks
I’m not fully sure I’m following here, I'm also happy to schedule a Zoom meeting if you’d like to discuss it in more detail.
Let me try: if we simply flip the monitor, the mapping rule stays the same; only the on-screen positions change, so the results should not reverse.
Let me try: if we simply flip the monitor, the mapping rule stays the same; only the on-screen positions change, so the results should not reverse.
December 2, 2025 at 4:46 PM
I’m not fully sure I’m following here, I'm also happy to schedule a Zoom meeting if you’d like to discuss it in more detail.
Let me try: if we simply flip the monitor, the mapping rule stays the same; only the on-screen positions change, so the results should not reverse.
Let me try: if we simply flip the monitor, the mapping rule stays the same; only the on-screen positions change, so the results should not reverse.
We also examined these one-dimensional task-relevant representations in the MEG data by projecting the original two-dimensional positions onto the relevant subspace, and we found the same positive and negative correlation effects.
December 2, 2025 at 2:45 PM
We also examined these one-dimensional task-relevant representations in the MEG data by projecting the original two-dimensional positions onto the relevant subspace, and we found the same positive and negative correlation effects.
The structure is not defined by the spatial positions themselves, but by the mapping between those positions and the task outcomes. It’s the association rule, the underlying manifold, that participants have to learn, not the absolute locations of the stimuli.
December 2, 2025 at 2:45 PM
The structure is not defined by the spatial positions themselves, but by the mapping between those positions and the task outcomes. It’s the association rule, the underlying manifold, that participants have to learn, not the absolute locations of the stimuli.