- Rating
- 1
- Confidence
- 3
- Informativeness
- 1
- Presentation
- 1
- Soundness
- 1
Read review
Paper summary
This article trains an adaptive RNN on a variable-delay sequence memory task. Oscillatory phase coding occurs without external rhythm or phase supervision: amplitude represents item content, while phase represents sequence order. The author identified the content and time cells and proposed a manually designed circuit to reproduce this organization.
Critical strengths
1.The phase coding that appears in task training is an interesting discovery.
2.Ablation links neural adaptation, delay variability, and readout timing with learned representations.
3.The separation of content and time provides a concrete and testable hypothesis.
Critical weaknesses
1.The main paper is not self-contained enough. The dataset source, train/test split, delay range, network size, and definitions of key analysis metrics are missing or deferred to the appendix. These details are necessary for evaluating the core claims and must be included in the main text.
2.The claim of necessity exceeds the evidence, and the failure to obtain stable phase coding in the ablation test does not necessarily mean that adaptation, variable delay, and flexible readout are usually necessary. The main paper still lacks aggregate results and uncertainty across independently trained networks.
3.The correlation between phase consistency and retrieval accuracy was reported, but it cannot be determined whether the network relies on phase coding to achieve successful recall. Targeted perturbations of phase relationships or identified cell populations will provide stronger evidence.
4.The circuit model builds in most of the proposed explanations, and its prescribed coupling gates and serial-order phases demonstrate a possible implementation, but it does not determine how these mechanisms are generated through learning or operate in biological systems.
Questions for authors
1.Can the main paper report on the basic experimental setup, metric definitions, accuracy, and phase consistency results across seeds?
2.Will disrupting phase relationships impair recall while retaining item representations?
3.Will a non-adaptive control with state and parameter matching yield the same conclusion about adaptation?
4.How does the proposed coding scheme handle repeated digits and different sequence lengths?
4 comments
Response to Reviewer yod7 (Part II/III)
Second, to test directly whether recall relies on internal phase relationships while preserving item representations (Q2), we performed the exact perturbation requested by the reviewer (Appendix C.3). During the delay (at or , one cycle before the cue at step ), we advanced selected neuronal populations along their own single-trial trajectories by , , or of a cycle ( steps)—shifting phase while preserving amplitude—and then let the network evolve freely ( trials; unperturbed baseline ; results for the half-cycle shift at shown below):
| Neurons shifted (Appendix C.3) | Accuracy | Order swaps among errors | Cued-item amplitude ratio (perturbed / intact) |
|---|---|---|---|
| All 500 neurons (global time shift) | 97.6% | 12.0% | 1.03 |
| 74 time cells only | 13.5% | 96.8% | 1.02 |
| 370 content cells only | 0.0% | 95.4% | 0.97 |
| 74 random non-time neurons | 96.8% | 15.4% | 0.84 |
Selectively disrupting the relative phase between the time-cell reference and content cells abolishes recall, and of the induced errors are pure serial-order swaps while the cued item's readout amplitude remains intact (). By contrast, applying the exact same shift globally to all neurons—or to random non-time neurons—leaves recall intact ( and ). Full results across steps and both intervention times (), as well as pathway-removal experiments, are reported in Appendix C.3.
W4. Role and scope of the circuit model
The circuit model (Section 3.4; Appendix F, F.1–F.4, Figs. 10–11) is a constructive, biologically grounded implementation inspired by the functional organization uncovered in the trained RNN, rather than a model of synaptic learning. Built from established neurobiological motifs—continuous-attractor dynamics, spike-frequency adaptation, and NMDA-like activity-dependent coupling—it demonstrates that the time-ring plus content-assembly architecture can sustain and read out OPC in a mechanistic neural circuit across a broad connected parameter region (Appendix F.4, Fig. 11), yielding concrete experimental predictions (Appendix I; highlighted as a strength by Reviewer medk). At the functional level, the A-RNN itself shows that this organization emerges from task optimization alone; how biological circuits realize and learn these cellular and synaptic couplings remains an open question and is discussed in Section 4.3 (Limitations and Future Directions).
We disagree that the absence of a biological learning rule is a weakness of the circuit model. This objection conflates two distinct explanatory targets: how a circuit performs a computation, and how an organism acquires that circuit. The former has been studied without requiring experimental verification of the latter in models like grid-cell path integration (Burak & Fiete, 2009, PLoS Computational Biology), relational memory (Whittington et al., 2020, Cell), hippocampal phase precession (Chu et al., 2024, eLife), and grid-cell modularity (Khona et al., 2025, Nature), and etc.
Q3. State- and parameter-matched non-adaptive controls
Because each A-RNN unit has two state variables ( and , totaling states and parameters at ), our original four-item non-adaptive control (502 hidden units; Appendix D) approximately matched the parameter budget but not recurrent states. We therefore trained controls with the same number of recurrent states and nearly the same number of trainable parameters ( states, parameters at , plus ; Appendix J.1, Fig. 18a–g) that retain the second state and modify only the feedback structure:
| Model ( / , 3 seeds each at ; Appendix J.1) | Feedback type | Learned gain (median) | Robust | Period | Long-delay accuracy () |
|---|---|---|---|---|---|
| A-RNN (paired seed) | Slow negative | 1.13 / 1.48 | 0.97 / 0.96 | (60 / 55) | 84.8% / 88.6% |
| Sign-free feedback (positive init.) | Learned, unconstrained sign | −0.03 / 0.00 | 0.43 / 0.40–0.61 | None at task scale† | 67.0% / 68.9–80.1% |
| Sign-free feedback (zero init.) | Learned, unconstrained sign | 0.00 / 0.00 | 0.44 / 0.31–0.39 | None at task scale† | 55.8% / 53.8–59.2% |
| Fast negative feedback ( as fast as ) | Fast negative (learnable rate) | 1.07 / 1.28 | 0.55 / 0.89–0.94 | (18–21) | 63.0% / 58.0–72.5% |
| Non-adaptive RNN () | None | — | 0.35 / 0.22 | None at task scale† | 79.7% / 63.0% |
(†Spectral peak falls in the lowest resolvable frequency bin of the analysis window, indicating no oscillation at the task timescale.)
Response to Reviewer yod7 (Part III/III)
- Sign-free feedback: When gradient descent is free to learn the feedback sign, it never discovers negative feedback—the learned gain collapses to near zero (). Although these networks learn in-distribution recall quickly and accurately (; zero-init reaches validation accuracy at epochs vs. epoch for the A-RNN; Fig. 18g), they develop no task-timescale oscillation or OPC (; Fig. 18b,c,e), keep their readout gate pinned to the last retrieval step like vanilla RNNs, and generalize poorly to unseen long delays ( vs. ; Fig. 18f).
- Fast negative feedback: Making as fast as produces a rhythm roughly three times faster than the retrieval window ( steps; Fig. 18d), so each phase recurs multiple times within , slowing learning (reaching at epochs ; Fig. 18g) and degrading long-delay generalization (; Fig. 18f).
- Conclusion: Adding an extra slow state variable by itself does not explain OPC or its long-delay generalization advantage; slow negative feedback (spike-frequency adaptation) is the essential mechanism, fully confirming our original conclusion (Appendix J.1).
Q4. Repeated digits and sequence length
- Sequence length: OPC emerges consistently across in our multi-configuration CNN-A-RNNs ( in three-item, four-item, and five-item networks; Appendix H.2, Table 5, Fig. 14) and across in all one-hot A-RNNs (Appendix K, K.1–K.2, Fig. 19).
- Repeated digits (Appendix J.2, Fig. 18h–m): In our standard task, digits represent distinct abstract items (following standard serial-recall paradigms without replacement). To test how the phase code handles item repeats, we trained new networks with sampling with replacement ( of trials contain repeats; 2 seeds):
- High recall accuracy & human-like Ranschburg cost: Repeat-trained networks achieve on standard no-repeat trials (main network: ) and on repeat trials, with slightly lower accuracy when the cued item itself is the repeated digit () than when a unique digit in the sequence is cued (), mirroring the classic Ranschburg effect in human serial recall.
- Zero-shot prediction on adjacent repeats vs. learned gate alignment on non-adjacent repeats: Even the original main network (never trained on repeats) generalizes zero-shot when a repeated digit occupies adjacent positions (), though its accuracy drops on non-adjacent repeats (; Fig. 18h,i,l). Training with replacement equips the temporal gating head with much stronger alignment to the repeated digit's retrieval peak across different ordinal cues (gate-to-peak alignment within 3 circular steps more than doubles from to ; Fig. 18m), thereby successfully resolving non-adjacent repeats as well (; Fig. 18j–l).
- Distinct-item maintenance during the delay (single peak per cycle): Crucially, in both the untrained and repeat-trained networks, the delay-period OPC maintains mutually distinct digit identities—a repeated digit exhibits only a single prominent readout peak per oscillation cycle ( in the distinct-only model and in the repeat-trained models; Fig. 18m) rather than peaking twice within a cycle, while position disambiguation at retrieval is handled by the cue-dependent gating head.
Summary on informativeness
In summary, our existing analyses and new experiments directly address each of your questions: we have clarified the quantitative cell-classification criteria (W1/Q1), reported full multi-seed statistics across all ablations and trained A-RNNs (W2/Q5), demonstrated via targeted phase-shift perturbations that recall causally relies on the relative phase between time and content cells (W3/Q2), clarified the constructive role of the biological circuit model (W4), and verified that OPC requires slow negative feedback rather than extra state variables (Q3) and extends robustly across sequence lengths () and repeated items (Q4). We hope these clarifications and new experiments fully resolve your concerns and would greatly appreciate a reassessment of our work.
Reminder to reply
Dear Reviewer, We kindly ask you to consider our responses and would be happy to address any further questions. As the author–reviewer discussion period ends on November 18 (AoE), we would greatly appreciate it if you could reply at your earliest convenience.
Thank you for your time and constructive feedback.
Response to Reviewer yod7 (Part I/III)
We thank the reviewer for finding the emergent phase coding "an interesting discovery", the ablations informative, and the content/time separation "a concrete and testable hypothesis". Several concerns (notably W2 on multi-seed validation and W3 on causal interventions) are already addressed in the submission; below we point to their locations, clarify our presentation, and report the requested new experiments directly.
W1/Q1. Self-containedness of the main text
Within the strict page limit, our initial main text focused on the core scientific findings while documenting full implementation details and metric definitions in Appendices A and C, following standard ICLR practice (e.g., Dosovitskiy et al., 2021; Keller et al., 2024). Using the additional rebuttal page, we have added a dedicated "Experimental setup and metrics" paragraph directly to Section 2.2 of the main text (and clarified the definition of Rϕ in Section 3.2) so that the main text is self-contained:
W2/Q1. Multi-seed replications and scope of the necessity claim
Results across independently trained networks are already reported in the submission and cited in the main text (Section 3.1: "Appendix H.2; Figure 14"; Section 3.3: "Appendix H.3" and "Appendix H.4; Figure 16"). Specifically, Appendix E (Table 3; see also Table 4 for T2=220) reports each ablation across 4–6 independently initialized networks (mean ± sample SD at T2=100), which we summarize below alongside our new state-matched controls (Appendix J.1, Fig. 18a–g):
*Oscillates at roughly 3× the task frequency (period ≈T3/3=18–21 steps) with degraded long-delay generalization (58.0–72.5% vs. 84.8–88.6% for A-RNNs; see Q3 and Appendix J.1).
Furthermore, Appendix H.2–H.4 (Figs. 14–16, Table 5) replicates the phase code, time/content cell populations, and connectivity structure across 12 independently trained A-RNNs varying in N∈{3,4,5}, T3∈{60,65,70}, and H∈{450,500,550}:
Because 0/14 ablation runs and 0/8 sign-free state-matched controls develop stable OPC (Rϕ>0.9), these multi-seed results confirm that all three factors are essential for generating stable OPC (and we have updated "necessary" in the Introduction to "required for ... in this working memory task" to avoid any formal-logic ambiguity).
W3/Q2. Does recall causally rely on the phase code? Targeted perturbations
First, the submission already includes a direct causal intervention on the phase readout: shifting the learned temporal gate away from the target phase while holding digit logits fixed drops recall accuracy from 95.7–98.7% to 16.7–29.1% across 12 networks (Appendix H.2, Fig. 14b; see also Section 3.1, Fig. 2e). Pathway-removal (connection-pruning) analyses in Appendix C.3 likewise confirm distinct input/output routing for time and content populations.