- Rating
- 2
- Confidence
- 2
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper studies training-loss spikes using power-law random-feature models and proposes a checkpoint-calibrated simulator for layer-wise neural-network dynamics. It connects spikes to stochastic gradient noise and spectral concentration, with experiments on synthetic models and a 1.3B-parameter Transformer trained on C4. It also proposes a measurement-informed Adam schedule and evaluates its effect on training stability and validation loss.
Critical strengths
-
Interpretable modeling framework. Section 2 and Figure 2.1 demonstrate spike-like behavior in simple quadratic random-feature models under SGD and Adam. These controlled experiments provide a useful setting for studying how spectral structure and stochastic optimization relate to training instability.
-
Concrete and lightweight simulator. Section 3.3 and Algorithm 1 translate checkpoint measurements into a procedure for simulating layer-wise dynamics. After calibration, the simulation is computationally inexpensive: Table 1 reports five seconds on CPU for 1,500 steps of one layer.
-
Substantial Transformer case study. Section 4 evaluates a 1.3B-parameter Transformer trained on C4, with controlled changes to Adam’s second-moment parameter. Figures 1.1 and 4.1 connect the synthetic investigation to observed training instability and measured layer-wise behavior.
-
Empirical checks of modeling assumptions. Appendix D.2 examines activation spectra, while Appendix D.3 tests the scalar gradient-norm relationship in Equation (6). These measurements help assess the plausibility of specific assumptions underlying the simulator (Figures D.2 and D.4).
Critical weaknesses
-
The gradient approximation lacks sufficient theoretical justification (major). Section 3.2, Eq. (5) neglects dependence terms that can remain significant when output gradients align with input eigenvectors. Additional conditions or error bounds are needed. Appendix D.3 validates only the scalar relation in Eq. (6), leaving the full second-moment approximation insufficiently supported.
-
Real-model experiments provide limited evidence of predictive accuracy (major). Section 4 evaluates six layers with simulations initialized at one training stage. Figure 4.1 compares only second-order loss changes, while Appendix D.6 acknowledges unexplained first order changes and different recovery behavior. Quantitative predictions from previously unused checkpoints, tested across runs, would better establish whether the simulator reliably identifies future spikes.
-
Quantitative simulation accuracy is limited at larger batch sizes (moderate). Section 3.3’s simulator reproduces qualitative spike behavior, but its quantitative accuracy varies across settings. At batch size 64, simulations spend 36–38% of steps above threshold versus 3.5% for PLRF (Appendix C, Table 2). The authors acknowledge this limitation; the main paper should clarify when accurate spike statistics can be expected.
-
The spectral explanation needs stronger validation in the Transformer (moderate). Section 3.4 proposes a plausible mechanism supported by Section 2’s synthetic experiments. However, Section 4 changes , which affects several aspects of optimization, without isolating the contribution of spectral concentration. More targeted evidence would strengthen the connection between the proposed mechanism and the observed Transformer behavior.
-
Generality beyond one Transformer configuration remains untested (minor). Section 4 evaluates one 1.3B Transformer on C4, without comparisons across architectures, model sizes, or training-data budgets. The synthetic sensitivity experiments do not establish broader real-model generality. Given the case-study framing, this is a minor limitation, but broader applicability claims should remain qualified.
-
The simulator’s value for selecting preventive interventions is unclear (moderate). The optimizer guidance presented in the Introduction is not evaluated through interventions selected using Section 3.3’s simulator. Instead, the schedule is fitted to observed metrics (Appendix D.5). Comparing simulator-selected settings with direct metric-based tuning and tuned constants on independent training continuations would clarify its added diagnostic value.
-
The optimizer schedule’s practical advantage is not established (moderate). The proposed schedule improves validation loss by only 0.0001 over the best tested constant , without reported uncertainty (Appendix D.5). This does not establish a reproducible advantage over a well-tuned constant setting. Practical comparisons should also account for the calibration costs reported in Table 1.
-
Missing targeted ablations limit attribution of the results (moderate). Sections 2–4 include useful sensitivity experiments, but additional controls are needed to separate Gaussian sampling from the gradient-moment approximation, spectral shape from overall scale and proximity to instability, and the proposed rule from simpler schedules with comparable ranges and average decay. Matched comparisons across seeds would clarify which mechanisms drive the simulated behavior and optimization benefits (Section 3.3; Appendix D.5).
Questions for authors
-
Under what conditions can the dependence terms in Section 3.2, Eq. (5), be neglected? Could you provide an error bound and empirical checks of the full second-moment approximation, beyond the scalar check in Eq. (6)?
-
How do Figure 4.1’s second-order comparisons support prediction of actual validation-loss spikes? Could you report predictive results across additional checkpoints and runs, including missed spikes and false alarms, and clarify the discrepancies in first-order changes and recovery behavior?
-
Under which batch-size conditions does the simulator accurately reproduce spike statistics? Could you report errors in spike magnitude, frequency, and duration across independent synthetic runs to help identify its reliable operating range?
-
Could you provide spectral-shape comparisons controlling trace, sharpness, and distance to instability to help isolate the role of spectral concentration in Figure 2.1? Which observations from the Transformer experiments support the same mechanism?
-
Which assumptions limit applicability beyond the Transformer configuration in Section 4? Could you clarify the supported scope across architectures, model sizes, and data settings, including any available broader evidence and conclusions currently limited to this case study?
-
Does the simulator improve intervention choices beyond direct metric-based tuning? Where feasible, could simulator-selected settings be compared with tuned constants and simpler indicators through short, matched continuations from an existing checkpoint, considering both stability and calibration cost?
-
Is the schedule’s 0.0001 improvement consistent across independent training runs? Could you report the mean and standard deviation of final validation loss for the proposed schedule, the best fixed , and a simple schedule with comparable range and average decay?
-
Which approximation causes the errors in Table 2? Could you compare actual minibatch gradients, Gaussian gradients with exact state-dependent mean and covariance, and Gaussian gradients using the paper’s approximated moments, under matched settings, and report their spike frequency, magnitude, and duration?
-
How far in advance can the simulator provide a useful warning? Could you report prediction quality across forecast horizons, indicate when recalibration becomes necessary, and compare warning times with simpler instability indicators?
-
Can the simulator identify which layers contribute most to instability? If feasible, could interventions on predicted high-risk layers be compared with matched interventions on other layers to assess whether the layer-wise diagnosis supports effective, targeted prevention?
5 comments
Full Reply to Questions and Critical Weaknesses
Question #1 (related to Critical Weakness #1): Under what conditions can the dependence terms in Section 3.2, Eq. (5), be neglected? Could you provide an error bound and empirical checks of the full second-moment approximation, beyond the scalar check in Eq. (6)?
The colab presented above, https://colab.research.google.com/drive/19fOfDvr6TAp9a7PhZ-wo1Ndm8-cetVNB?usp=sharing , should answer this question. We have provided empirical measurements of the error. Theoretically, for the case of random feature models one can prove that the error tends to 0 as the dimension goes to infinity. We have provided a sketch of the proof in the paper as well (Appendix B.2).
Question #2 (related to Critical Weakness #2): How do Figure 4.1’s second-order comparisons support prediction of actual validation-loss spikes? Could you report predictive results across additional checkpoints and runs, including missed spikes and false alarms, and clarify the discrepancies in first-order changes and recovery behavior?
Is the colab https://colab.research.google.com/drive/19qjgh8Je16FcZJrYnNREN3QRj4iFwUWe?usp=sharing presented in our top-level comment what you ask for? The colab combined with our discussion in the paper should have documented most we know about the proposed simulation. It is what it is. We do not intend to provide a "recommended use-case" for the proposed simulation algorithm in the paper; one probably needs several extra components for empirical applications, which are strictly out of scope of this work. If merely for the empirical usage of indicating unhealthy training, several metrics that the simulation relies on, such as the distribution of the eigenvalues and the power-law exponent alpha, should be predictive already.
Question #3 (related to Critical Weakness #3): Under which batch-size conditions does the simulator accurately reproduce spike statistics? Could you report errors in spike magnitude, frequency, and duration across independent synthetic runs to help identify its reliable operating range?
The theory suggests that the condition is . Specific to our example, , so the batch-size is the condition. This kind of explains why at the simulation shows different spiking statistics. The condition appears in several places in our theory (e.g., on Page 4, 20), and we will make it more explicit in our revision of the paper.
Question #4 (related to Critical Weakness #4): Could you provide spectral-shape comparisons controlling trace, sharpness, and distance to instability to help isolate the role of spectral concentration in Figure 2.1? Which observations from the Transformer experiments support the same mechanism?
Is Figure D.2 (in Appendix D.2) what you are asking?
Question #5 (related to Critical Weakness #5) Which assumptions limit applicability beyond the Transformer configuration in Section 4? Could you clarify the supported scope across architectures, model sizes, and data settings, including any available broader evidence and conclusions currently limited to this case study?
Question #6 (related to Critical Weakness #6) Does the simulator improve intervention choices beyond direct metric-based tuning? Where feasible, could simulator-selected settings be compared with tuned constants and simpler indicators through short, matched continuations from an existing checkpoint, considering both stability and calibration cost?
Question #9 (related to Critical Weakness #6) How far in advance can the simulator provide a useful warning? Could you report prediction quality across forecast horizons, indicate when recalibration becomes necessary, and compare warning times with simpler instability indicators?
Question #10 (related to Critical Weakness #6) Can the simulator identify which layers contribute most to instability? If feasible, could interventions on predicted high-risk layers be compared with matched interventions on other layers to assess whether the layer-wise diagnosis supports effective, targeted prevention?
I cannot tell you that. (Seriously!) The paper is a case-study, and these topics are strictly beyond the scope. If the reviewer really wants to know, we need to figure out a way to collaborate, or the reviewer uses their own resources to find out.
Does the simulator help issue interventions when there are training spikes? Not really. No intervention really works when there are already training spikes. The simulator does help understand normal training and can be informative in figuring out the best hyperparameters; but that also requires many other components to work together.
Full Reply to Questions and Critical Weaknesses (cont.)
Question #7 (related to Critical Weakness #7) Is the schedule’s 0.0001 improvement consistent across independent training runs? Could you report the mean and standard deviation of final validation loss for the proposed schedule, the best fixed, and a simple schedule with comparable range and average decay?
Please see the report of our additional experiments on different beta2 settings, presented above in a top-level comment.
Question #8 (related to Critical Weakness #1) Which approximation causes the errors in Table 2? Could you compare actual minibatch gradients, Gaussian gradients with exact state-dependent mean and covariance, and Gaussian gradients using the paper’s approximated moments, under matched settings, and report their spike frequency, magnitude, and duration?
I think one misunderstanding that the reviewer might have, is that the discrepancies in spiking statistics as reported Table 2, are somehow caused by approximation errors in Prediction 5 -- they are not.
Any approximation errors in Prediction 5 will directly cause the simulation to deviate from the average case behavior of the loss. But as you can see, the average case behavior matches (i.e., the underlying flows of the curves match, when ignoring spikes), even at batch-size=64. So there are really no approximation error from Prediction 5, for high-dimensional random feature models. We showed that the empirical error is mostly less than 5% for Transformer LM; but it should be less than 1% for PLRF.
What could cause the discrepancies in spiking statistics, is the short-term gradient noise distribution being not Gaussian. This is the only reason. The two points I wrote in the paper, are meant to be the two factors that make the accurate theoretical calculation of this distribution difficult -- I don't know how to calculate the short-term gradient noise distribution at all. The comparison we present in the paper, is really the best we can do for this observation. There is no further decomposition into separate factors.
Re: Critical Weakness #1
We present a "per-eigenmode verification" in this colab: https://colab.research.google.com/drive/19fOfDvr6TAp9a7PhZ-wo1Ndm8-cetVNB?usp=sharing
Combining this with the already existing scalar verification (i.e., verification of relation Eq.(6) in Appendix D.3), we should have covered the full second-moment approximation that is asked by Reviewer PJ6M, which should be enough to theoretically support our simulation algorithm.
We would like to discuss a bit further on this intuition. Using the notation in the paper, let A=UDiag(λ)V⊤ be the singular value decomposition of A. It is true that the "dependence terms" can "remain significant" when output gradients align with eigenvectors in V. This can be seen by taking the expectation of Eq.(11) in the paper, in Appendix B.2. Where you have probably get confused, is that there are two sets of "eigenvectors", U and V, in the above singular value decomposition. By conflating U and V in one concept as "input eigenvectors", you thought the per-eigenmode version of Prediction 5 is against your intuition -- But, note that Eq.(5), the per-eigenmode version of Eq.(6), is purely based on the eigen decomposition defined by U. Because U is irrelevant to V, it almost never "aligns with output gradients". This is why when Eq.(6) is empirically verified, we can almost be sure that the per-eigenmode version Eq.(5) is likely true. The "dependence terms" can "remain significant" in the sense that, its overall norm indeed occupies a non-zero portion of E(∥∇BW∥2), which causes errors that are already checked when we verify Eq.(6) in Appendix D.3 -- you have already witnessed the most essential step. But that "dependence terms" cannot really affect the eigenmodes defined by U. Please take a deeper thought on this one, and if you have other related questions, I'm here and happy to answer them.