The large learning rate phase of deep learning: the catapult mechanism | Paper Citation Record

Observation aa995168-f597-4e21-9019-e9e4d2652a75 · inbound

Scaling Laws for Autoregressive Generative Modeling cites this paper.

Scaling Laws for Autoregressive Generative Modeling The large learning rate phase of deep learning: the catapult mechanism

Reference 11

Resolution: verified exact; arxiv_id, observed 2026-05-13T07:49:43.819446Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 19ece7e6-6614-46d0-b7da-52a487f6eaa4 · inbound

Scaling Laws for Transfer cites this paper.

Scaling Laws for Transfer The large learning rate phase of deep learning: the catapult mechanism

Reference 65

Resolution: metadata mismatch; arxiv_id, observed 2026-05-18T00:58:13.693119Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 617e79ba-f6d4-4378-84b8-de100a8cfc73 · inbound

A General Language Assistant as a Laboratory for Alignment cites this paper.

A General Language Assistant as a Laboratory for Alignment The large learning rate phase of deep learning: the catapult mechanism

Reference 94

Resolution: metadata mismatch; arxiv_id, observed 2026-05-11T14:22:58.981470Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 6261559f-2e0e-4b22-b45a-4d8df32b9d36 · inbound

Language Models (Mostly) Know What They Know cites this paper.

Language Models (Mostly) Know What They Know The large learning rate phase of deep learning: the catapult mechanism

Reference 152

Resolution: metadata mismatch; arxiv_id, observed 2026-05-10T15:42:47.810297Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation a15c71d2-eff2-4c84-a603-85886ace9913 · inbound

A ghost mechanism: An analytical model of abrupt learning in recurrent networks cites this paper.

A ghost mechanism: An analytical model of abrupt learning in recurrent networks The large learning rate phase of deep learning: the catapult mechanism

Reference 51

Resolution: verified exact; arxiv_id, observed 2026-05-23T06:32:38.851832Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 2b72b369-736d-41cd-b45f-90aec27f8d62 · inbound

Right Time to Learn:Promoting Generalization via Bio-inspired Spacing Effect in Knowledge Distillation cites this paper.

Right Time to Learn:Promoting Generalization via Bio-inspired Spacing Effect in Knowledge Distillation The large learning rate phase of deep learning: the catapult mechanism

Reference 2019

Resolution: unresolved; no resolver link, observed 2026-08-08T16:31:36.516166Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9f45d863-9706-4ebf-8ea7-1d696367aee8 · inbound

Adaptive Preconditioners Trigger Loss Spikes in Adam cites this paper.

Adaptive Preconditioners Trigger Loss Spikes in Adam The large learning rate phase of deep learning: the catapult mechanism

Reference 22

Resolution: unresolved; no resolver link, observed 2026-08-07T10:42:54.707684Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2d4679b0-9100-4b52-8d3a-6bd1049139c2 · inbound

Large Spikes in Stochastic Gradient Descent: A Large-Deviations View cites this paper.

Large Spikes in Stochastic Gradient Descent: A Large-Deviations View The large learning rate phase of deep learning: the catapult mechanism

Reference 35

Resolution: verified exact; arxiv_id, observed 2026-05-15T13:30:51.043994Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 6f90fd1c-8a12-40b0-bdc5-5f8197777bfc · inbound

(How) Learning Rates Regulate Catastrophic Overtraining cites this paper.

(How) Learning Rates Regulate Catastrophic Overtraining The large learning rate phase of deep learning: the catapult mechanism

Reference 13

Resolution: verified exact; arxiv_id, observed 2026-05-10T13:40:26.497872Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 758cdf53-4afe-46cd-94f4-7c0e3a9d543b · inbound

(How) Learning Rates Regulate Catastrophic Overtraining cites this paper.

(How) Learning Rates Regulate Catastrophic Overtraining The large learning rate phase of deep learning: the catapult mechanism

Reference 13

Resolution: unresolved; no resolver link, observed 2026-08-04T05:30:35.676773Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation abbf4d66-e621-4a0d-b426-2cab6f18c99e · inbound

Zeroth-Order Optimization at the Edge of Stability cites this paper.

Zeroth-Order Optimization at the Edge of Stability The large learning rate phase of deep learning: the catapult mechanism

Reference 3

Resolution: verified exact; arxiv_id, observed 2026-05-10T11:20:10.144195Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation f1f63203-12e3-4bab-a3b4-d460b16298df · inbound

Zeroth-Order Optimization at the Edge of Stability cites this paper.

Zeroth-Order Optimization at the Edge of Stability The large learning rate phase of deep learning: the catapult mechanism

Reference 3

Resolution: unresolved; no resolver link, observed 2026-07-12T20:06:39.488359Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9f1484d4-1a70-4671-b587-2d69ce24461c · inbound

The Origin of Edge of Stability cites this paper.

The Origin of Edge of Stability The large learning rate phase of deep learning: the catapult mechanism

Reference 39

Resolution: metadata mismatch; arxiv_id, observed 2026-05-11T13:36:07.948407Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 9602fa3f-807d-4725-8944-0d29c099e53f · inbound

SGD at the Edge of Stability: The Stochastic Sharpness Gap cites this paper.

SGD at the Edge of Stability: The Stochastic Sharpness Gap The large learning rate phase of deep learning: the catapult mechanism

Reference 58

Resolution: verified exact; arxiv_id, observed 2026-05-10T00:59:49.391917Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 82f3b924-b8a0-420e-8775-d27a2173f4d6 · inbound

There Will Be a Scientific Theory of Deep Learning cites this paper.

There Will Be a Scientific Theory of Deep Learning The large learning rate phase of deep learning: the catapult mechanism

Reference 46

Resolution: metadata mismatch; arxiv_id, observed 2026-05-11T15:21:08.996891Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation fa3bea90-131e-48d7-94a2-4a0d1c1a8b4d · inbound

Finite-Size Gradient Transport in Large Language Model Pretraining: From Cascade Size to Intensive Transport Efficiency cites this paper.

Finite-Size Gradient Transport in Large Language Model Pretraining: From Cascade Size to Intensive Transport Efficiency The large learning rate phase of deep learning: the catapult mechanism

Reference 12

Resolution: verified exact; arxiv_id, observed 2026-05-10T15:00:31.040454Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation f0c3f25a-a64b-4ad5-9fd7-6b6673039343 · inbound

Endogenous Regime Switching Driven by Scalar-Irreducible Learning Dynamics cites this paper.

Endogenous Regime Switching Driven by Scalar-Irreducible Learning Dynamics The large learning rate phase of deep learning: the catapult mechanism

Reference 7

Resolution: verified exact; arxiv_id, observed 2026-05-11T06:36:01.856918Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 6bbaf906-a3d7-44ce-9fde-e2b624d0f59e · inbound

A Rod Flow Model for Adam at the Edge of Stability cites this paper.

A Rod Flow Model for Adam at the Edge of Stability The large learning rate phase of deep learning: the catapult mechanism

Reference 11

Resolution: metadata mismatch; arxiv_id, observed 2026-05-11T05:00:56.647680Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 22c671cb-62d2-45e7-a830-c0bd1247e408 · inbound

Can Muon Fine-tune Adam-Pretrained Models? cites this paper.

Can Muon Fine-tune Adam-Pretrained Models? The large learning rate phase of deep learning: the catapult mechanism

Reference 77

Resolution: metadata mismatch; arxiv_id, observed 2026-05-12T06:51:28.677058Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation dcff6084-c81d-47e3-a253-5b8e87ae07eb · inbound

How to Scale Mixture-of-Experts: From muP to the Maximally Scale-Stable Parameterization cites this paper.

How to Scale Mixture-of-Experts: From muP to the Maximally Scale-Stable Parameterization The large learning rate phase of deep learning: the catapult mechanism

Reference 90

Resolution: metadata mismatch; arxiv_id, observed 2026-05-15T04:49:44.826858Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 49dd278e-4c23-412f-8a62-8c5246844818 · inbound

Fine-Tuning Without Forgetting via Loss-Adaptive Learning Rates cites this paper.

Fine-Tuning Without Forgetting via Loss-Adaptive Learning Rates The large learning rate phase of deep learning: the catapult mechanism

Reference 33

Resolution: verified exact; arxiv_id, observed 2026-05-20T07:18:07.041459Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation d622889e-88bb-4c7b-a0d8-efc3a61dd255 · inbound

Large-Step Training Dynamics of a Two-Factor Linear Transformer Model cites this paper.

Large-Step Training Dynamics of a Two-Factor Linear Transformer Model The large learning rate phase of deep learning: the catapult mechanism

Reference 12

Resolution: verified exact; arxiv_id, observed 2026-05-21T03:53:56.329810Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation a38262b2-cb62-44b6-a9ba-85ffa08037ec · inbound

Accurate Large-sample Uncertainty Quantification using Stochastic Gradient Markov Chain Monte Carlo cites this paper.

Accurate Large-sample Uncertainty Quantification using Stochastic Gradient Markov Chain Monte Carlo The large learning rate phase of deep learning: the catapult mechanism

Reference 39

Resolution: verified exact; arxiv_id, observed 2026-07-01T19:16:01.182666Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation aea36370-95e3-4e8b-a880-45dff5b07730 · inbound

Edge of Stability Selectively Shapes Learning Across the Data Distribution cites this paper.

Edge of Stability Selectively Shapes Learning Across the Data Distribution The large learning rate phase of deep learning: the catapult mechanism

Reference 35

Resolution: verified exact; arxiv_id, observed 2026-07-02T02:56:29.501282Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 5c7b22c6-ee91-4a3a-b8ca-083ab82e857a · inbound

Gradient Descent with Large Step Size Restores Symmetry in Deep Linear Networks with Multi-Pathway cites this paper.

Gradient Descent with Large Step Size Restores Symmetry in Deep Linear Networks with Multi-Pathway The large learning rate phase of deep learning: the catapult mechanism

Reference 26

Resolution: verified exact; arxiv_id, observed 2026-06-28T23:22:46.406450Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 3f38c97f-f7c8-48c7-bcb0-6ee12b2e4c95 · inbound

Edge Flow: A Tractable and Predictive Continuous-Time Model for Gradient Descent at the Edge of Stability cites this paper.

Edge Flow: A Tractable and Predictive Continuous-Time Model for Gradient Descent at the Edge of Stability The large learning rate phase of deep learning: the catapult mechanism

Reference 1

Resolution: verified exact; arxiv_id, observed 2026-07-03T20:28:55.905813Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation af75fc6f-3063-471f-97e7-19aaa4a2d3d2 · inbound

Gradient-Descent Steps to Success over Mean Accuracy: A Paradigm Shift for ML cites this paper.

Gradient-Descent Steps to Success over Mean Accuracy: A Paradigm Shift for ML The large learning rate phase of deep learning: the catapult mechanism

Reference 9

Resolution: verified exact; arxiv_id, observed 2026-07-04T07:49:39.677056Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 147cf3b2-b3b7-46b3-a6fb-387931b0fe8e · inbound

Convergence of Gradient Descent for General Neural Network Architectures Beyond the NTK Regime cites this paper.

Convergence of Gradient Descent for General Neural Network Architectures Beyond the NTK Regime The large learning rate phase of deep learning: the catapult mechanism

Reference 49

Resolution: metadata mismatch; arxiv_id, observed 2026-07-04T10:29:44.290800Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation f5590f2d-b115-4234-8a37-eca57540006c · inbound

Implicit Bias of SGD in Multivariate ReLU Networks: Effective Width Collapse cites this paper.

Implicit Bias of SGD in Multivariate ReLU Networks: Effective Width Collapse The large learning rate phase of deep learning: the catapult mechanism

Reference 189

Resolution: unresolved; no resolver link, observed 2026-07-12T01:11:50.444218Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 99af2c4f-03dd-481a-b5fc-2eeb432f6685 · inbound

Directional Curvature from Armijo Backtracking: A Low-Cost Sharpness Probe and a Calibration-Free Learning-Rate Safeguard for Adam cites this paper.

Directional Curvature from Armijo Backtracking: A Low-Cost Sharpness Probe and a Calibration-Free Learning-Rate Safeguard for Adam The large learning rate phase of deep learning: the catapult mechanism

Reference 15

Resolution: unresolved; no resolver link, observed 2026-07-11T22:26:03.646701Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 69446ef4-97dc-40ba-8097-aa60a77e02ff · inbound

Directional Curvature from Armijo Backtracking: A Low-Cost Sharpness Probe and a Calibration-Free Learning-Rate Safeguard for Adam cites this paper.

Directional Curvature from Armijo Backtracking: A Low-Cost Sharpness Probe and a Calibration-Free Learning-Rate Safeguard for Adam The large learning rate phase of deep learning: the catapult mechanism

Reference 17

Resolution: unresolved; no resolver link, observed 2026-07-14T16:26:22.217462Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation bcb3f7be-9112-4054-a632-51be920dd956 · inbound

Directional Curvature from Armijo Backtracking: A Low-Cost Sharpness Probe and a Calibration-Free Learning-Rate Safeguard for Adam cites this paper.

Directional Curvature from Armijo Backtracking: A Low-Cost Sharpness Probe and a Calibration-Free Learning-Rate Safeguard for Adam The large learning rate phase of deep learning: the catapult mechanism

Reference 17

Resolution: unresolved; no resolver link, observed 2026-08-02T08:48:43.532111Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 8ed05444-e79c-417c-b07c-cc62055f153e · inbound

The Map Behind the Flow: Finite-Step Gradient Descent as a Dynamical System cites this paper.

The Map Behind the Flow: Finite-Step Gradient Descent as a Dynamical System The large learning rate phase of deep learning: the catapult mechanism

Reference 9

Resolution: unresolved; no resolver link, observed 2026-07-11T10:24:00.719150Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 71e7f561-54f2-4a81-ad47-b6e249b99043 · inbound

Weight-norm Criticality: A Mechanism for Loss Spikes Induced by the Normalization and Weight Decay cites this paper.

Weight-norm Criticality: A Mechanism for Loss Spikes Induced by the Normalization and Weight Decay The large learning rate phase of deep learning: the catapult mechanism

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-01T08:50:41.724748Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ce33ad15-ee68-4472-b792-648ee5d9b15b · inbound

A Defense of the Quadratic Model cites this paper.

A Defense of the Quadratic Model The large learning rate phase of deep learning: the catapult mechanism

Reference 12

Resolution: unresolved; no resolver link, observed 2026-08-01T06:58:09.772532Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 54c77455-4e96-4490-aced-f7b75d138663 · inbound

The Fourth Quadrant: A Stylized View of Benign Misfitting cites this paper.

The Fourth Quadrant: A Stylized View of Benign Misfitting The large learning rate phase of deep learning: the catapult mechanism

Reference 118

Resolution: unresolved; no resolver link, observed 2026-08-06T00:43:14.981584Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Coverage vector

Reference resolution

Outbound references

Pith citing papers