REVIEW 10 cited by
Dynamics of Transient Structure in In-Context Linear Regression Transformers
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Dynamics of Transient Structure in In-Context Linear Regression Transformers
read the original abstract
Modern deep neural networks display striking examples of rich internal computational structure. Uncovering principles governing the development of such structure is a priority for the science of deep learning. In this paper, we explore the transient ridge phenomenon: when transformers are trained on in-context linear regression tasks with intermediate task diversity, they initially behave like ridge regression before specializing to the tasks in their training distribution. This transition from a general solution to a specialized solution is revealed by joint trajectory principal component analysis. Further, we draw on the theory of Bayesian internal model selection to suggest a general explanation for the phenomena of transient structure in transformers, based on an evolving tradeoff between loss and complexity. We empirically validate this explanation by measuring the model complexity of our transformers as defined by the local learning coefficient.
Forward citations
Cited by 10 Pith papers
-
Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior
Manifold steering along activation geometry induces behavioral trajectories matching the natural manifold of outputs, while linear steering produces off-manifold unnatural behaviors.
-
Task Vector Geometry Underlies Dual Modes of Task Inference in Transformers
In a controlled synthetic setting, transformers implement in-distribution task inference via convex combinations of task vectors and out-of-distribution inference via nearly orthogonal extrapolative representations.
-
Influence Dynamics and Stagewise Data Attribution
Using Bayesian influence functions and singular learning theory, the authors show that a sample's influence on a model varies non-monotonically over training, peaking and flipping sign at phase transitions.
-
What does a Bayes-filtered transformer believe? A predictive Monte Carlo approach
Predictive Monte Carlo recovers a Bayes-filtered transformer's implicit prior and posterior over latent tasks from next-token rollouts alone; the task-diversity threshold and transient generalization appear in this re...
-
Temporal Task Diversity: Inductive Biases Under Non-Stationarity in Synthetic Sequence Modelling
Temporal diversity in task distribution during training increases generalization bias over memorization in transformers for in-context linear regression.
-
Mechanistic Anomaly Detection via Functional Attribution
Functional attribution with influence functions detects anomalous mechanisms in neural networks, achieving SOTA backdoor detection (average DER 0.93) on vision benchmarks and improvements on LLMs.
-
Structure and Scale in Simplicial Sequence Modelling
Small transformers on HMM prediction tasks exhibit correlated scaling between performance and linear encoding of belief distributions in residual activations.
-
Mechanistic Anomaly Detection via Functional Attribution
A grid-free attention-based deep learning model interpolates surface-level PM2.5 concentrations across the US using sparse sensor data and auxiliary geospatial features.
-
High-Dimensional Statistics: Reflections on Progress and Open Problems
A survey synthesizing representative advances, common themes, and open problems in high-dimensional statistics while pointing to key entry-point works.
-
High-Dimensional Statistics: Reflections on Progress and Open Problems
This review synthesizes representative advances in high-dimensional statistics, highlights common themes and open problems, and points to key entry works.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.