arXiv Machine Learning September 29, 2026 By Barak Gahtan, Ido Galil, Alex M. Bronstein

The Residual Stream's Effective Depth

Read the original on arXiv Machine Learning →

The paper introduces effective depth (Deff), a scalar diagnostic that treats a transformer’s layer‑wise residual stream as a discrete‑time process and measures how representation similarity decays with layer distance. Across sixteen decoder‑only language models, Deff reveals that most models exhibit a lower similarity decay than the closed‑form reference, indicating correlated residual updates rather than unused depth. The study also shows that this effect is robust to various controls and persists early in training, suggesting Deff is a global accumulated‑state diagnostic rather than a capability score.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

arXiv Machine Learning

Aug 14

Excess Separability: Nuisance-Controlled Residual-Stream Probing for Benchmark Contamination Detection

arXiv:2608. 12652v1 Announce Type: cross Abstract: Benchmark contamination is diagnosed today with n-gram overlap, with likelihood-based membership inference, or with canary strings, and each needs something usually unavailable: the training corpus, a well-chosen test statistic, or foresight at dataset release.

By Florian Braun

llms benchmarks

arXiv Machine Learning

1d ago

The Geometry of Inference in Transformer Residual Streams

arXiv:2609.37824v1 Announce Type: cross Abstract: Transformer language models build predictions through successive residual updates, but how their representations become specific to an eventual outco...

By Timur Mudarisov, Mikhail Burtsev, Radu State

llms safety

arXiv AI

Jul 22

An Analysis of Residual-Stream Geometry Across Transformer Depth

arXiv:2607. 18348v1 Announce Type: cross Abstract: We propose a transition-centred geometric analysis of transformer residual streams.

By Sunit Bhattacharya, Ravi Shankar Kolli

llms

arXiv Machine Learning

Sep 14

Where Decoder Cosine Similarity Fails for SAE Feature Flow Discovery

arXiv:2609. 12591v1 Announce Type: new Abstract: Foundation models are increasingly adapted through fine-tuning, model editing, and alignment procedures while retaining previously acquired capabilities.

By Hendrik Droste, Christian Medeiros Adriano, Kathrin Korte, Holger Giese

fine-tuning benchmarks safety

arXiv AI

Jul 9

At-Grok Is Not Converged:A Measurement-Validity Audit for Grokking Representation Metrics

arXiv:2607. 06639v1 Announce Type: cross Abstract: On modular arithmetic, a network's embedding keeps compressing for tens of thousands of steps after it has already generalized.

By Truong Xuan Khanh

llms rag benchmarks safety

arXiv Computation and Language

Sep 10

Through the Looking Glass: Directly Reading and Writing Transformers

The paper investigates how many transformer components influence a token prediction by measuring the absolute contribution of each unit and channel to the logit. It finds that thousands of components contribute to a single prediction, yet a small subset—often just dozens—carries the majority of the predictive mass. Across models ranging from 124 M to 7 B parameters, the proportion of the model involved in a prediction remains around one to three percent, independent of size, and the study demonstrates that specific components can be directly read and written to modify model behavior without additional training.

By Mark Oskin

llms rag

One email a morning, machine-written

Related stories

Excess Separability: Nuisance-Controlled Residual-Stream Probing for Benchmark Contamination Detection

The Geometry of Inference in Transformer Residual Streams

An Analysis of Residual-Stream Geometry Across Transformer Depth

Where Decoder Cosine Similarity Fails for SAE Feature Flow Discovery

At-Grok Is Not Converged:A Measurement-Validity Audit for Grokking Representation Metrics

Through the Looking Glass: Directly Reading and Writing Transformers