REVIEW 6 cited by
Critical attention scaling in long-context 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
Critical attention scaling in long-context transformers
read the original abstract
As large language models scale to longer contexts, attention layers suffer from a fundamental pathology: attention scores collapse toward uniformity as context length $n$ increases, causing tokens to cluster excessively, a phenomenon known as rank-collapse. While $\textit{attention scaling}$ effectively addresses this deficiency by rescaling attention scores with a polylogarithmic factor $\beta_n$, theoretical justification for this approach remains lacking. We analyze a simplified yet tractable model that magnifies the effect of attention scaling. In this model, attention exhibits a phase transition governed by the scaling factor $\beta_n$: insufficient scaling collapses all tokens to a single direction, while excessive scaling reduces attention to identity, thereby eliminating meaningful interactions between tokens. Our main result identifies the critical scaling $\beta_n \asymp \log n$ and provides a rigorous justification for attention scaling in YaRN and Qwen, clarifying why logarithmic scaling maintains sparse, content-adaptive attention at large context lengths.
Forward citations
Cited by 6 Pith papers
-
A Unified Framework for Critical Scaling of Inverse Temperature in Self-Attention
The upper-tail accumulation scale derived from the gap-counting function N_n sets the critical inverse temperature for softmax attention concentration, unifying prior conflicting laws as special cases of different N_n.
-
On the Diverse Dynamical Behaviors Arising in Deep Linear Transformers
Linear self-attention in 2D reduces to a second-harmonic Kuramoto model whose order parameter obeys one ODE; explicit matrix conditions yield clustering, rotating clusters, Hamiltonian oscillations, and bifurcations.
-
FreeSpec: Training-Free Long Video Generation via Singular-Spectrum Reconstruction
FreeSpec uses SVD-based spectral reconstruction to fuse global low-rank and local high-rank features, reducing content drift and preserving temporal dynamics in long video generation.
-
Perceptrons and localization of attention's mean-field landscape
In the mean-field limit of attention with perceptron blocks, critical points of the energy landscape are generically atomic and localized on subsets of the unit sphere.
-
Quantifying Concentration Phenomena of Mean-Field Transformers in the Low-Temperature Regime
In the low-temperature regime, the token distribution in mean-field transformers concentrates onto the push-forward under a key-query-value projection with Wasserstein distance scaling as √(log(β+1)/β) exp(Ct) + exp(-ct).
-
Visual-Language-Guided Task Planning for Horticultural Robots
A vision-language model drives a simulated greenhouse robot through simple crop-inspection tasks with ~87% success, but long multi-target tasks collapse to under 10% success.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.