REVIEW 13 cited by
The emergence of clusters in self-attention dynamics
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
The emergence of clusters in self-attention dynamics
read the original abstract
Viewing Transformers as interacting particle systems, we describe the geometry of learned representations when the weights are not time dependent. We show that particles, representing tokens, tend to cluster toward particular limiting objects as time tends to infinity. Cluster locations are determined by the initial tokens, confirming context-awareness of representations learned by Transformers. Using techniques from dynamical systems and partial differential equations, we show that the type of limiting object that emerges depends on the spectrum of the value matrix. Additionally, in the one-dimensional case we prove that the self-attention matrix converges to a low-rank Boolean matrix. The combination of these results mathematically confirms the empirical observation made by Vaswani et al. [VSP'17] that leaders appear in a sequence of tokens when processed by Transformers.
Forward citations
Cited by 13 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.
-
Transient Reserves, Sink Dampers, and the Failure of Eigenvalue Reasoning in the Attention Propagator
Resolvent analysis of trained causal attention shows sinks act as transient dampers, routing heads carry excess Kreiss reserve, and eigenvalue depth predictions fail by 7–11 orders of magnitude.
-
Phase transitions for the noisy transformer model in arbitrary dimension
In every dimension d≥2 there exists a unique β_*^{(d)}>0 such that the uniform density on the sphere is the unique global minimizer of the USA free energy up to the linear-stability threshold K_# for β≤β_*, yielding a...
-
Gradient Flow Structure and Quantitative Dynamics of Multi-Head Self-Attention
Multi-head self-attention is modeled as a gradient flow with a non-decreasing energy functional under conditions on score matrices, yielding closed-form clustering thresholds in simplified regimes and monotonic entrop...
-
Clustering in pure-attention hardmax transformers and its role in sentiment analysis
Hardmax transformers converge to leader-determined clusters, enabling an interpretable model for sentiment analysis.
-
Many-body Tipping Dynamics of ChatGPT-like AIs
Tipping of ChatGPT-like AI to undesirable outputs is modeled as first-passage transport of a residual-state spin across an output-basin wall, with attention disorder controlling the crossing.
-
Self-Attention Dynamics with Rotary Position Embeddings: Twisted States and Explicit Consensus Rates on the Sphere
Normalized query/key-only RoPE attention on the sphere has reversible consensus kernels with exact Bessel-aliasing spectra, explicit regional contraction rates from a sharp softmax floor, and RoPE-selected twisted equ...
-
On Transformer Dynamics
A universal, finitely parametrized family of geometric interaction laws realizes any prescribed attention digraph, with cost governed by the biclique cover number and a new hub-chromatic index.
-
Correcting Influence: Unboxing LLM Outputs with Orthogonal Latent Spaces
A latent mediation framework with sparse autoencoders enables non-additive token-level influence attribution in LLMs by learning orthogonal features and back-propagating attributions.
-
Gradient Flow Structure and Quantitative Dynamics of Multi-Head Self-Attention
Multi-head self-attention dynamics admit a non-decreasing energy functional under suitable score-matrix conditions, with closed-form clustering thresholds and monotonic entropy production in simplified regimes.
-
Krause Synchronization Transformers
Krause Attention replaces global softmax self-attention with localized, distance-based bounded-confidence interactions to promote local synchronization, reduce complexity to linear in sequence length, and alleviate at...
-
What Makes Position Zero Special? A Mechanistic Study of Position Zero Attention Sinks in LLMs
Position-zero attention sinks in transformers emerge from causal-masking asymmetry: position zero attends only to itself, and an MLP then amplifies its representation into a stable, high-norm 'sink'.
-
Krause Synchronization Transformers
Krause Attention, a local distance-based attention with top-k sparsity, improves accuracy and efficiency on vision and language benchmarks and, per its analysis, promotes multi-cluster rather than global token synchro...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.