REVIEW 8 cited by
Spectral Filters, Dark Signals, and Attention Sinks
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
read the original abstract
Projecting intermediate representations onto the vocabulary is an increasingly popular interpretation tool for transformer-based LLMs, also known as the logit lens. We propose a quantitative extension to this approach and define spectral filters on intermediate representations based on partitioning the singular vectors of the vocabulary embedding and unembedding matrices into bands. We find that the signals exchanged in the tail end of the spectrum are responsible for attention sinking (Xiao et al. 2023), of which we provide an explanation. We find that the loss of pretrained models can be kept low despite suppressing sizable parts of the embedding spectrum in a layer-dependent way, as long as attention sinking is preserved. Finally, we discover that the representation of tokens that draw attention from many tokens have large projections on the tail end of the spectrum.
Forward citations
Cited by 8 Pith papers
-
When Sinks Help or Hurt: Unified Framework for Attention Sink in Large Vision-Language Models
Attention sinks in LVLM create a global-vs-local trade-off that a layer-wise gating module can balance to improve multimodal benchmark performance.
-
When Sinks Help or Hurt: Unified Framework for Attention Sink in Large Vision-Language Models
A V-sink/L-sink taxonomy plus a frozen-backbone, NTP-trained layer-wise key gate (LSG) improves LLaVA-1.5-7B by up to +1.55pp on MMStar and +3.08pp on CVBench.
-
Contribution Weights: A Geometrical Analysis of Self-Attention Transformers
Contribution Weights combine attention, value magnitude, and directional alignment to measure token influence more faithfully than attention alone, and show attention sinks actively suppress information via a convex s...
-
Vocabulary Hijacking in LVLMs: Unveiling Critical Attention Heads by Excluding Inert Tokens to Mitigate Hallucination
LVLMs show vocabulary hijacking by inert tokens that decode to hijacking anchors; HABI locates them, NHAR finds resilient heads, and HAVAE boosts those heads to cut hallucinations.
-
The Structural Origin of Attention Sink: Variance Discrepancy, Super Neurons, and Dimension Disparity
Attention sinks arise from variance discrepancy in self-attention value aggregation, amplified by super neurons and first-token dimension disparity, and can be mitigated by head-wise RMSNorm to accelerate pre-training...
-
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'.
-
When Attention Sink Emerges in Language Models: An Empirical View
Attention sinks emerge in language models from softmax-induced token dependence on attention scores and do not appear when using sigmoid attention without normalization in models up to 1B parameters.
-
When Sinks Help or Hurt: Unified Framework for Attention Sink in Large Vision-Language Models
Layer-wise Sink Gating scales vision and LLM attention sinks in LVLMs to balance global priors and local evidence, improving multimodal benchmarks with a frozen backbone.
Discussion (0). Sign in to comment.