Paper Citation Record · LEDGER

What are you sinking? A geometric approach on attention sink

As of 19 August 2026, this Paper Citation Record lists 17 of 17 outbound references and 13 inbound Pith citation observations for arXiv:2508.02546.

A citation records a reference. It does not transfer a finding from one paper to another.

pith.paper-citation-record.v1
2508.02546 v1

Coverage vector

measured 17 of 17 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-08-06T05:02:17.520047Z

measured 30 of 30 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-18T06:34:40.430872+00:00

measured 13 of 13 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links, observed 2026-08-02T17:04:24.334432Z

measured 0 of 1 external citation measurements

A source-named dated measurement, never combined with another source.

Source: arxiv_reference, observed 2026-07-01T22:36:17.024576Z

Reference resolution

17 of 17 outbound references displayed

verified exact3
verified fuzzy2
unresolved11
parse uncertain0
malformed identifier1
metadata mismatch0

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation ff4d1241-4470-4032-bc16-0aad637d2030 · outbound

This paper cites Why do LLMs attend to the first token?.

What are you sinking? A geometric approach on attention sink Why do LLMs attend to the first token?

Reference 1

Resolution: unresolved; no resolver link, observed 2026-08-06T05:02:17.448635Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation f80c0435-9951-4d17-aa5d-fc8a11f3bdd4 · outbound

This paper cites Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation.

What are you sinking? A geometric approach on attention sink Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation

Reference 6

Resolution: unresolved; no resolver link, observed 2026-08-06T05:02:17.471437Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 439cfdfe-36da-495c-a9f2-08c04cd7646f · outbound

This paper cites Beyond Position: the emergence of wavelet-like properties in Transformers.

What are you sinking? A geometric approach on attention sink Beyond Position: the emergence of wavelet-like properties in Transformers

Reference 7

Resolution: verified exact; local_arxiv, observed 2026-08-06T05:02:17.713076Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation 0b9c8182-960b-4e86-a81f-3278f5ed6c74 · outbound

This paper cites Learning High-Frequency Functions Made Easy with Sinusoidal Positional Encoding.

What are you sinking? A geometric approach on attention sink Learning High-Frequency Functions Made Easy with Sinusoidal Positional Encoding

Reference 8

Resolution: unresolved; no resolver link, observed 2026-08-06T05:02:17.480424Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 951c1050-658c-4ed2-94ce-fa2741b81036 · outbound

This paper cites The information bottleneck method.

What are you sinking? A geometric approach on attention sink The information bottleneck method

Reference 9

Resolution: unresolved; no resolver link, observed 2026-08-06T05:02:17.484806Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 873d8730-bb06-4622-94d8-2a7b9d02cd2a · outbound

This paper cites an unresolved cited work.

What are you sinking? A geometric approach on attention sink Unresolved cited work

Reference 16

Resolution: unresolved; raw_fallback, observed 2026-08-06T05:02:17.832957Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation d55be1aa-f0e1-453f-9b64-f4dbb50cf30f · outbound

This paper cites layer 5 in BERT), indicating that larger models can maintain the initial coordinate system longer before transitioning to the integration phase.

What are you sinking? A geometric approach on attention sink layer 5 in BERT), indicating that larger models can maintain the initial coordinate system longer before transitioning to the integration phase

Reference 19

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T05:02:17.846822Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation e566a805-7c24-423d-9132-7c1fed8c0960 · outbound

This paper cites inverted U.

What are you sinking? A geometric approach on attention sink inverted U

Reference 21

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T05:02:17.874782Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation 617f5fe5-37dd-4b02-9632-fd260a1f2eb9 · outbound

This paper cites an unresolved cited work.

What are you sinking? A geometric approach on attention sink Unresolved cited work

Reference 26

Resolution: verified exact; raw_fallback, observed 2026-08-06T05:02:17.648313Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation 91ef5f96-460f-40aa-b49d-2739c38c6499 · outbound

This paper cites an unresolved cited work.

What are you sinking? A geometric approach on attention sink Unresolved cited work

Reference 27

Resolution: unresolved; raw_fallback, observed 2026-08-06T05:02:17.861307Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation 31c82b98-1349-4138-827b-02f9559b4682 · outbound

This paper cites an unresolved cited work.

What are you sinking? A geometric approach on attention sink Unresolved cited work

Reference 35

Resolution: unresolved; raw_fallback, observed 2026-08-06T05:02:17.818953Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation 5257c462-f981-4186-a987-7f1cca06a322 · outbound

This paper cites Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned.

What are you sinking? A geometric approach on attention sink Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned

Reference 2000

Resolution: unresolved; no resolver link, observed 2026-08-06T05:02:17.488894Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 7cf4a7b5-f628-4810-b7ef-51113e237266 · outbound

This paper cites For each checkpoint during training, we extracted attention matrices from all layers and heads, computed their eigendecomposition, and tracked the evolution of these metrics.

What are you sinking? A geometric approach on attention sink For each checkpoint during training, we extracted attention matrices from all layers and heads, computed their eigendecomposition, and tracked the evolution of these metrics

Reference 2019

Resolution: malformed identifier; raw_fallback, observed 2026-08-06T05:02:17.889415Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation d70b2dc0-3368-4edd-8f11-bd32aeb06a8e · outbound

This paper cites Relative representations enable zero-shot latent space communication.

What are you sinking? A geometric approach on attention sink Relative representations enable zero-shot latent space communication

Reference 2020

Resolution: unresolved; no resolver link, observed 2026-08-06T05:02:17.466914Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 68ea8d55-1106-457b-a093-5c08e6ba8130 · outbound

This paper cites When Attention Sink Emerges in Language Models: An Empirical View.

What are you sinking? A geometric approach on attention sink When Attention Sink Emerges in Language Models: An Empirical View

Reference 2021

Resolution: unresolved; no resolver link, observed 2026-08-06T05:02:17.457751Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6415500a-423c-4bb5-a8d1-30dc0664fdb0 · outbound

This paper cites Quantifying Context Mixing in Transformers.

What are you sinking? A geometric approach on attention sink Quantifying Context Mixing in Transformers

Reference 2024

Resolution: verified exact; local_arxiv, observed 2026-08-06T05:02:17.761767Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation 0411558a-34f9-47f4-972b-e806a2042f9a · outbound

This paper cites Spectral Filters, Dark Signals, and Attention Sinks.

What are you sinking? A geometric approach on attention sink Spectral Filters, Dark Signals, and Attention Sinks

Reference 2025

Resolution: unresolved; no resolver link, observed 2026-08-06T05:02:17.453309Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Pith citing papers

Observation 42458b3e-bb50-494e-86d3-36d84ce231a1 · inbound

Scene Generation at Absolute Scale: Utilizing Semantic and Geometric Guidance From Text for Accurate and Interpretable 3D Indoor Scene Generation cites this paper.

Scene Generation at Absolute Scale: Utilizing Semantic and Geometric Guidance From Text for Accurate and Interpretable 3D Indoor Scene Generation What are you sinking? A geometric approach on attention sink

Reference 20

Resolution: unresolved; no resolver link, observed 2026-07-14T21:37:04.895721Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3d5600c3-c67c-456b-83f1-492c1311749b · inbound

When Sinks Help or Hurt: Unified Framework for Attention Sink in Large Vision-Language Models cites this paper.

When Sinks Help or Hurt: Unified Framework for Attention Sink in Large Vision-Language Models What are you sinking? A geometric approach on attention sink

Reference 32

Resolution: metadata mismatch; arxiv_id, observed 2026-05-13T23:23:26.817989Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation 2f9c1edc-a811-4f5a-b270-ced1340ca7c9 · inbound

When Sinks Help or Hurt: Unified Framework for Attention Sink in Large Vision-Language Models cites this paper.

When Sinks Help or Hurt: Unified Framework for Attention Sink in Large Vision-Language Models What are you sinking? A geometric approach on attention sink

Reference 32

Resolution: unresolved; no resolver link, observed 2026-08-02T17:04:24.334432Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 058012c7-eeb7-4738-816f-a4485e5dc497 · inbound

SinkTrack: Attention Sink based Context Anchoring for Large Language Models cites this paper.

SinkTrack: Attention Sink based Context Anchoring for Large Language Models What are you sinking? A geometric approach on attention sink

Reference 13

Resolution: verified exact; arxiv_id, observed 2026-05-11T09:51:01.052651Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation a199941f-9665-4718-9f86-0167199ca53e · inbound

SinkTrack: Attention Sink based Context Anchoring for Large Language Models cites this paper.

SinkTrack: Attention Sink based Context Anchoring for Large Language Models What are you sinking? A geometric approach on attention sink

Reference 13

Resolution: verified exact; arxiv_id, observed 2026-05-21T01:39:22.630672Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation b30700eb-2bc6-49c3-b85b-879086e3a713 · inbound

A Mechanistic Account of Attention Sinks in GPT-2: One Circuit, Broader Implications for Mitigation cites this paper.

A Mechanistic Account of Attention Sinks in GPT-2: One Circuit, Broader Implications for Mitigation What are you sinking? A geometric approach on attention sink

Reference 4

Resolution: metadata mismatch; arxiv_id, observed 2026-05-10T11:10:09.236858Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation ee00f45c-ab82-4ac1-ba67-10647c69643a · inbound

Parameter Efficiency Is Not Memory Efficiency: Rethinking Fine-Tuning for On-Device LLM Adaptation cites this paper.

Parameter Efficiency Is Not Memory Efficiency: Rethinking Fine-Tuning for On-Device LLM Adaptation What are you sinking? A geometric approach on attention sink

Reference 3

Resolution: verified exact; arxiv_id, observed 2026-05-13T20:38:15.045078Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation d52557bd-daee-4ce7-8838-8514ff43e7b6 · inbound

A Single Layer to Explain Them All:Understanding Massive Activations in Large Language Models cites this paper.

A Single Layer to Explain Them All:Understanding Massive Activations in Large Language Models What are you sinking? A geometric approach on attention sink

Reference 17

Resolution: verified exact; arxiv_id, observed 2026-05-12T07:36:41.328382Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation 6b193806-a39b-433c-bc6f-b1f01da6483e · inbound

A Single Layer to Explain Them All:Understanding Massive Activations in Large Language Models cites this paper.

A Single Layer to Explain Them All:Understanding Massive Activations in Large Language Models What are you sinking? A geometric approach on attention sink

Reference 17

Resolution: verified exact; arxiv_id, observed 2026-05-14T21:19:29.102910Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation 2be136a7-154b-4666-a0f7-8725af75d026 · inbound

Retrieval and competition: how a protein foundation model starts a protein cites this paper.

Retrieval and competition: how a protein foundation model starts a protein What are you sinking? A geometric approach on attention sink

Reference 16

Resolution: verified exact; arxiv_id, observed 2026-05-21T00:13:53.235015Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation 0359e233-667e-47d2-9a96-bf1ac1fea572 · inbound

Retrieval and competition: how a protein foundation model starts a protein cites this paper.

Retrieval and competition: how a protein foundation model starts a protein What are you sinking? A geometric approach on attention sink

Reference 16

Resolution: verified exact; arxiv_id, observed 2026-07-01T00:25:10.456684Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation 619d8b94-a283-41b2-b083-726587c39371 · inbound

ASAP: Attention Sink Anchored Pruning cites this paper.

ASAP: Attention Sink Anchored Pruning What are you sinking? A geometric approach on attention sink

Reference 16

Resolution: verified exact; arxiv_id, observed 2026-05-22T08:14:45.569516Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.

Observation 80c7379f-9e03-4434-b55d-bc1c5b5c6b1c · inbound

Massive Spikes in LLMs are Bias Vectors: Mechanistic Uncovering and Spike-Free Quantization cites this paper.

Massive Spikes in LLMs are Bias Vectors: Mechanistic Uncovering and Spike-Free Quantization What are you sinking? A geometric approach on attention sink

Reference 2

Resolution: metadata mismatch; arxiv_id, observed 2026-07-01T22:36:17.027283Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-18T06:34:40.430872+00:00.