Paper Citation Record · LEDGER

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization

As of 11 August 2026, this Paper Citation Record lists 38 of 38 outbound references and 1 inbound Pith citation observation for arXiv:2506.22396.

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

pith.paper-citation-record.v1
2506.22396 v1

Coverage vector

measured 38 of 38 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-08-06T22:10:23.901164Z

measured 39 of 39 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-11T06:34:44.6726+00:00

measured 1 of 1 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links, observed 2026-05-14T22:59:07.065193Z

measured 0 of 1 external citation measurements

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

Source: arxiv_reference, observed 2026-05-14T22:59:34.279645Z

Reference resolution

38 of 38 outbound references displayed

verified exact0
verified fuzzy19
unresolved19
parse uncertain0
malformed identifier0
metadata mismatch0

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation cbdea40f-a78e-4bf2-a233-45def0e3706a · outbound

This paper cites This ": processed all layers.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization This ": processed all layers

Reference 1

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:29.345819Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 55d60984-3f8a-4332-ad5b-83b8aa8f4132 · outbound

This paper cites this ": KV diff 1.00 < 0.30 -> Write.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization this ": KV diff 1.00 < 0.30 -> Write

Reference 2

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:29.210914Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation ceffd707-cb7d-43bb-a4ab-7af4522662a4 · outbound

This paper cites This " +.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization This " +

Reference 3

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:29.042472Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation f923ebb3-9e92-4cd0-9245-b8626840ec47 · outbound

This paper cites Sparks of Artificial General Intelligence: Early experiments with GPT-4.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Sparks of Artificial General Intelligence: Early experiments with GPT-4

Reference 4

Resolution: unresolved; no resolver link, observed 2026-08-06T22:10:20.844437Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 7fc147b7-69e6-4816-baf2-79182af41f32 · outbound

This paper cites GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers

Reference 5

Resolution: unresolved; no resolver link, observed 2026-08-06T22:10:20.927295Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d563ab34-0ddb-4def-8e66-9e57cdad4304 · outbound

This paper cites Adaptive Computation Time for Recurrent Neural Networks.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Adaptive Computation Time for Recurrent Neural Networks

Reference 6

Resolution: unresolved; no resolver link, observed 2026-08-06T22:10:20.989489Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 10675910-233b-4717-b9d8-d1a0a413bb93 · outbound

This paper cites Estimating the Carbon Footprint of BLOOM, a 176B Parameter Language Model.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Estimating the Carbon Footprint of BLOOM, a 176B Parameter Language Model

Reference 8

Resolution: unresolved; no resolver link, observed 2026-08-06T22:10:21.184745Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6e1305a7-c7ab-4ee0-80cc-4a5aeb0c3cb4 · outbound

This paper cites OpenAI Blog.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization OpenAI Blog

Reference 9

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:29.706660Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 9830d11b-602f-42ed-966e-906875262a7b · outbound

This paper cites Transactions of the Association for Computational Linguistics, 8:842–866.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Transactions of the Association for Computational Linguistics, 8:842–866

Reference 10

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:29.525120Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 3005d7e8-64b6-4768-9db1-f58c93b9e6b5 · outbound

This paper cites and ": entropy 0.23 -> 2 - bit quant Token.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization and ": entropy 0.23 -> 2 - bit quant Token

Reference 15

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:28.888395Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 5de69b14-c194-45c1-8a33-36faf45a01f1 · outbound

This paper cites an unresolved cited work.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Unresolved cited work

Reference 16

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:10:28.721423Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 70b860aa-9582-4900-a298-c52495bfbb84 · outbound

This paper cites an unresolved cited work.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Unresolved cited work

Reference 17

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:10:28.464567Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation be1bc39a-c227-454f-9c9b-023fc1d7c74b · outbound

This paper cites the, ” “of,.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization the, ” “of,

Reference 18

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:28.130054Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 340a6210-5183-4572-bc9e-972aa58128c8 · outbound

This paper cites an unresolved cited work.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Unresolved cited work

Reference 19

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:10:27.797728Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 8e967772-e3fd-432c-971b-ea39e29bc34a · outbound

This paper cites an unresolved cited work.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Unresolved cited work

Reference 20

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:10:27.481988Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation b44abbd4-911d-4b2d-8854-64d7dd03a1fb · outbound

This paper cites skipped after Layer 2.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization skipped after Layer 2

Reference 21

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:27.259969Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation ac9c5c76-258e-483a-9422-72051c13e320 · outbound

This paper cites an unresolved cited work.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Unresolved cited work

Reference 22

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:10:26.990889Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 50b1b291-d620-430c-afe5-a794443dc6fb · outbound

This paper cites an unresolved cited work.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Unresolved cited work

Reference 23

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:10:26.661016Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation ff59b215-10fa-4728-a5df-f16c00033a54 · outbound

This paper cites an unresolved cited work.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Unresolved cited work

Reference 24

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:10:26.307801Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 320abb27-1600-4517-9b8f-8980131a5ebe · outbound

This paper cites This conservative threshold ensures fusion only when representational collapse is semantically safe.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization This conservative threshold ensures fusion only when representational collapse is semantically safe

Reference 25

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:26.002069Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 13489094-6c33-4013-9521-78b40884a72b · outbound

This paper cites an unresolved cited work.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Unresolved cited work

Reference 26

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:10:25.699415Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation a3a64abf-0ac8-46a2-bd14-9a857917d78e · outbound

This paper cites We choose the pair (τlow, τhigh) = (0.3, 0.6) that achieves a strong Pareto frontier: ∼ 8.6% additional FLOP reduction with less than 0.1 perplexity change.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization We choose the pair (τlow, τhigh) = (0.3, 0.6) that achieves a strong Pareto frontier: ∼ 8.6% additional FLOP reduction with less than 0.1 perplexity change

Reference 27

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:25.606207Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation dff6614e-5875-45ae-b55d-43a4879dbd19 · outbound

This paper cites an unresolved cited work.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Unresolved cited work

Reference 28

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:10:25.456484Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 9b1963e1-8a94-4e7c-b5d8-77fe94faa165 · outbound

This paper cites an unresolved cited work.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Unresolved cited work

Reference 29

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:10:25.334389Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 5780c87f-a0be-4ef7-aac8-637ceecbbc95 · outbound

This paper cites Pareto Surface Analysis.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Pareto Surface Analysis

Reference 30

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:25.204098Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 2632d43f-6fdb-40dc-a5a3-e86ccce7e131 · outbound

This paper cites an unresolved cited work.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Unresolved cited work

Reference 31

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:10:25.073338Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 5a48624b-df50-437f-a210-2fa41d532593 · outbound

This paper cites an unresolved cited work.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Unresolved cited work

Reference 32

Resolution: unresolved; raw_fallback, observed 2026-08-06T22:10:24.929281Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 109bdff8-33a9-4348-a66a-093ce5cc00fb · outbound

This paper cites We hope this work encourages the community to adopt tools like CodeCarbon not as post-hoc profil- ers, but as first-class citizens in the deployment pipeline.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization We hope this work encourages the community to adopt tools like CodeCarbon not as post-hoc profil- ers, but as first-class citizens in the deployment pipeline

Reference 33

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:24.813090Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 21bf7a74-74a1-40a4-88b5-0dc614ac3f85 · outbound

This paper cites I.2 holds, halt the token unconditionally.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization I.2 holds, halt the token unconditionally

Reference 34

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:24.715856Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 7f17e496-1fa0-4250-a091-099349017816 · outbound

This paper cites I.3 holds for any u, merge (t, u).

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization I.3 holds for any u, merge (t, u)

Reference 35

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:24.575835Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 6a014c21-1968-4921-a1bf-73f7835c4458 · outbound

This paper cites This priority is grounded in halting the provision of computational savings without representational loss, while fusion entails approximation.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization This priority is grounded in halting the provision of computational savings without representational loss, while fusion entails approximation

Reference 36

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:24.483090Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation a247a1bb-80cf-4184-a20b-6a4175ef868f · outbound

This paper cites the”, “of.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization the”, “of

Reference 37

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:24.333934Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 53f694ef-434c-44fb-8cba-210f44d7265b · outbound

This paper cites tokens.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization tokens

Reference 103

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:24.153732Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 049f2228-6a3f-4ccf-86d7-8905790b5e24 · outbound

This paper cites Shaojie Bai, J Zico Kolter, and Vladlen Koltun.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Shaojie Bai, J Zico Kolter, and Vladlen Koltun

Reference 2018

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:29.876336Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 76f29783-9953-4308-a35a-b0f1be593238 · outbound

This paper cites Fast Inference from Transformers via Speculative Decoding.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Fast Inference from Transformers via Speculative Decoding

Reference 2019

Resolution: unresolved; no resolver link, observed 2026-08-06T22:10:21.087718Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation f82ae740-22a5-4662-835f-af65acadd9b6 · outbound

This paper cites In International Conference on Machine Learning (ICML).

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization In International Conference on Machine Learning (ICML)

Reference 2020

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T22:10:30.034930Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation aa67e5e7-8c7a-4cbc-8cd2-f5f616578b5c · outbound

This paper cites Post-training 4-bit quantization of convolution networks for rapid-deployment.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Post-training 4-bit quantization of convolution networks for rapid-deployment

Reference 2021

Resolution: unresolved; no resolver link, observed 2026-08-06T22:10:20.694641Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9daf933b-ddf0-44b3-9907-a617827aab08 · outbound

This paper cites Toolformer: Language Models Can Teach Themselves to Use Tools.

QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization Toolformer: Language Models Can Teach Themselves to Use Tools

Reference 2022

Resolution: unresolved; no resolver link, observed 2026-08-06T22:10:21.546065Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Pith citing papers

Observation c9037c0f-f0b6-4073-b2a2-f85c2c71e183 · inbound

Two-dimensional early exit optimisation of LLM inference cites this paper.

Two-dimensional early exit optimisation of LLM inference QuickSilver -- Speeding up LLM Inference through Dynamic Token Halting, KV Skipping, Contextual Token Fusion, and Adaptive Matryoshka Quantization

Reference 11

Resolution: verified exact; arxiv_id, observed 2026-05-14T22:59:34.284901Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.