Paper Citation Record · LEDGER

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage

As of 10 August 2026, this Paper Citation Record lists 39 of 39 outbound references and 0 inbound Pith citation observations for arXiv:2506.06472.

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

pith.paper-citation-record.v1
2506.06472 v1

Coverage vector

measured 39 of 39 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-08-07T06:00:33.427109Z

measured 39 of 39 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-10T06:31:04.303077+00:00

measured 0 of 0 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links

measured 0 of 1 external citation measurements

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

Source: cited_works

Reference resolution

39 of 39 outbound references displayed

verified exact0
verified fuzzy22
unresolved17
parse uncertain0
malformed identifier0
metadata mismatch0

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation f2e46f6e-6bdf-45b9-938b-f02312d76d6c · outbound

This paper cites In19th USENIX Conference on File and Storage Technologies (FAST 21), pages 387–401, 2021.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage In19th USENIX Conference on File and Storage Technologies (FAST 21), pages 387–401, 2021

Reference 1

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:38.224953Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 8b9ac2a1-8192-4455-a34a-31ae374be6b4 · outbound

This paper cites Efficient combination of rematerialization and offloading for training dnns.Advances in Neural Information Processing Systems, 34:23844–23857, 2021.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Efficient combination of rematerialization and offloading for training dnns.Advances in Neural Information Processing Systems, 34:23844–23857, 2021

Reference 2

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:38.076966Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 75a2931f-a17e-40b1-a1bb-47b4253cdbf5 · outbound

This paper cites BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

Reference 3

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.219076Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4010036c-c454-4239-ae2b-b2339ce45d89 · outbound

This paper cites The Llama 3 Herd of Models.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage The Llama 3 Herd of Models

Reference 4

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.223943Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 157e3f5e-29db-458b-8529-bd6344290d2c · outbound

This paper cites Exxact.https://www.exxactcorp.com/, 2025.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Exxact.https://www.exxactcorp.com/, 2025

Reference 5

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:37.885091Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 79269ad2-b570-4a8c-a116-40c88095f66c · outbound

This paper cites T5 11b.https://huggingface.co/google-t5/t5-11b, 2025.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage T5 11b.https://huggingface.co/google-t5/t5-11b, 2025

Reference 6

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:37.668737Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 0220b629-eb06-4ca6-af42-eb96375c0285 · outbound

This paper cites nvidia.com/blog/gpudirect-storage/.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage nvidia.com/blog/gpudirect-storage/

Reference 7

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:37.498608Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation fa229a1c-7736-48aa-901c-63497343213c · outbound

This paper cites Swapadvisor: Pushingdeeplearningbeyondthegpu memory limit via smart swapping.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Swapadvisor: Pushingdeeplearningbeyondthegpu memory limit via smart swapping

Reference 8

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:37.315978Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 565f4070-ef28-41fb-a56a-b5144f07d578 · outbound

This paper cites Gpipe: Efficient training of giant neural networks using pipeline parallelism.Advances in neural information processing systems, 32, 2019.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Gpipe: Efficient training of giant neural networks using pipeline parallelism.Advances in neural information processing systems, 32, 2019

Reference 9

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.255025Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 90eeb0ea-75ae-43c7-9042-18b66e18393f · outbound

This paper cites Ibm granite.https://huggingface.co/ibm-granite, 2025.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Ibm granite.https://huggingface.co/ibm-granite, 2025

Reference 10

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:37.132856Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 05d469d2-ce6e-4e5c-8cd2-08bfb9f65c09 · outbound

This paper cites Checkmate: Breaking the memory wall with optimal tensor rematerialization.Proceedings of Machine Learning and Systems, 2:497–511, 2020.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Checkmate: Breaking the memory wall with optimal tensor rematerialization.Proceedings of Machine Learning and Systems, 2:497–511, 2020

Reference 11

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:36.954625Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 66b4a927-a40b-4dd6-84cc-2fc2347ddce8 · outbound

This paper cites Smart-infinity: Fastlargelanguagemodeltrainingusingnear-storageprocessingonarealsystem.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Smart-infinity: Fastlargelanguagemodeltrainingusingnear-storageprocessingonarealsystem

Reference 12

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:36.734315Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 62e93b52-97ee-480b-9a82-f474a889d2ba · outbound

This paper cites Deepum: Tensormigrationandprefetchinginunified memory.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Deepum: Tensormigrationandprefetchinginunified memory

Reference 13

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:36.517578Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 9e4d01fd-f963-4c16-aac4-afdd38782399 · outbound

This paper cites Scaling Laws for Neural Language Models.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Scaling Laws for Neural Language Models

Reference 14

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.301738Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1ee83164-88a9-44c8-8c30-d5f8387bc992 · outbound

This paper cites Beyond the memory wall: A case for memory-centric hpc system for deep learning.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Beyond the memory wall: A case for memory-centric hpc system for deep learning

Reference 15

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:36.317589Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 25d743ac-8fb4-4732-845d-9763d4d96a9a · outbound

This paper cites TFLMS: Large Model Support in TensorFlow by Graph Rewriting.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage TFLMS: Large Model Support in TensorFlow by Graph Rewriting

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.323175Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6c2374d8-4fe3-4c22-b633-27febef3d874 · outbound

This paper cites TorchTitan: One-stop PyTorch native solution for production ready LLM pre-training.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage TorchTitan: One-stop PyTorch native solution for production ready LLM pre-training

Reference 17

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.332267Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9347a12c-dc38-4126-87ec-492546f5dec0 · outbound

This paper cites LoHan: Low-Cost High-Performance Framework to Fine-Tune 100B Model on a Consumer GPU.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage LoHan: Low-Cost High-Performance Framework to Fine-Tune 100B Model on a Consumer GPU

Reference 18

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.342788Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 93c3f85c-3e80-4487-aace-506d2cf81670 · outbound

This paper cites An Empirical Model of Large-Batch Training.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage An Empirical Model of Large-Batch Training

Reference 19

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.354308Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6b5ea083-c2da-4c5a-a012-ecf0710c1997 · outbound

This paper cites Mixed Precision Training.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Mixed Precision Training

Reference 20

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.365297Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c7725067-0716-481a-ba8e-9b15d1ccdbdd · outbound

This paper cites Pipedream: Generalized pipeline parallelism for dnn training.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Pipedream: Generalized pipeline parallelism for dnn training

Reference 21

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.377846Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4244f784-58d3-4f9b-b7ce-5b136bfbc231 · outbound

This paper cites Angel-PTM: A Scalable and Economical Large-scale Pre-training System in Tencent.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Angel-PTM: A Scalable and Economical Large-scale Pre-training System in Tencent

Reference 22

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.388782Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation e21e5ca9-0316-495e-81fd-55617bd9174e · outbound

This paper cites Automatic differentiation in pytorch.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Automatic differentiation in pytorch

Reference 23

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.401389Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 0f628802-5c38-4ba1-8595-23609fd97d9c · outbound

This paper cites Scalable diffusion models with transformers.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Scalable diffusion models with transformers

Reference 24

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.412497Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 14a0829f-bdb2-48d5-916a-ca38b86e1ebd · outbound

This paper cites Training Large Neural Networks with Constant Memory using a New Execution Algorithm.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Training Large Neural Networks with Constant Memory using a New Execution Algorithm

Reference 25

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.422749Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5047daf3-6af8-4ed3-8baa-742822822aaf · outbound

This paper cites Gpu-initiated on-demand high-throughput storage access in the bam system architecture.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Gpu-initiated on-demand high-throughput storage access in the bam system architecture

Reference 26

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:36.064077Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 635b8491-36e0-499a-ae3f-3c6646cb1c40 · outbound

This paper cites Language models are unsupervised multitask learners.OpenAI blog, 1(8):9, 2019.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Language models are unsupervised multitask learners.OpenAI blog, 1(8):9, 2019

Reference 27

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.441563Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 01ad0130-5142-4528-bb29-3d84f7a75259 · outbound

This paper cites Exploring the limits of transfer learning with a unified text-to-text transformer.Journal of machine learning research, 21(140):1–67, 2020.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Exploring the limits of transfer learning with a unified text-to-text transformer.Journal of machine learning research, 21(140):1–67, 2020

Reference 28

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:32.485821Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ea7d54cc-26f5-4bec-8c70-07fc2c4a603c · outbound

This paper cites Zero- infinity: Breaking the gpu memory wall for extreme scale deep learning.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Zero- infinity: Breaking the gpu memory wall for extreme scale deep learning

Reference 29

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:35.776160Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 60312ea8-13a6-49bd-acc7-eda91f132159 · outbound

This paper cites {Zero-offload}: Democratizing{billion-scale}model training.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage {Zero-offload}: Democratizing{billion-scale}model training

Reference 30

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:35.580069Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 612eff07-8439-4277-b73d-1afec967f1e8 · outbound

This paper cites vdnn: Virtualizeddeepneuralnetworksforscalable,memory-efficientneuralnetworkdesign.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage vdnn: Virtualizeddeepneuralnetworksforscalable,memory-efficientneuralnetworkdesign

Reference 31

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:35.359528Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation fe8d81db-435b-4143-9666-f0460fa68928 · outbound

This paper cites Building ai agents for autonomous clouds: Challenges and design principles.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Building ai agents for autonomous clouds: Challenges and design principles

Reference 32

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:35.090008Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 4899a902-1259-4f04-a872-4a13a2e6d2b3 · outbound

This paper cites Stronghold: fast and affordable billion-scale deep learning model training.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Stronghold: fast and affordable billion-scale deep learning model training

Reference 33

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:34.872110Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 87b62046-85c8-4127-ae0d-7ee6f187fbb5 · outbound

This paper cites Superneurons: Dynamic gpu memory management for training deep neural networks.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Superneurons: Dynamic gpu memory management for training deep neural networks

Reference 34

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:34.654946Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 60397c44-59f8-443c-abe6-5f9a2c279258 · outbound

This paper cites Netllm: Adapting large language models for networking.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Netllm: Adapting large language models for networking

Reference 35

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:34.439392Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation 108aa304-7f91-4023-931c-365f0d00c4d5 · outbound

This paper cites SSDTrain: An Activation Offloading Framework to SSDs for Faster Large Language Model Training.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage SSDTrain: An Activation Offloading Framework to SSDs for Faster Large Language Model Training

Reference 36

Resolution: unresolved; no resolver link, observed 2026-08-07T06:00:33.164055Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 57779a89-1c9f-4080-9835-0f54f9743c4d · outbound

This paper cites Acceleratingthetrainingoflargelanguagemodelsusingefficientactivation rematerializationandoptimalhybridparallelism.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Acceleratingthetrainingoflargelanguagemodelsusingefficientactivation rematerializationandoptimalhybridparallelism

Reference 37

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:34.182335Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation ca18bf44-aead-4cb4-bb78-9fb324554960 · outbound

This paper cites Zng: Architecting gpu multi-processors with new flash for scalable data analysis.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Zng: Architecting gpu multi-processors with new flash for scalable data analysis

Reference 38

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:34.008635Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Observation f5f669cb-a2c0-4bdd-b073-8e9a4d7e7030 · outbound

This paper cites Flashgpu: Placing new flash next to gpu cores.

Cost-Efficient LLM Training with Lifetime-Aware Tensor Offloading via GPUDirect Storage Flashgpu: Placing new flash next to gpu cores

Reference 39

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T06:00:33.797888Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-10T06:31:04.303077+00:00.

Pith citing papers

No inbound Pith citation observations are available.