Paper Citation Record · LEDGER

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models

As of 17 August 2026, this Paper Citation Record lists 31 of 31 outbound references and 3 inbound Pith citation observations for arXiv:2505.17769.

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

pith.paper-citation-record.v1
2505.17769 v2

Coverage vector

measured 31 of 31 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-08-07T14:44:44.021955Z

measured 34 of 34 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-17T06:30:58.91139+00:00

measured 3 of 3 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links, observed 2026-08-07T12:30:05.981400Z

measured 1 of 1 external citation measurements

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

Source: arxiv_reference, observed 2026-08-05T02:28:24.338817Z

Reference resolution

31 of 31 outbound references displayed

verified exact0
verified fuzzy12
unresolved19
parse uncertain0
malformed identifier0
metadata mismatch0

External citation measurements

0
arxiv_reference, observed 2026-08-05T02:28:24.338817Z

Outbound references

Observation 1eb949d6-74e7-4814-bbc5-c075ff6d2631 · outbound

This paper cites an unresolved cited work.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Unresolved cited work

Reference 1

Resolution: unresolved; raw_fallback, observed 2026-08-07T14:44:44.853884Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 95cf4d0e-bbed-420d-a53d-8889a9194442 · outbound

This paper cites Transcoders find interpretable llm feature circuits.NeurIPS 2024,.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Transcoders find interpretable llm feature circuits.NeurIPS 2024,

Reference 4

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:44:47.548746Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 553005ea-e8d6-45c2-96de-9ef16933b1f6 · outbound

This paper cites The Pile: An 800GB Dataset of Diverse Text for Language Modeling.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models The Pile: An 800GB Dataset of Diverse Text for Language Modeling

Reference 7

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:42.345620Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation f702ad0c-155d-4fb8-a23a-8fbcada5dd0d · outbound

This paper cites Sparse autoencoders can interpret randomly initialized transform- ers.arXiv preprint arXiv:2501.17727,.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Sparse autoencoders can interpret randomly initialized transform- ers.arXiv preprint arXiv:2501.17727,

Reference 8

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:42.427740Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation fccb8e5c-b84c-4302-bd2b-387b3bae6f71 · outbound

This paper cites Saebench: A comprehensive benchmark for sparse autoencoders, December 2024a.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Saebench: A comprehensive benchmark for sparse autoencoders, December 2024a

Reference 11

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:44:46.831434Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation bb2d5175-7072-4bf8-8535-571c78840d20 · outbound

This paper cites Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2

Reference 13

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:42.856429Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ec0bdb8c-202d-4452-97e6-e8969bf8b555 · outbound

This paper cites Automatically Interpreting Millions of Features in Large Language Models.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Automatically Interpreting Millions of Features in Large Language Models

Reference 17

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:43.094342Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3da134b9-22d9-4cf2-a558-34b803115ff3 · outbound

This paper cites Open Problems in Mechanistic Interpretability.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Open Problems in Mechanistic Interpretability

Reference 19

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:43.200839Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 7846871d-511e-456a-9974-81962e89e6e5 · outbound

This paper cites Gemma 2: Improving Open Language Models at a Practical Size.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Gemma 2: Improving Open Language Models at a Practical Size

Reference 20

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:43.266471Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation fe01fa85-282d-4f04-bed8-48b550b34cd2 · outbound

This paper cites This provides a gradient for training unlike the L0-norm, but suppresses latent activations harming reconstruction performance (Rajamanoharan et al., 2025).

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models This provides a gradient for training unlike the L0-norm, but suppresses latent activations harming reconstruction performance (Rajamanoharan et al., 2025)

Reference 21

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:44:46.666826Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 51713ab3-57c2-46f3-a3b1-4110885fe4af · outbound

This paper cites an unresolved cited work.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Unresolved cited work

Reference 22

Resolution: unresolved; raw_fallback, observed 2026-08-07T14:44:46.498614Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 6f0f4803-afca-4f43-a14f-e38d55c8cd70 · outbound

This paper cites an unresolved cited work.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Unresolved cited work

Reference 23

Resolution: unresolved; raw_fallback, observed 2026-08-07T14:44:46.240958Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation d7319a43-23fd-46ef-b04f-c6ea37c880bc · outbound

This paper cites How Idris Elba’s ’Luther’ Puts Us in the Mind set of a Renegade Detective. “Luther.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models How Idris Elba’s ’Luther’ Puts Us in the Mind set of a Renegade Detective. “Luther

Reference 24

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:44:45.984384Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation a68772e7-5935-42b2-ba35-6f68e4c4d106 · outbound

This paper cites an unresolved cited work.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Unresolved cited work

Reference 28

Resolution: unresolved; raw_fallback, observed 2026-08-07T14:44:45.051445Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 7991f56e-153a-4269-85b8-96986cfe076d · outbound

This paper cites Robot-assisted laparoscopic renal artery aneurysm repair with selective arterial clamping. Renal artery aneurysms represent a rare clinical.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Robot-assisted laparoscopic renal artery aneurysm repair with selective arterial clamping. Renal artery aneurysms represent a rare clinical

Reference 30

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:44:45.264819Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation a63b0e38-20d1-4c67-939a-c4b5fe721a7e · outbound

This paper cites Note the higher similarity within model architectures.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Note the higher similarity within model architectures

Reference 31

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:44:44.489730Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation b4cb0a66-9821-4f67-8879-f555e90d1bc7 · outbound

This paper cites Activations of 0 are omitted for legibility.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Activations of 0 are omitted for legibility

Reference 50

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:44:45.716451Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 47f1ae1d-9832-43e7-8e90-68d2d089344c · outbound

This paper cites In-context Learning and Induction Heads.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models In-context Learning and Induction Heads

Reference 1997

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:43.035887Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9a02851f-305b-48d0-81b0-d0c0a4809ffe · outbound

This paper cites Understanding intermediate layers using linear classifier probes.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Understanding intermediate layers using linear classifier probes

Reference 2006

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:41.885452Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 65bad93b-77de-4a8e-b1ae-a764b8847d95 · outbound

This paper cites Are Sparse Autoencoders Useful? A Case Study in Sparse Probing.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Are Sparse Autoencoders Useful? A Case Study in Sparse Probing

Reference 2008

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:42.625025Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 70983da0-1086-49b3-831e-33292b241dce · outbound

This paper cites Quantifying Feature Space Universality Across Large Language Models via Sparse Autoencoders.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Quantifying Feature Space Universality Across Large Language Models via Sparse Autoencoders

Reference 2009

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:42.785494Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b1cb3b1f-d3e1-4ee0-8524-64eaff27958d · outbound

This paper cites NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internals.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internals

Reference 2010

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:42.266153Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 0df4e643-97e8-468a-b609-12cafd42f614 · outbound

This paper cites k-Sparse Autoencoders.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models k-Sparse Autoencoders

Reference 2014

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:42.916690Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 06cee725-277b-45a1-a9c7-2dc6f15635da · outbound

This paper cites Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders

Reference 2017

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:43.148233Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b0e495c8-5b73-49fe-a311-4d8fd5a12a49 · outbound

This paper cites SAEBench metrics were created for evaluating SAEs, which limits their applicability to ITDAs.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models SAEBench metrics were created for evaluating SAEs, which limits their applicability to ITDAs

Reference 2020

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:44:44.655171Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 3a4bcadf-cc16-4c91-bef6-0cd69aef0a59 · outbound

This paper cites Erhan, D., Courville, A., Bengio, Y ., and Vincent, P.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Erhan, D., Courville, A., Bengio, Y ., and Vincent, P

Reference 2021

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:44:47.202702Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 0b6b2d8b-ff9a-4d24-8d73-e38c49e9022e · outbound

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

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Relative representations enable zero-shot latent space communication

Reference 2022

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:42.982828Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation fea16326-328c-405d-9541-4796f430d6eb · outbound

This paper cites The Llama 3 Herd of Models.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models The Llama 3 Herd of Models

Reference 2023

Resolution: unresolved; no resolver link, observed 2026-08-07T14:44:42.057515Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 59f5debc-0196-472d-bb19-6f0a0cfe9eed · outbound

This paper cites Sparse autoencoders find highly interpretable features in language models.ICLR 2024,.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models Sparse autoencoders find highly interpretable features in language models.ICLR 2024,

Reference 2024

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:44:47.800969Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 777e870b-12c5-495c-9393-e10c7463cf43 · outbound

This paper cites and Manning, C.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models and Manning, C

Reference 2025

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:44:47.000622Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 38989650-1d08-4e50-80f6-328da5afd4ed · outbound

This paper cites 1 . Field of the Invention \n The present invention relates to a camera system for transmitting and receiving data to and from a camera by obtaining information /hlon.

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models 1 . Field of the Invention \n The present invention relates to a camera system for transmitting and receiving data to and from a camera by obtaining information /hlon

Reference 7000

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:44:45.502512Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Pith citing papers

Observation 9a01235e-b8c4-41b5-b0e7-4b43a850fa71 · inbound

Interpreting Large Text-to-Image Diffusion Models with Dictionary Learning cites this paper.

Interpreting Large Text-to-Image Diffusion Models with Dictionary Learning Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models

Reference 31

Resolution: unresolved; no resolver link, observed 2026-08-07T12:30:05.981400Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 07c2b787-83be-431b-982d-1ba09205994c · inbound

Subspace-Aware Sparse Autoencoders for Effective Mechanistic Interpretability cites this paper.

Subspace-Aware Sparse Autoencoders for Effective Mechanistic Interpretability Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models

Reference 31

Resolution: verified exact; arxiv_id, observed 2026-06-28T02:11:29.076262Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

Observation 1bc492b3-4fc2-4105-8c10-ad0773b3201f · inbound

ICA Lens: Interpreting Language Models Without Training Another Dictionary cites this paper.

ICA Lens: Interpreting Language Models Without Training Another Dictionary Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models

Reference 12

Resolution: verified exact; arxiv_id, observed 2026-07-03T09:37:49.254240Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.