Paper Citation Record · LEDGER

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs

As of 21 August 2026, this Paper Citation Record lists 56 of 56 outbound references and 2 inbound Pith citation observations for arXiv:2505.20254.

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

pith.paper-citation-record.v1
2505.20254 v1

Coverage vector

measured 56 of 56 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-08-07T14:03:03.819990Z

measured 58 of 58 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-20T06:33:59.587034+00:00

measured 2 of 2 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links, observed 2026-07-04T17:01:39.362411Z

measured 0 of 1 external citation measurements

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

Source: arxiv_reference, observed 2026-07-04T17:09:58.298491Z

Reference resolution

56 of 56 outbound references displayed

verified exact0
verified fuzzy19
unresolved37
parse uncertain0
malformed identifier0
metadata mismatch0

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation d985f89e-c3d7-481f-bcb8-a67ba9aeba66 · outbound

This paper cites SAFE: A Sparse Autoencoder-Based Framework for Robust Query Enrichment and Hallucination Mitigation in LLMs.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs SAFE: A Sparse Autoencoder-Based Framework for Robust Query Enrichment and Hallucination Mitigation in LLMs

Reference 1

Resolution: unresolved; no resolver link, observed 2026-08-07T14:02:58.487711Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 84fc0e70-98a5-443c-a3cd-cbfa08c00f8f · outbound

This paper cites an unresolved cited work.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Unresolved cited work

Reference 2

Resolution: unresolved; raw_fallback, observed 2026-08-07T14:03:08.646727Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation 645c8277-9cca-4b46-8479-21efa3b5726c · outbound

This paper cites New algorithms for learning incoherent and overcomplete dictionaries.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs New algorithms for learning incoherent and overcomplete dictionaries

Reference 3

Resolution: unresolved; no resolver link, observed 2026-08-07T14:02:58.726604Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 34c35eff-2e77-4f2d-a7b9-d27f17667d93 · outbound

This paper cites Pythia: A suite for analyzing large language models across training and scaling.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Pythia: A suite for analyzing large language models across training and scaling

Reference 4

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:08.494837Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation d388b3e2-e670-49b7-8ca1-79094ac30591 · outbound

This paper cites Turner, Cem Anil, Carson Denison, Amanda Askell, et al.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Turner, Cem Anil, Carson Denison, Amanda Askell, et al

Reference 5

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:08.359800Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation e0457bd2-f38e-4f23-a0f5-9e841a08dcc5 · outbound

This paper cites Towards monosemanticity: Decomposing language models with dictionary learning.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Towards monosemanticity: Decomposing language models with dictionary learning

Reference 6

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:08.219243Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation 6a38c3d4-af46-42a5-87b0-f690c11b3c21 · outbound

This paper cites BatchTopK Sparse Autoencoders.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs BatchTopK Sparse Autoencoders

Reference 7

Resolution: unresolved; no resolver link, observed 2026-08-07T14:02:58.976794Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4f48c707-9311-4ae1-a416-49bbd095833e · outbound

This paper cites Learning Multi-Level Features with Matryoshka Sparse Autoencoders.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Learning Multi-Level Features with Matryoshka Sparse Autoencoders

Reference 8

Resolution: unresolved; no resolver link, observed 2026-08-07T14:02:59.051713Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d04aa17b-e3fd-4de2-a6cf-85e93322aba3 · outbound

This paper cites Improving Steering Vectors by Targeting Sparse Autoencoder Features.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Improving Steering Vectors by Targeting Sparse Autoencoder Features

Reference 9

Resolution: unresolved; no resolver link, observed 2026-08-07T14:02:59.154435Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1363a7c9-9a08-4aba-8d9a-1e8ee822cd3a · outbound

This paper cites A is for absorption: Studying feature splitting and absorption in sparse autoencoders.arXiv preprint arXiv:2409.14507, 2024.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs A is for absorption: Studying feature splitting and absorption in sparse autoencoders.arXiv preprint arXiv:2409.14507, 2024

Reference 10

Resolution: unresolved; no resolver link, observed 2026-08-07T14:02:59.217821Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 052532a1-9c86-440b-96ad-92209798469f · outbound

This paper cites Sparse Autoencoders Find Highly Interpretable Features in Language Models.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Sparse Autoencoders Find Highly Interpretable Features in Language Models

Reference 11

Resolution: unresolved; no resolver link, observed 2026-08-07T14:02:59.277009Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation eb97890a-4497-4757-91c6-f1b8787df341 · outbound

This paper cites Optimally sparse representation in general (nonorthogonal) dictionaries via l1 minimization.Proceedings of the National Academy of Sciences, 100(5):2197– 2202, 2003.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Optimally sparse representation in general (nonorthogonal) dictionaries via l1 minimization.Proceedings of the National Academy of Sciences, 100(5):2197– 2202, 2003

Reference 12

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:08.092039Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation ea1036f8-040f-4298-a3da-7ae2ee0a1cdf · outbound

This paper cites Toy Models of Superposition.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Toy Models of Superposition

Reference 13

Resolution: unresolved; no resolver link, observed 2026-08-07T14:02:59.477693Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6e73a26d-6fa2-4fbd-b629-66a9067e20c9 · outbound

This paper cites A mathematical framework for transformer circuits.Transformer Circuits Thread, 1(1):12, 2021.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs A mathematical framework for transformer circuits.Transformer Circuits Thread, 1(1):12, 2021

Reference 14

Resolution: unresolved; no resolver link, observed 2026-08-07T14:02:59.553114Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation db898c7e-37ed-42a2-aed4-64927b95be1a · outbound

This paper cites Archetypal SAE: Adaptive and Stable Dictionary Learning for Concept Extraction in Large Vision Models.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Archetypal SAE: Adaptive and Stable Dictionary Learning for Concept Extraction in Large Vision Models

Reference 15

Resolution: unresolved; no resolver link, observed 2026-08-07T14:02:59.630371Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 17262248-3e93-4e91-99ee-b4c16cfb830b · outbound

This paper cites Scientific inference with interpretable machine learning: Analyzing models to learn about real-world phenomena.Minds and Machines, 34(3):32, 2024.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Scientific inference with interpretable machine learning: Analyzing models to learn about real-world phenomena.Minds and Machines, 34(3):32, 2024

Reference 16

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:07.963257Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation 9a340735-39cf-4e46-b67a-61471657af26 · outbound

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

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs The Pile: An 800GB Dataset of Diverse Text for Language Modeling

Reference 17

Resolution: unresolved; no resolver link, observed 2026-08-07T14:02:59.781474Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 44650b81-0291-42d5-9ccd-10c19e3d95ab · outbound

This paper cites Scaling and evaluating sparse autoencoders.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Scaling and evaluating sparse autoencoders

Reference 18

Resolution: unresolved; no resolver link, observed 2026-08-07T14:02:59.886828Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 16e0a001-c085-469c-9db9-6eaa35db7c86 · outbound

This paper cites Scaling and evaluating sparse autoencoders.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Scaling and evaluating sparse autoencoders

Reference 19

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:07.848712Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation ed653d77-400e-476f-b68f-b3871cd5a005 · outbound

This paper cites Causal Abstraction: A Theoretical Foundation for Mechanistic Interpretability.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Causal Abstraction: A Theoretical Foundation for Mechanistic Interpretability

Reference 20

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:00.147887Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 711f7380-ba74-4da7-866d-6b0f7556bf10 · outbound

This paper cites SCAR: Sparse Conditioned Autoencoders for Concept Detection and Steering in LLMs.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs SCAR: Sparse Conditioned Autoencoders for Concept Detection and Steering in LLMs

Reference 21

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:00.224128Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3dc34d7f-6197-46c2-b0e1-71ca5554b36e · outbound

This paper cites When can dictionary learning uniquely recover sparse data from subsamples?IEEE Transactions on Information Theory, 61(11):6290–6297, 2015.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs When can dictionary learning uniquely recover sparse data from subsamples?IEEE Transactions on Information Theory, 61(11):6290–6297, 2015

Reference 22

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:07.725224Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation 1e0a86e9-fb72-408a-aca9-8bd88cdc2b1d · outbound

This paper cites Projecting assumptions: The duality between sparse autoencoders and concept geometry.arXiv preprint arXiv:2503.01822, 2025.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Projecting assumptions: The duality between sparse autoencoders and concept geometry.arXiv preprint arXiv:2503.01822, 2025

Reference 23

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:00.464472Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 30ea7564-3ac3-4918-8c04-284def90e3d4 · outbound

This paper cites Independent component analysis: algorithms and applications.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Independent component analysis: algorithms and applications

Reference 24

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:00.545532Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 40145398-b7ba-4508-9d24-4cac2ad7d40a · outbound

This paper cites Identifiable steering via sparse autoencoding of multi-concept shifts.arXiv preprint arXiv:2502.12179, 2025.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Identifiable steering via sparse autoencoding of multi-concept shifts.arXiv preprint arXiv:2502.12179, 2025

Reference 25

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:00.599280Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c34b9b99-f69f-4e55-acd7-16470e758eac · outbound

This paper cites SAEBench: A Comprehensive Benchmark for Sparse Autoencoders.https://www.neuronpedia.org/sae-bench/info, 2024.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs SAEBench: A Comprehensive Benchmark for Sparse Autoencoders.https://www.neuronpedia.org/sae-bench/info, 2024

Reference 26

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:07.558514Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation d5a3890e-9477-4eb5-adb4-83caf2541240 · outbound

This paper cites Measuring progress in dictionary learning for language model interpretability with board game models.Advances in Neural Information Processing Systems, 37:83091–83118, 2024.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Measuring progress in dictionary learning for language model interpretability with board game models.Advances in Neural Information Processing Systems, 37:83091–83118, 2024

Reference 27

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:07.323869Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation 9472c785-10ea-4cc4-bee1-c5ab6b941e6d · outbound

This paper cites Sparse Autoencoders Do Not Find Canonical Units of Analysis.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Sparse Autoencoders Do Not Find Canonical Units of Analysis

Reference 28

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:00.843301Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5afc856f-5def-4f83-a651-10d75b848acb · outbound

This paper cites an unresolved cited work.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Unresolved cited work

Reference 29

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:00.948695Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a6d3d15b-1428-4035-a894-0823aaa09cf9 · outbound

This paper cites The mythos of model interpretability: In machine learning, the concept of interpretability is both important and slippery.Queue, 16(3):31–57, 2018.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs The mythos of model interpretability: In machine learning, the concept of interpretability is both important and slippery.Queue, 16(3):31–57, 2018

Reference 30

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:01.063367Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9e74e2b7-c5fd-44be-881c-5fd1e2935ffc · outbound

This paper cites Is This the Subspace You Are Looking for? An Interpretability Illusion for Subspace Activation Patching.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Is This the Subspace You Are Looking for? An Interpretability Illusion for Subspace Activation Patching

Reference 31

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:01.213348Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4ccd8b59-f9fb-46b4-bddd-424074957b94 · outbound

This paper cites Enhancing Neural Network Interpretability with Feature-Aligned Sparse Autoencoders.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Enhancing Neural Network Interpretability with Feature-Aligned Sparse Autoencoders

Reference 32

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:01.307047Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c2c6a64f-b920-4ca4-ab4c-4408c1615757 · outbound

This paper cites Dictionary learning.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Dictionary learning

Reference 33

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:07.152154Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation 2aa0ffdd-ad11-4ee0-a36f-eb6a806197e4 · outbound

This paper cites Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models

Reference 34

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:01.502160Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 70443d27-4dcd-47a8-af9a-664c3dc8acec · outbound

This paper cites Sparse feature circuits: Discovering and editing interpretable causal graphs in language models.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Sparse feature circuits: Discovering and editing interpretable causal graphs in language models

Reference 35

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:06.970419Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation 611694cd-002a-4a4c-9c56-819eca673eed · outbound

This paper cites Everything, Everywhere, All at Once: Is Mechanistic Interpretability Identifiable?.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Everything, Everywhere, All at Once: Is Mechanistic Interpretability Identifiable?

Reference 36

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:01.725857Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5aa98ec1-5520-44bb-8dfb-b821c908129a · outbound

This paper cites SAEs $\textit{Can}$ Improve Unlearning: Dynamic Sparse Autoencoder Guardrails for Precision Unlearning in LLMs.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs SAEs $\textit{Can}$ Improve Unlearning: Dynamic Sparse Autoencoder Guardrails for Precision Unlearning in LLMs

Reference 37

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:01.830004Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 8971053d-2e39-4d5b-9596-5976a58bb55b · outbound

This paper cites Steering Language Model Refusal with Sparse Autoencoders.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Steering Language Model Refusal with Sparse Autoencoders

Reference 38

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:01.923666Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c0744883-b547-4b5c-b50f-e89ef3fdaee1 · outbound

This paper cites Mechanistic interpretability, variables, and the importance of interpretable bases.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Mechanistic interpretability, variables, and the importance of interpretable bases

Reference 39

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:06.824381Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation f6606495-3294-41fe-b978-60cefaafdf67 · outbound

This paper cites Zoom in: An introduction to circuits.Distill, 5(3):e00024–001, 2020.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Zoom in: An introduction to circuits.Distill, 5(3):e00024–001, 2020

Reference 40

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:02.169916Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3db0a845-75cf-492c-8210-cf42f735372d · outbound

This paper cites The building blocks of interpretability.Distill, 3(3):e10, 2018.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs The building blocks of interpretability.Distill, 3(3):e10, 2018

Reference 41

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:06.551552Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation 80efe505-6c96-40c4-806b-a586fdc29537 · outbound

This paper cites Compute Optimal Inference and Provable Amortisation Gap in Sparse Autoencoders.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Compute Optimal Inference and Provable Amortisation Gap in Sparse Autoencoders

Reference 42

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:02.401440Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 962b0a1f-79ea-4677-a29a-dddb71067f29 · outbound

This paper cites Sparse autoencoders learn monosemantic features in vision-language models.arXiv preprint arXiv:2504.02821, 2025.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Sparse autoencoders learn monosemantic features in vision-language models.arXiv preprint arXiv:2504.02821, 2025

Reference 43

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:02.554813Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 0174c757-c952-4a4b-86e9-ca37bda1d101 · outbound

This paper cites Sparse Autoencoders Trained on the Same Data Learn Different Features.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Sparse Autoencoders Trained on the Same Data Learn Different Features

Reference 44

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:02.656998Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 69d5d226-67de-45f7-aa85-cd7999a7ea30 · outbound

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

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Automatically Interpreting Millions of Features in Large Language Models

Reference 45

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:02.798784Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9157edce-d3e0-47ee-afc8-fd9ec6246fb9 · outbound

This paper cites Improving Dictionary Learning with Gated Sparse Autoencoders.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Improving Dictionary Learning with Gated Sparse Autoencoders

Reference 46

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:02.963645Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 49a5bc79-cd46-4749-b6f5-eef6fb7db0f6 · outbound

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

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders

Reference 47

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:03.074303Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c62a6bcc-f559-49e7-8447-343d0259e375 · outbound

This paper cites Global identifiability of overcomplete dictionary learning via l1 and volume minimization.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Global identifiability of overcomplete dictionary learning via l1 and volume minimization

Reference 48

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:06.409527Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation c4230207-d7dc-463a-b104-9ce6c21900af · outbound

This paper cites Gemma: Open Models Based on Gemini Research and Technology.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Gemma: Open Models Based on Gemini Research and Technology

Reference 49

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:03.259155Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 34708252-3bf4-45cc-8e99-1421f135b8af · outbound

This paper cites Identifiability of overcomplete independent component analysis.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Identifiability of overcomplete independent component analysis

Reference 50

Resolution: unresolved; no resolver link, observed 2026-08-07T14:03:03.323826Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6f2f630b-77d6-4bae-8949-39cd1ddfbc23 · outbound

This paper cites Only the most frequent clusters show consistent reproducibility, indicating severe capacity limitations where the dictionary should prioritize only the dominant clusters.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Only the most frequent clusters show consistent reproducibility, indicating severe capacity limitations where the dictionary should prioritize only the dominant clusters

Reference 53

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:05.923848Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation 9d6bfcb5-31d0-4e1f-84e2-1d5cb285e950 · outbound

This paper cites an unresolved cited work.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Unresolved cited work

Reference 54

Resolution: unresolved; raw_fallback, observed 2026-08-07T14:03:05.688187Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation 05aabd42-d69e-481d-9abc-b6c47954fd7c · outbound

This paper cites The substantial increase in capacity 28 Figure 31:Two-phase model with dictionary size.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs The substantial increase in capacity 28 Figure 31:Two-phase model with dictionary size

Reference 55

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:05.440082Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation 475cb85b-8abe-441b-a0b3-9d13000a2038 · outbound

This paper cites Figure 30:Two-phase model with dictionary size.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Figure 30:Two-phase model with dictionary size

Reference 80

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:06.246169Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation c5fb5432-f866-431a-b510-6ae0accdb7c2 · outbound

This paper cites Figures 29 through 32 demonstrate how dictionary size affects feature reproducibility across the activation frequency spectrum.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs Figures 29 through 32 demonstrate how dictionary size affects feature reproducibility across the activation frequency spectrum

Reference 160

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:06.087258Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation 8cfb17c6-1cd2-4de3-bf46-92a29dd0e9ad · outbound

This paper cites the same.

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs the same

Reference 1000

Resolution: verified fuzzy; raw_fallback, observed 2026-08-07T14:03:05.279790Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Pith citing papers

Observation 29fe38af-348b-4e31-95a9-11839d546684 · inbound

Graph-Regularized Sparse Autoencoders for LLM Safety Steering cites this paper.

Graph-Regularized Sparse Autoencoders for LLM Safety Steering Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs

Reference 16

Resolution: verified exact; arxiv_id, observed 2026-05-21T18:40:28.916617Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.

Observation bab755f3-6fa3-47e9-87e4-fef0bc161a45 · inbound

Make Mechanistic Interpretability Auditable: A Call to Develop Guidelines via Continuous Collaborative Reviewing cites this paper.

Make Mechanistic Interpretability Auditable: A Call to Develop Guidelines via Continuous Collaborative Reviewing Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs

Reference 4

Resolution: verified exact; arxiv_id, observed 2026-07-04T17:09:58.300413Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-20T06:33:59.587034+00:00.