Pith. sign in

Paper Citation Record · LEDGER

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders

As of 7 August 2026, this Paper Citation Record lists 23 of 23 outbound references and 2 inbound Pith citation observations for arXiv:2508.16560.

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

pith.paper-citation-record.v1
2508.16560 v4

Coverage vector

measured 23 of 23 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-08-05T17:18:56.858736Z

measured 25 of 25 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-06T06:34:29.942622+00:00

measured 2 of 2 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links, observed 2026-08-05T17:26:48.667582Z

measured 0 of 1 external citation measurements

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

Source: arxiv_reference, observed 2026-07-02T01:36:25.794218Z

Reference resolution

23 of 23 outbound references displayed

  • verified exact0
  • verified fuzzy8
  • unresolved15
  • parse uncertain0
  • malformed identifier0
  • metadata mismatch0

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation ea91efb9-c09b-47e8-970d-cccdc32e32e5 · outbound

This paper cites Feature hedging: Correlated features break narrow sparse autoencoders.arXiv preprint arXiv:2505.11756,.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders Feature hedging: Correlated features break narrow sparse autoencoders.arXiv preprint arXiv:2505.11756,

Reference 4

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:54.509236Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:54.509236Z digest=sha256:94926017ff403c94b5185f90ca735676c0bfe8586c418d456f10f75f6010c305

Observation 5c05d986-410d-4ef0-b766-33eb185b1a40 · outbound

This paper cites Dictionary learning optimization techniques.https: //transformer-circuits.pub/2025/january-update,.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders Dictionary learning optimization techniques.https: //transformer-circuits.pub/2025/january-update,

Reference 5

Resolution
verified fuzzy
raw_fallback, observed 2026-08-05T17:18:59.568468Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-06T06:34:29.942622+00:00.

source=pdf_text observed=2026-08-05T17:18:54.614738Z digest=sha256:3f0c84bec7c992667327d996565c573d6b888831dd4dc92ff04dd9f4ba29a0de

Observation f0e611eb-21d1-488a-91b6-ac57c2747d49 · outbound

This paper cites The Llama 3 Herd of Models.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders The Llama 3 Herd of Models

Reference 6

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:54.777391Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:54.777391Z digest=sha256:433914e711009618b675c56ce3891e595d26d5ce92e2d02227cce69287691048

Observation 5a2fde88-c213-45d6-b2c7-4630a1c72e8b · outbound

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

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders The Pile: An 800GB Dataset of Diverse Text for Language Modeling

Reference 8

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:55.043237Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:55.043237Z digest=sha256:35fadc9ad868f860a284cc8e927c1f3c29e9f11293ccbc605e9e36174aaee53b

Observation 85e4a851-3abf-439e-8848-9eda8427f7e1 · outbound

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

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders Are Sparse Autoencoders Useful? A Case Study in Sparse Probing

Reference 10

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:55.289843Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:55.289843Z digest=sha256:96d415bd8f1800e6b0ac9813c2461229811076297f38a7849f18f8f83bddfb73

Observation 17437237-317a-4538-84de-91207aabd307 · outbound

This paper cites Adam: A Method for Stochastic Optimization.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders Adam: A Method for Stochastic Optimization

Reference 11

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:55.423730Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:55.423730Z digest=sha256:f3186711548fdf77d7d620384acb2c5db1e930974aada51b1da4746e5cfec1a5

Observation 99b5aca8-e888-426f-8387-ad271dc4b08a · outbound

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

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders

Reference 12

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:55.529394Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:55.529394Z digest=sha256:a38f404a5b3a2a76b34c26d200c1bc492385cc5633785916c296f5dffc3e6f4f

Observation beab2e75-a3f0-4462-af17-39b6f5dce6a8 · outbound

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

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders Gemma 2: Improving Open Language Models at a Practical Size

Reference 13

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:55.632316Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:55.632316Z digest=sha256:b35a4efb6e4729954ee09b7383fd95dce1e8e955f30814e4b33b7c3d9ca5acf0

Observation 5f90aaef-7103-4c7f-a616-dc6de6694890 · outbound

This paper cites AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders

Reference 14

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:55.749890Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:55.749890Z digest=sha256:6322a2437aec73f13ed042e5f70c216d8870b423330045428e7bb8c4f67628cc

Observation 525d2484-d664-4c1a-a87e-d8cd12acdb0d · outbound

This paper cites an unresolved cited work.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders Unresolved cited work

Reference 15

Resolution
unresolved
raw_fallback, observed 2026-08-05T17:18:59.288062Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-06T06:34:29.942622+00:00.

source=pdf_text observed=2026-08-05T17:18:55.883653Z digest=sha256:6c1dda9a1f04b43ccd22604bc3fd7a93095febb6f3dabb884cf751e19aa31bb8

Observation 239fed1d-0bc9-4536-9b64-629195747781 · outbound

This paper cites For BatchTopK SAEs, there is no sparsity penalty as sparsity is enforced by the BatchTopK function.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders For BatchTopK SAEs, there is no sparsity penalty as sparsity is enforced by the BatchTopK function

Reference 16

Resolution
verified fuzzy
raw_fallback, observed 2026-08-05T17:18:59.014988Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-06T06:34:29.942622+00:00.

source=pdf_text observed=2026-08-05T17:18:55.992231Z digest=sha256:53690d2ce850e3ec41be1442a73bed147d525a6bde141054abecf9da3cf23189

Observation 4a8456d8-9368-475d-a7eb-61530f26297d · outbound

This paper cites We use a learning rate of3e −4 with no warmup or decay.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders We use a learning rate of3e −4 with no warmup or decay

Reference 17

Resolution
verified fuzzy
raw_fallback, observed 2026-08-05T17:18:58.683267Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-06T06:34:29.942622+00:00.

source=pdf_text observed=2026-08-05T17:18:56.102650Z digest=sha256:860d3c26271da4e4725fed84ab1e957bab26a270fedb82ed1a4edc9cea6151a7

Observation d987020c-7963-48b9-bb95-b9b2ee45d246 · outbound

This paper cites The metric is minimized at the true L0.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders The metric is minimized at the true L0

Reference 18

Resolution
verified fuzzy
raw_fallback, observed 2026-08-05T17:18:58.510680Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-06T06:34:29.942622+00:00.

source=pdf_text observed=2026-08-05T17:18:56.234427Z digest=sha256:bcd73a4a8ee9a7dca043b6fe5ca7030af75e00fef21eae127b5366d94f5c00ab

Observation 6919888b-c2b2-4854-9f47-0059caaf9dce · outbound

This paper cites We include all SAEs cross-listed in both SAELens and Neuronpedia with an L0 reported in SAELens.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders We include all SAEs cross-listed in both SAELens and Neuronpedia with an L0 reported in SAELens

Reference 19

Resolution
verified fuzzy
raw_fallback, observed 2026-08-05T17:18:58.332535Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-06T06:34:29.942622+00:00.

source=pdf_text observed=2026-08-05T17:18:56.373953Z digest=sha256:3f9c5f97f0c37831459069fe733091f50e00c8373a9ca6528901f8c45b307a5b

Observation 8dc7b0e2-e097-45f2-bea0-1b5457b6d0f3 · outbound

This paper cites However, we find that most open-source SAEs have L0 below 100, much lower than our analysis expects to be ideal.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders However, we find that most open-source SAEs have L0 below 100, much lower than our analysis expects to be ideal

Reference 20

Resolution
verified fuzzy
raw_fallback, observed 2026-08-05T17:18:58.019875Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-06T06:34:29.942622+00:00.

source=pdf_text observed=2026-08-05T17:18:56.502070Z digest=sha256:77827aa7150525e3f9a0d070b05c4705cd0a3a411e83427520c789f5662681d0

Observation 4fbd6256-801c-46a9-8c34-eebb4dfd2ef2 · outbound

This paper cites elbow” point, but the “elbow.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders elbow” point, but the “elbow

Reference 21

Resolution
verified fuzzy
raw_fallback, observed 2026-08-05T17:18:57.791364Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-06T06:34:29.942622+00:00.

source=pdf_text observed=2026-08-05T17:18:56.624234Z digest=sha256:881e3b7fa32c35ff414ed437ba675d72c7421b3ebe6e39a42f49d14d2f1c0476

Observation a44e8dae-29e9-4470-863e-96abbc676ec7 · outbound

This paper cites an unresolved cited work.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders Unresolved cited work

Reference 22

Resolution
unresolved
raw_fallback, observed 2026-08-05T17:18:57.520272Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-06T06:34:29.942622+00:00.

source=pdf_text observed=2026-08-05T17:18:56.724145Z digest=sha256:fb32cc6b3ac46ec3ce33a0df8d16f7fc5425daf780db69aff176e0e0dafbb680

Observation b65905ce-f407-444d-8b5a-fdc66c4b27fb · outbound

This paper cites The threshold for BatchTopK is much higher than it is for JumpReLU, and the threshold decreases as L0 increses.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders The threshold for BatchTopK is much higher than it is for JumpReLU, and the threshold decreases as L0 increses

Reference 23

Resolution
verified fuzzy
raw_fallback, observed 2026-08-05T17:18:57.294683Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-06T06:34:29.942622+00:00.

source=pdf_text observed=2026-08-05T17:18:56.858736Z digest=sha256:551d137f95d8b8a16bdc13a8cb1d53e0522f0cbbe61ecfc0d634f7174036308d

Observation 5400b549-3046-479c-8abe-3340cb3bbf18 · outbound

This paper cites Scaling and evaluating sparse autoencoders.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders Scaling and evaluating sparse autoencoders

Reference 2020

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:55.166754Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:55.166754Z digest=sha256:17a8f2854a2daa99db4f48b535df268b09d3997a8b50a5e5bdb1f3cd6a907f16

Observation ba039a99-074f-4e6d-bd04-a9cf50401773 · outbound

This paper cites Decomposing The Dark Matter of Sparse Autoencoders.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders Decomposing The Dark Matter of Sparse Autoencoders

Reference 2022

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:54.886073Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:54.886073Z digest=sha256:fbc96d57f25c04f5ffb75ec64cd75349ca93184b9eec80d8fad6f3b92a88a564

Observation c0c8dfd4-f44e-49bd-9257-744ca2bd4923 · outbound

This paper cites BatchTopK Sparse Autoencoders.

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders BatchTopK Sparse Autoencoders

Reference 2023

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:54.161687Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:54.161687Z digest=sha256:34ad33448d46a666d6ce05cca7deb08f25818ae2f3a844cc70019ab7678d5cd0

Observation dd93be14-1c3c-4914-b46a-4a3b389dff4b · outbound

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

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders Learning Multi-Level Features with Matryoshka Sparse Autoencoders

Reference 2024

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:54.254819Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:54.254819Z digest=sha256:9328862f989ae3e97b1b48b7327bdef52ae2330debf4f6497a86f46de379b739

Observation e9240924-e2ae-45e8-81d5-21e6a10bb076 · outbound

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

Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders A is for absorption: Studying feature splitting and absorption in sparse autoencoders.arXiv preprint arXiv:2409.14507,

Reference 2025

Resolution
unresolved
no resolver link, observed 2026-08-05T17:18:54.344796Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:18:54.344796Z digest=sha256:5c0f592d3cd5aa27f95731ad37e6f06d17a9a59efa7dbbc66d29493d33bff4d3

Pith citing papers

Observation 5582e680-8fbd-4b78-9a34-99908386a6a4 · inbound

SATORI: Static Test Oracle Generation for REST APIs cites this paper.

SATORI: Static Test Oracle Generation for REST APIs Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders

Reference 60

Resolution
unresolved
no resolver link, observed 2026-08-05T17:26:48.667582Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-05T17:26:48.667582Z digest=sha256:eed3827e31d2d568b88a6c86b94f08d0ba51bba2be3f7aa341c0bf5fb8865495

Observation dfa80f82-a054-4b72-a244-305da658db4a · inbound

Perplexity Can Miss SAE Feature Damage Under Quantization cites this paper.

Perplexity Can Miss SAE Feature Damage Under Quantization Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders

Reference 4

Resolution
verified exact
arxiv_id, observed 2026-07-08T02:18:07.154573Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-06T06:34:29.942622+00:00.

source=pdf_text observed=2026-06-28T11:41:18.460538Z digest=sha256:987a161be07e762df8c63afe49cd8bd2304daf05740424d598dae42ada483f66