Not All Language Model Features Are One-Dimensionally Linear | Paper Citation Record

Observation 096ba099-9cbd-4319-80cd-71e900b35d8d · inbound

Modular addition without black-boxes: Compressing explanations of MLPs that compute numerical integration cites this paper.

Modular addition without black-boxes: Compressing explanations of MLPs that compute numerical integration Not All Language Model Features Are One-Dimensionally Linear

Reference 6

Resolution: unresolved; no resolver link, observed 2026-08-11T22:13:00.050986Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 10966fcf-b421-4196-ba98-0ec0ad365332 · inbound

A Survey on Uncertainty Quantification of Large Language Models: Taxonomy, Open Research Challenges, and Future Directions cites this paper.

A Survey on Uncertainty Quantification of Large Language Models: Taxonomy, Open Research Challenges, and Future Directions Not All Language Model Features Are One-Dimensionally Linear

Reference 52

Resolution: unresolved; no resolver link, observed 2026-08-11T20:37:54.701796Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation f1649718-66fc-49e3-b8ab-b4a12b9dfa19 · inbound

Beyond Scalars: Concept-Based Alignment Analysis in Vision Transformers cites this paper.

Beyond Scalars: Concept-Based Alignment Analysis in Vision Transformers Not All Language Model Features Are One-Dimensionally Linear

Reference 11

Resolution: unresolved; no resolver link, observed 2026-08-11T19:32:02.608302Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c6c496f7-5c29-4497-b0ad-e1c114b92182 · inbound

GPT-2 Through the Lens of Vector Symbolic Architectures cites this paper.

GPT-2 Through the Lens of Vector Symbolic Architectures Not All Language Model Features Are One-Dimensionally Linear

Reference 10

Resolution: unresolved; no resolver link, observed 2026-08-11T18:26:31.774536Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a8055c1c-0c4d-4676-b528-4b47258fb230 · inbound

ICLR: In-Context Learning of Representations cites this paper.

ICLR: In-Context Learning of Representations Not All Language Model Features Are One-Dimensionally Linear

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-10T23:24:01.441554Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2ed65907-e655-418a-b366-f4ce1fabd5ce · inbound

Analyzing Finetuning Representation Shift for Multimodal LLMs Steering cites this paper.

Analyzing Finetuning Representation Shift for Multimodal LLMs Steering Not All Language Model Features Are One-Dimensionally Linear

Reference 14

Resolution: unresolved; no resolver link, observed 2026-08-10T22:03:24.935618Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5874e8a3-6850-4b3d-b82f-6b03d75d2222 · inbound

Interpretability in Parameter Space: Minimizing Mechanistic Description Length with Attribution-based Parameter Decomposition cites this paper.

Interpretability in Parameter Space: Minimizing Mechanistic Description Length with Attribution-based Parameter Decomposition Not All Language Model Features Are One-Dimensionally Linear

Reference 23

Resolution: unresolved; no resolver link, observed 2026-08-10T14:55:02.009949Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation f1ef7e34-c91d-4568-bb8a-66a4bfb724d0 · inbound

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

Sparse Autoencoders Trained on the Same Data Learn Different Features Not All Language Model Features Are One-Dimensionally Linear

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-10T11:58:37.064303Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3865ada9-88fb-44ec-8644-3551e37bdef6 · inbound

Language Models Use Trigonometry to Do Addition cites this paper.

Language Models Use Trigonometry to Do Addition Not All Language Model Features Are One-Dimensionally Linear

Reference 7

Resolution: unresolved; no resolver link, observed 2026-08-09T17:29:00.125291Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation bd548730-d06c-4725-b01c-4da06c7ccaeb · inbound

Harmonic Loss Trains Interpretable AI Models cites this paper.

Harmonic Loss Trains Interpretable AI Models Not All Language Model Features Are One-Dimensionally Linear

Reference 21

Resolution: unresolved; no resolver link, observed 2026-08-09T14:52:24.255125Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 53139257-4cc2-4178-a89a-e7ad60874756 · inbound

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

Sparse Autoencoders Do Not Find Canonical Units of Analysis Not All Language Model Features Are One-Dimensionally Linear

Reference 2022

Resolution: unresolved; no resolver link, observed 2026-08-08T21:12:13.824293Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation da61c799-cd6e-46da-bef2-457e7d73ce2e · inbound

The Hidden Dimensions of LLM Alignment: A Multi-Dimensional Analysis of Orthogonal Safety Directions cites this paper.

The Hidden Dimensions of LLM Alignment: A Multi-Dimensional Analysis of Orthogonal Safety Directions Not All Language Model Features Are One-Dimensionally Linear

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-07T23:08:02.254996Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2d3c7a0a-447f-40cd-854d-2819fa76b881 · inbound

Representation Learning on a Random Lattice cites this paper.

Representation Learning on a Random Lattice Not All Language Model Features Are One-Dimensionally Linear

Reference 2022

Resolution: unresolved; no resolver link, observed 2026-08-16T05:41:19.077273Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9a93bcfe-f4cc-4a52-9e56-a716c5208000 · inbound

Interpretable Risk Mitigation in LLM Agent Systems cites this paper.

Interpretable Risk Mitigation in LLM Agent Systems Not All Language Model Features Are One-Dimensionally Linear

Reference 23

Resolution: unresolved; no resolver link, observed 2026-08-15T21:11:01.891128Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 44a87f73-8562-4a2b-98f8-8fb523dc1847 · inbound

From Directions to Cones: Exploring Multidimensional Representations of Propositional Facts in LLMs cites this paper.

From Directions to Cones: Exploring Multidimensional Representations of Propositional Facts in LLMs Not All Language Model Features Are One-Dimensionally Linear

Reference 13

Resolution: unresolved; no resolver link, observed 2026-08-07T13:28:00.819179Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b3f89a17-2f54-4b91-b149-3593b7a074d4 · inbound

Sparsification and Reconstruction from the Perspective of Representation Geometry cites this paper.

Sparsification and Reconstruction from the Perspective of Representation Geometry Not All Language Model Features Are One-Dimensionally Linear

Reference 9

Resolution: unresolved; no resolver link, observed 2026-08-07T13:11:10.172162Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation da87d8a2-5ba9-4978-81d9-3c72e69790c1 · inbound

Incorporating Hierarchical Semantics in Sparse Autoencoder Architectures cites this paper.

Incorporating Hierarchical Semantics in Sparse Autoencoder Architectures Not All Language Model Features Are One-Dimensionally Linear

Reference 8

Resolution: unresolved; no resolver link, observed 2026-08-07T11:56:15.166919Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a2ef8bf8-f3bb-49ea-bc85-52577a1bc164 · inbound

Internal Value Alignment in Large Language Models through Controlled Value Vector Activation cites this paper.

Internal Value Alignment in Large Language Models through Controlled Value Vector Activation Not All Language Model Features Are One-Dimensionally Linear

Reference 12

Resolution: unresolved; no resolver link, observed 2026-08-06T17:17:28.984932Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 57afb69c-33db-495f-9a62-2aa63a6f8628 · inbound

Understanding sparse autoencoder scaling in the presence of feature manifolds cites this paper.

Understanding sparse autoencoder scaling in the presence of feature manifolds Not All Language Model Features Are One-Dimensionally Linear

Reference 22

Resolution: unresolved; no resolver link, observed 2026-08-15T16:41:17.926938Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4cad7f70-6925-454b-84c0-1e076421cc8e · inbound

How Do Transformers Learn to Associate Tokens: Gradient Leading Terms Bring Mechanistic Interpretability cites this paper.

How Do Transformers Learn to Associate Tokens: Gradient Leading Terms Bring Mechanistic Interpretability Not All Language Model Features Are One-Dimensionally Linear

Reference 7

Resolution: metadata mismatch; arxiv_id, observed 2026-05-16T11:20:52.634674Z

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 caee3fe2-910b-4c01-a668-51ceeffa284f · inbound

Logit Distance Bounds Representational Similarity cites this paper.

Logit Distance Bounds Representational Similarity Not All Language Model Features Are One-Dimensionally Linear

Reference 2006

Resolution: unresolved; no resolver link, observed 2026-08-02T22:57:36.025860Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b64c2577-70da-4c6d-aee8-65fc99871aa1 · inbound

The Lattice Representation Hypothesis of Large Language Models cites this paper.

The Lattice Representation Hypothesis of Large Language Models Not All Language Model Features Are One-Dimensionally Linear

Reference 5

Resolution: verified exact; arxiv_id, observed 2026-05-21T12:04:09.560248Z

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 1401b968-426c-41ed-9981-730516cecf67 · inbound

The Lattice Representation Hypothesis of Large Language Models cites this paper.

The Lattice Representation Hypothesis of Large Language Models Not All Language Model Features Are One-Dimensionally Linear

Reference 2025

Resolution: unresolved; no resolver link, observed 2026-08-02T19:43:25.400771Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1971c98c-41e4-433a-b9eb-9d18adedf9b8 · inbound

Predicting Where Steering Vectors Succeed cites this paper.

Predicting Where Steering Vectors Succeed Not All Language Model Features Are One-Dimensionally Linear

Reference 5

Resolution: metadata mismatch; arxiv_id, observed 2026-05-10T11:00:03.823175Z

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 c6473e49-605c-40b6-8393-ef777b005572 · inbound

There Will Be a Scientific Theory of Deep Learning cites this paper.

There Will Be a Scientific Theory of Deep Learning Not All Language Model Features Are One-Dimensionally Linear

Reference 217

Resolution: metadata mismatch; arxiv_id, observed 2026-05-11T15:21:09.121872Z

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 fdb5656b-db9c-45e3-9f20-601071dc2f16 · inbound

H-Probes: Extracting Hierarchical Structures From Latent Representations of Language Models cites this paper.

H-Probes: Extracting Hierarchical Structures From Latent Representations of Language Models Not All Language Model Features Are One-Dimensionally Linear

Reference 5

Resolution: metadata mismatch; arxiv_id, observed 2026-05-10T14:10:28.842529Z

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 332ab266-1b4f-47a3-93f8-997d92b30daf · inbound

Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior cites this paper.

Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior Not All Language Model Features Are One-Dimensionally Linear

Reference 205

Resolution: metadata mismatch; arxiv_id, observed 2026-05-11T17:16:06.969598Z

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 d7a541ed-ae90-49c3-b09f-74f8fa3620d5 · inbound

Tree SAE: Learning Hierarchical Feature Structures in Sparse Autoencoders cites this paper.

Tree SAE: Learning Hierarchical Feature Structures in Sparse Autoencoders Not All Language Model Features Are One-Dimensionally Linear

Reference 8

Resolution: metadata mismatch; arxiv_id, observed 2026-05-11T03:15:54.387341Z

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 a57d3192-6627-4e4c-8f6b-b0834607e294 · inbound

Tree SAE: Learning Hierarchical Feature Structures in Sparse Autoencoders cites this paper.

Tree SAE: Learning Hierarchical Feature Structures in Sparse Autoencoders Not All Language Model Features Are One-Dimensionally Linear

Reference 8

Resolution: metadata mismatch; arxiv_id, observed 2026-05-12T07:16:25.371063Z

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 f78a01aa-39b5-4d40-ab9e-17cd95f9cd34 · inbound

Tool Calling is Linearly Readable and Steerable in Language Models cites this paper.

Tool Calling is Linearly Readable and Steerable in Language Models Not All Language Model Features Are One-Dimensionally Linear

Reference 48

Resolution: verified exact; arxiv_id, observed 2026-05-11T03:10:52.721023Z

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 629d436a-31e6-4e78-a889-54c1723729ec · inbound

Tensor Product Representation Probes Reveal Shared Structure Across Linear Directions cites this paper.

Tensor Product Representation Probes Reveal Shared Structure Across Linear Directions Not All Language Model Features Are One-Dimensionally Linear

Reference 6

Resolution: metadata mismatch; arxiv_id, observed 2026-05-12T07:16:30.298763Z

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 9bc12940-3a01-4c8c-8e36-5dcc4ce30711 · inbound

WriteSAE: Sparse Autoencoders for Recurrent State cites this paper.

WriteSAE: Sparse Autoencoders for Recurrent State Not All Language Model Features Are One-Dimensionally Linear

Reference 39

Resolution: metadata mismatch; arxiv_id, observed 2026-05-14T20:59:28.564457Z

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 0e9c5996-8459-4dac-9d93-5d79ffec6cdf · inbound

WriteSAE: Sparse Autoencoders for Recurrent State cites this paper.

WriteSAE: Sparse Autoencoders for Recurrent State Not All Language Model Features Are One-Dimensionally Linear

Reference 39

Resolution: metadata mismatch; arxiv_id, observed 2026-05-15T04:59:45.200561Z

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 6bc4200b-a4f3-4338-9320-f39f80371e64 · inbound

WriteSAE: Sparse Autoencoders for Recurrent State cites this paper.

WriteSAE: Sparse Autoencoders for Recurrent State Not All Language Model Features Are One-Dimensionally Linear

Reference 16

Resolution: metadata mismatch; arxiv_id, observed 2026-05-21T07:49:50.100762Z

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 383a1443-61e9-4889-b6b8-cbd31759d547 · inbound

Rethinking Layer Relevance in Large Language Models Beyond Cosine Similarity cites this paper.

Rethinking Layer Relevance in Large Language Models Beyond Cosine Similarity Not All Language Model Features Are One-Dimensionally Linear

Reference 28

Resolution: metadata mismatch; arxiv_id, observed 2026-05-15T05:09:46.256471Z

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 7cf0129d-1f53-4d86-85b0-50bef9abe05b · inbound

Geometry of Human Perceptual Domains Emerges Transiently in LLM Representations cites this paper.

Geometry of Human Perceptual Domains Emerges Transiently in LLM Representations Not All Language Model Features Are One-Dimensionally Linear

Reference 5

Resolution: verified exact; arxiv_id, observed 2026-06-29T12:33:24.317059Z

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 58141d5e-4aea-4be7-8f3c-c61e8ef479d1 · inbound

Sign-Aware Gated Sparse Autoencoders: Modeling Anticorrelated Features with Bi-Jump-ReLU Activations cites this paper.

Sign-Aware Gated Sparse Autoencoders: Modeling Anticorrelated Features with Bi-Jump-ReLU Activations Not All Language Model Features Are One-Dimensionally Linear

Reference 7

Resolution: verified exact; arxiv_id, observed 2026-06-29T14:23:30.789091Z

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 34173d6d-fdf3-4433-848d-c5ddf4e0862f · inbound

Sign-Aware Gated Sparse Autoencoders: Modeling Anticorrelated Features with Bi-Jump-ReLU Activations cites this paper.

Sign-Aware Gated Sparse Autoencoders: Modeling Anticorrelated Features with Bi-Jump-ReLU Activations Not All Language Model Features Are One-Dimensionally Linear

Reference 7

Resolution: unresolved; no resolver link, observed 2026-08-04T05:02:51.288089Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation e59b3cf3-c9b0-459a-8df8-a775b5f23dc1 · inbound

Temporal Preference Concepts and their Functions in a Large Language Model cites this paper.

Temporal Preference Concepts and their Functions in a Large Language Model Not All Language Model Features Are One-Dimensionally Linear

Reference 25

Resolution: verified exact; arxiv_id, observed 2026-07-01T14:05:47.187058Z

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 1db4c80b-7165-4016-96cd-421e242dd481 · inbound

Temporal Preference Concepts and their Functions in a Large Language Model cites this paper.

Temporal Preference Concepts and their Functions in a Large Language Model Not All Language Model Features Are One-Dimensionally Linear

Reference 25

Resolution: unresolved; no resolver link, observed 2026-07-12T17:03:44.315006Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 05e3641c-e8cd-4b4f-a8cd-6ecc0c91b122 · inbound

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

Subspace-Aware Sparse Autoencoders for Effective Mechanistic Interpretability Not All Language Model Features Are One-Dimensionally Linear

Reference 1

Resolution: metadata mismatch; arxiv_id, observed 2026-06-28T02:11:29.085215Z

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 4252f86c-9887-4d15-8bd0-7ade27ff6397 · inbound

Closure-Validated Circuit Discovery in Attention Heads: Co-activation Proposes, Ablation Disposes cites this paper.

Closure-Validated Circuit Discovery in Attention Heads: Co-activation Proposes, Ablation Disposes Not All Language Model Features Are One-Dimensionally Linear

Reference 4

Resolution: verified exact; arxiv_id, observed 2026-07-03T00:17:29.333025Z

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 548f1b33-918c-4b0f-90f1-a7202210d2ca · inbound

Muon Learns More Robust and Transferable Features than Adam cites this paper.

Muon Learns More Robust and Transferable Features than Adam Not All Language Model Features Are One-Dimensionally Linear

Reference 120

Resolution: metadata mismatch; arxiv_id, observed 2026-07-03T00:27:30.214165Z

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 c31441fa-cd1c-408f-b70e-46770a6a16db · inbound

Density Ridge Selective Prediction for LLM and VLM Hallucination Detection under Calibration Label Scarcity cites this paper.

Density Ridge Selective Prediction for LLM and VLM Hallucination Detection under Calibration Label Scarcity Not All Language Model Features Are One-Dimensionally Linear

Reference 14

Resolution: metadata mismatch; arxiv_id, observed 2026-07-03T00:27:29.503504Z

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 bdb436bc-7722-4224-897d-5d9c96f4e129 · inbound

Size Doesn't Matter: Cosine-Scored Sparse Autoencoders cites this paper.

Size Doesn't Matter: Cosine-Scored Sparse Autoencoders Not All Language Model Features Are One-Dimensionally Linear

Reference 32

Resolution: metadata mismatch; arxiv_id, observed 2026-07-01T07:15:29.669716Z

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 abdc302e-9d48-4cd8-99e0-f773d1af635c · inbound

Structuring Sparsity: Block-Sparse Featurizers Capture Visual Concept Manifolds cites this paper.

Structuring Sparsity: Block-Sparse Featurizers Capture Visual Concept Manifolds Not All Language Model Features Are One-Dimensionally Linear

Reference 35

Resolution: verified exact; arxiv_id, observed 2026-07-04T17:09:58.969579Z

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 055b7f6e-8c13-43d7-a417-8df7f05eb430 · inbound

At the Edge of Understanding: Sparse Autoencoders Trace The Limits of Transformer Generalization cites this paper.

At the Edge of Understanding: Sparse Autoencoders Trace The Limits of Transformer Generalization Not All Language Model Features Are One-Dimensionally Linear

Reference 41

Resolution: metadata mismatch; arxiv_id, observed 2026-06-26T01:28:50.560096Z

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 b9b88ed0-c644-481f-9383-ead7b3cdb1a3 · inbound

Do Models Read What They Write? Causal Registers in Scratchpad Reasoning cites this paper.

Do Models Read What They Write? Causal Registers in Scratchpad Reasoning Not All Language Model Features Are One-Dimensionally Linear

Reference 1

Resolution: verified exact; arxiv_id, observed 2026-06-30T07:34:21.854001Z

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 6fa62143-c4a6-4e01-b16f-2f525b11630a · inbound

Training, Reading, and Editing Legible Transformers cites this paper.

Training, Reading, and Editing Legible Transformers Not All Language Model Features Are One-Dimensionally Linear

Reference 55

Resolution: unresolved; no resolver link, observed 2026-07-13T05:35:58.568346Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b8f70879-0299-4241-8ecb-01922e686687 · inbound

Are Single-Token Sparse Autoencoder Features Causally Necessary? Layer-Depth and SAE-Family Effects cites this paper.

Are Single-Token Sparse Autoencoder Features Causally Necessary? Layer-Depth and SAE-Family Effects Not All Language Model Features Are One-Dimensionally Linear

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-01T10:03:54.939594Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c5e874e5-8fb3-4fe0-9415-af59f4f32906 · inbound

Context Is King: How In-Context Specification Shapes the Geometry of Concepts cites this paper.

Context Is King: How In-Context Specification Shapes the Geometry of Concepts Not All Language Model Features Are One-Dimensionally Linear

Reference 7

Resolution: unresolved; no resolver link, observed 2026-07-31T15:17:59.776360Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3787f0e2-4aa9-42ae-a8ed-f1d15f4245f8 · inbound

Where You Measure Decides What You Measure: Position Selection in Ablation-Based SAE Evaluation cites this paper.

Where You Measure Decides What You Measure: Position Selection in Ablation-Based SAE Evaluation Not All Language Model Features Are One-Dimensionally Linear

Reference 21

Resolution: unresolved; no resolver link, observed 2026-08-14T13:29:00.130261Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Coverage vector

Reference resolution

Outbound references

Pith citing papers