Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So | Paper Citation Record

Observation 67cb0b3f-d3c2-4087-8501-0e5a49364dff · outbound

This paper cites Three-Phase Transformer.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Three-Phase Transformer

Reference 1

Resolution: verified exact; local_arxiv, observed 2026-08-14T04:19:00.688533Z

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 491e695e-f116-4f30-b042-32049cb4b451 · outbound

This paper cites SliceGPT: Compress Large Language Models by Deleting Rows and Columns.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So SliceGPT: Compress Large Language Models by Deleting Rows and Columns

Reference 2

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:55.234099Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6bc73440-a997-4251-9010-52b871594082 · outbound

This paper cites Converting MLPs into Polynomials in Closed Form.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Converting MLPs into Polynomials in Closed Form

Reference 3

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:55.271257Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a79e7f4e-b4f4-4229-a98f-c8bf44976b81 · outbound

This paper cites Eliciting Latent Predictions from Transformers with the Tuned Lens.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Eliciting Latent Predictions from Transformers with the Tuned Lens

Reference 4

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:55.305254Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation eaeb0e25-f803-4e12-9b6c-9f985f7c8f19 · outbound

This paper cites An Analysis of Residual-Stream Geometry Across Transformer Depth.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So An Analysis of Residual-Stream Geometry Across Transformer Depth

Reference 5

Resolution: verified exact; local_arxiv, observed 2026-08-14T04:19:00.188307Z

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 01f809be-d134-4921-93ae-65ebf0cd7869 · outbound

This paper cites an unresolved cited work.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Unresolved cited work

Reference 6

Resolution: unresolved; raw_fallback, observed 2026-08-14T04:19:02.629677Z

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 ec333105-88f1-4b07-92c1-214968b881f8 · outbound

This paper cites Adam optimizer causes privileged basis in transformer lan- guage model residual stream.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Adam optimizer causes privileged basis in transformer lan- guage model residual stream

Reference 7

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:02.531491Z

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 c7c60df4-8a46-4d58-8ba1-a05ed650e128 · outbound

This paper cites PaLM: Scaling Language Modeling with Pathways.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So PaLM: Scaling Language Modeling with Pathways

Reference 8

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:55.486171Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4482c55d-e451-476d-94d4-8bd5cf7e28e1 · outbound

This paper cites A bias bound for least squares linear regression.Statistica Sinica, 1(1):127–136, 1991.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So A bias bound for least squares linear regression.Statistica Sinica, 1(1):127–136, 1991

Reference 9

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:02.454897Z

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 e62ee521-b9ae-48e6-99cf-19b4fab0160c · outbound

This paper cites A mathematical framework for transformer circuits.Transformer Circuits Thread, 2021.https://transformer-circuits.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So A mathematical framework for transformer circuits.Transformer Circuits Thread, 2021.https://transformer-circuits

Reference 10

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:02.320621Z

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 6b319088-fedb-4bcc-a893-47a03b44710b · outbound

This paper cites Toy models of superposition.Transformer Circuits Thread, 2022.https:// transformer-circuits.pub/2022/toy_model/index.html.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Toy models of superposition.Transformer Circuits Thread, 2022.https:// transformer-circuits.pub/2022/toy_model/index.html

Reference 11

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:02.245991Z

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 c5a5395c-237d-454d-a1a1-c66a05c66b58 · outbound

This paper cites Privileged bases in the transformer residual stream.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Privileged bases in the transformer residual stream

Reference 12

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:02.157889Z

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 21d02de8-3ffe-49ef-8844-291aec7d5d79 · outbound

This paper cites LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding

Reference 13

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:55.764751Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 40b765d8-9a37-4950-a3f1-1967c1c4fd8d · outbound

This paper cites Dynamics of the Transformer Residual Stream: Coupling Spectral Geometry to Network Topology.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Dynamics of the Transformer Residual Stream: Coupling Spectral Geometry to Network Topology

Reference 14

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:55.814750Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 7cea4851-5884-4ba0-8ae6-5f39328e8d25 · outbound

This paper cites A theory of multineuronal dimensionality, dynamics and measurement.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So A theory of multineuronal dimensionality, dynamics and measurement

Reference 15

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:02.090420Z

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 bdb4ef28-ad12-48ee-92c8-29e60142a138 · outbound

This paper cites Transformer Feed-Forward Layers Are Key-Value Memories.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Transformer Feed-Forward Layers Are Key-Value Memories

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:55.899622Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b1dbd7ae-d768-4932-a77c-4f847f52883c · outbound

This paper cites Transformer Feed-Forward Layers Build Predictions by Promoting Concepts in the Vocabulary Space.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Transformer Feed-Forward Layers Build Predictions by Promoting Concepts in the Vocabulary Space

Reference 17

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:55.935100Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 759fef3e-170c-4b6c-ae33-09dd618c175d · outbound

This paper cites Openwebtext corpus.http://Skylion007.github.io/ OpenWebTextCorpus, 2019.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Openwebtext corpus.http://Skylion007.github.io/ OpenWebTextCorpus, 2019

Reference 18

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:55.984862Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d5d5c9c0-f18c-422e-bd6e-389a6800a0fa · outbound

This paper cites Verbalizable representations form a global workspace in language models.Transformer Circuits Thread, 2026.https://transformer-circuits.pub/2026/workspace/, ICLR 2026.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Verbalizable representations form a global workspace in language models.Transformer Circuits Thread, 2026.https://transformer-circuits.pub/2026/workspace/, ICLR 2026

Reference 19

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:02.006478Z

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 1390a450-8c39-414c-a234-4ce3593d7827 · outbound

This paper cites The Concept Allocation Zone: Tracking How Concepts Form Across Transformer Depth.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So The Concept Allocation Zone: Tracking How Concepts Form Across Transformer Depth

Reference 20

Resolution: verified exact; local_arxiv, observed 2026-08-14T04:18:59.604343Z

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 9297b733-66b6-4db6-9c09-6bfd96e47443 · outbound

This paper cites Geometric Evolution Maps: Extracting Stable Concept Probes from Transformer Residual Streams.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Geometric Evolution Maps: Extracting Stable Concept Probes from Transformer Residual Streams

Reference 21

Resolution: verified exact; local_arxiv, observed 2026-08-14T04:18:59.504297Z

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 ff7fe12f-2802-4b22-81ae-fec3ecf9e396 · outbound

This paper cites Decomposing Representation Space into Interpretable Subspaces with Unsupervised Learning.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Decomposing Representation Space into Interpretable Subspaces with Unsupervised Learning

Reference 22

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.154748Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a62ce9ab-8f5e-498a-99c5-1a7a53f031a4 · outbound

This paper cites Residual connections encourage iterative inference.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Residual connections encourage iterative inference

Reference 23

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:01.937216Z

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 84bfa2d7-87e4-4c18-bfa3-a40e5349605a · outbound

This paper cites Unraveling Token Prediction Refinement and Identifying Essential Layers in Language Models.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Unraveling Token Prediction Refinement and Identifying Essential Layers in Language Models

Reference 24

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.244847Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 596967e7-8f74-4fd6-b836-d9086ebe00fc · outbound

This paper cites Similarity of Neural Network Representations Revisited.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Similarity of Neural Network Representations Revisited

Reference 25

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.294747Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d45b0c2a-3d55-4c66-8d6f-f16f8cf7aae4 · outbound

This paper cites The Remarkable Robustness of LLMs: Stages of Inference?.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So The Remarkable Robustness of LLMs: Stages of Inference?

Reference 26

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.305244Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b14cd72c-89f0-4a01-884b-5f7ddee3b0e0 · outbound

This paper cites To FP8 and Back Again: Quantifying Reduced Precision Effects on LLM Training Stability.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So To FP8 and Back Again: Quantifying Reduced Precision Effects on LLM Training Stability

Reference 27

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.315357Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ee100329-ced4-4ccf-a207-2f719f2e2c2f · outbound

This paper cites Regression analysis under link violation.The Annals of Statistics, 17(3):1009–1052, 1989.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Regression analysis under link violation.The Annals of Statistics, 17(3):1009–1052, 1989

Reference 28

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:01.869608Z

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 d9f96146-5725-42ed-b9e0-bbac7b7304f2 · outbound

This paper cites Optimal degrees of synaptic connectivity.Neuron, 93(5):1153–1164, 2017.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Optimal degrees of synaptic connectivity.Neuron, 93(5):1153–1164, 2017

Reference 29

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:01.781473Z

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 fc9f46ee-096a-4707-a1f1-fdd92fa8ac64 · outbound

This paper cites A Geometric Perspective on Next-Token Prediction in Large Language Models: Three Emerging Phases.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So A Geometric Perspective on Next-Token Prediction in Large Language Models: Three Emerging Phases

Reference 30

Resolution: verified exact; local_arxiv, observed 2026-08-14T04:18:59.197943Z

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 6fa04832-b3fe-4a9e-800b-7d6066bc1def · outbound

This paper cites Residual connections and the causal shift: Uncovering a structural misalignment in transformers.arXiv preprint arXiv:2602.14760, 2026.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Residual connections and the causal shift: Uncovering a structural misalignment in transformers.arXiv preprint arXiv:2602.14760, 2026

Reference 31

Resolution: verified exact; raw_fallback, observed 2026-08-14T04:18:59.093043Z

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 ccf13a42-fcef-4b42-b536-dcf3e84e61a3 · outbound

This paper cites Morcos, Maithra Raghu, and Samy Bengio.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Morcos, Maithra Raghu, and Samy Bengio

Reference 32

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:01.713595Z

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 81677136-64b0-45b4-ab56-e082997cd1ca · outbound

This paper cites Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depth.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depth

Reference 33

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.392638Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 66e6a2c6-04b3-4f7c-95b1-1eb7c2fc4cd6 · outbound

This paper cites Interpreting GPT: The logit lens.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Interpreting GPT: The logit lens

Reference 34

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.421234Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9d0f1ddc-92c4-47da-837a-759566557216 · outbound

This paper cites Legible-by-Construction: Attention and End-to-End Transformers.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Legible-by-Construction: Attention and End-to-End Transformers

Reference 35

Resolution: verified exact; local_arxiv, observed 2026-08-14T04:18:58.765292Z

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 e7ae5c3f-a2ff-4ee5-83c2-8872327d04c6 · outbound

This paper cites EE-Tuning: An Economical yet Scalable Solution for Tuning Early-Exit Large Language Models.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So EE-Tuning: An Economical yet Scalable Solution for Tuning Early-Exit Large Language Models

Reference 36

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.449695Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 986ab22f-0df5-4bdf-9a8b-fbd37e8c9cd9 · outbound

This paper cites The LAM- BADA dataset: Word prediction requiring a broad discourse context.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So The LAM- BADA dataset: Word prediction requiring a broad discourse context

Reference 37

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:01.525180Z

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 eff347bd-514f-4485-ae4c-5349ae05b3fa · outbound

This paper cites The Linear Representation Hypothesis and the Geometry of Large Language Models.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So The Linear Representation Hypothesis and the Geometry of Large Language Models

Reference 38

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.476172Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation cebea8f3-c269-43d9-ab2f-7557ef69fcd3 · outbound

This paper cites Language models are unsupervised multitask learners.OpenAI Technical Report, 2019.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Language models are unsupervised multitask learners.OpenAI Technical Report, 2019

Reference 39

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.503252Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 55f96599-a6f7-4f07-9291-96f93fa6016b · outbound

This paper cites SVCCA: Singu- lar vector canonical correlation analysis for deep learning dynamics and interpretability.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So SVCCA: Singu- lar vector canonical correlation analysis for deep learning dynamics and interpretability

Reference 40

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.525316Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation e50d0e10-f8e5-478e-a228-62c15d6b444e · outbound

This paper cites Steering Llama 2 via Contrastive Activation Addition.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Steering Llama 2 via Contrastive Activation Addition

Reference 41

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.574750Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b1282d08-e36a-4b73-9c3e-bb0264b24a21 · outbound

This paper cites The effective rank: A measure of effective dimensionality.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So The effective rank: A measure of effective dimensionality

Reference 42

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:01.241955Z

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 b7a02345-ba6e-4c53-9604-c2b9b9d63743 · outbound

This paper cites A generalized solution of the orthogonal Procrustes problem.Psychome- trika, 31(1):1–10, 1966.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So A generalized solution of the orthogonal Procrustes problem.Psychome- trika, 31(1):1–10, 1966

Reference 43

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:01.126783Z

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 889ab939-35e8-4903-9262-6010522f7628 · outbound

This paper cites Confident Adaptive Language Modeling.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Confident Adaptive Language Modeling

Reference 44

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.724751Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 337ef89c-9eb5-4eff-9ab0-e78077a6153f · outbound

This paper cites Confidence Regulation Neurons in Language Models.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Confidence Regulation Neurons in Language Models

Reference 45

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.774750Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 41323b35-b5b4-455e-8a4e-419507399902 · outbound

This paper cites RoFormer: Enhanced Transformer with Rotary Position Embedding.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So RoFormer: Enhanced Transformer with Rotary Position Embedding

Reference 46

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.834846Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1fa51a08-fa60-432e-8c5b-91a0ef461216 · outbound

This paper cites Signed-Permutation Coordinate Transport for RMSNorm Transformers.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Signed-Permutation Coordinate Transport for RMSNorm Transformers

Reference 47

Resolution: verified exact; local_arxiv, observed 2026-08-14T04:18:57.945348Z

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 31f3fbb8-be37-434c-8e11-b66f4fb3cf71 · outbound

This paper cites Codebook Features: Sparse and Discrete Interpretability for Neural Networks.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Codebook Features: Sparse and Discrete Interpretability for Neural Networks

Reference 48

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.934829Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 540f5297-4f5a-4996-b2ee-9c991b49f9d3 · outbound

This paper cites Steering Language Models With Activation Engineering.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Steering Language Models With Activation Engineering

Reference 49

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:56.974757Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 64693f2a-5add-4809-8303-24a0c2637d0f · outbound

This paper cites PolyPythias: Stabilityandoutliersacrossfiftylanguage model pre-training runs.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So PolyPythias: Stabilityandoutliersacrossfiftylanguage model pre-training runs

Reference 50

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:01.034824Z

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 a2b88ac1-330c-424d-ab61-a97bdcd1489a · outbound

This paper cites Gomez, Lukasz Kaiser, and Illia Polosukhin.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Gomez, Lukasz Kaiser, and Illia Polosukhin

Reference 51

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:00.924785Z

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 8847ced9-fce1-4eea-9327-62f37aa8f14c · outbound

This paper cites BLiMP: The benchmark of linguistic minimal pairs for English.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So BLiMP: The benchmark of linguistic minimal pairs for English

Reference 52

Resolution: verified fuzzy; raw_fallback, observed 2026-08-14T04:19:00.804821Z

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 b7e756b0-cc52-4123-9269-342163c343e0 · outbound

This paper cites Small-scale proxies for large-scale Transformer training instabilities.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Small-scale proxies for large-scale Transformer training instabilities

Reference 53

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:57.107627Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 52fde849-47cb-44e1-822f-431b7f6d426a · outbound

This paper cites Jump to Conclusions: Short-Cutting Transformers With Linear Transformations.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So Jump to Conclusions: Short-Cutting Transformers With Linear Transformations

Reference 54

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:57.118142Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 77de13b4-a1c3-4657-954c-54a199bf3c96 · outbound

This paper cites White-Box Transformers via Sparse Rate Reduction.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So White-Box Transformers via Sparse Rate Reduction

Reference 55

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:57.136528Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3bbac0d5-773b-4418-95a2-937fc9d85d58 · outbound

This paper cites White-Box Transformers via Sparse Rate Reduction: Compression Is All There Is?.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So White-Box Transformers via Sparse Rate Reduction: Compression Is All There Is?

Reference 56

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:57.151867Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d64144b9-1863-49c9-b6d2-6390aff7a2dd · outbound

This paper cites OPT: Open Pre-trained Transformer Language Models.

Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So OPT: Open Pre-trained Transformer Language Models

Reference 57

Resolution: unresolved; no resolver link, observed 2026-08-14T04:18:57.172236Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Coverage vector

Reference resolution

Outbound references

Pith citing papers