Paper Citation Record · LEDGER

The Geometry of Harmfulness in LLMs through Subconcept Probing

As of 9 August 2026, this Paper Citation Record lists 28 of 28 outbound references and 2 inbound Pith citation observations for arXiv:2507.21141.

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

pith.paper-citation-record.v1
2507.21141 v1

Coverage vector

measured 28 of 28 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-08-06T15:01:47.854894Z

measured 30 of 30 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-09T06:31:02.800959+00:00

measured 2 of 2 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links, observed 2026-08-02T07:09:32.736304Z

measured 0 of 1 external citation measurements

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

Source: arxiv_reference, observed 2026-05-14T21:38:00.071469Z

Reference resolution

28 of 28 outbound references displayed

verified exact0
verified fuzzy4
unresolved24
parse uncertain0
malformed identifier0
metadata mismatch0

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation ecb6aa75-1eaf-4e6c-89a3-701ad29a7dff · outbound

This paper cites On the Opportunities and Risks of Foundation Models.

The Geometry of Harmfulness in LLMs through Subconcept Probing On the Opportunities and Risks of Foundation Models

Reference 3

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.765176Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 82c5c459-c5e7-4430-96b4-4da7071c1999 · outbound

This paper cites Safety-Aware Fine-Tuning of Large Language Models.

The Geometry of Harmfulness in LLMs through Subconcept Probing Safety-Aware Fine-Tuning of Large Language Models

Reference 5

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.772974Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation f360be14-326c-43cc-8947-f019ae1cba39 · outbound

This paper cites Training Verifiers to Solve Math Word Problems.

The Geometry of Harmfulness in LLMs through Subconcept Probing Training Verifiers to Solve Math Word Problems

Reference 6

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.776571Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 7fb6beaf-c30a-4a6e-a4a4-0d5bf79bceec · outbound

This paper cites The Llama 3 Herd of Models.

The Geometry of Harmfulness in LLMs through Subconcept Probing The Llama 3 Herd of Models

Reference 10

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.791537Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a528a80c-c0b7-4f4d-9440-a7b03bb7d901 · outbound

This paper cites Safedpo: A simple approach to direct preference optimization with enhanced safety.

The Geometry of Harmfulness in LLMs through Subconcept Probing Safedpo: A simple approach to direct preference optimization with enhanced safety

Reference 13

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.801834Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6b0343f1-d2dc-41e9-9564-a4d601b82724 · outbound

This paper cites AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models.

The Geometry of Harmfulness in LLMs through Subconcept Probing AutoDAN: Generating Stealthy Jailbreak Prompts on Aligned Large Language Models

Reference 14

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.805088Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a1939382-08e1-4005-bcf6-ae0c11ce47c8 · outbound

This paper cites Enhancing LLM Safety via Constrained Direct Preference Optimization.

The Geometry of Harmfulness in LLMs through Subconcept Probing Enhancing LLM Safety via Constrained Direct Preference Optimization

Reference 15

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.808584Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ff0ef252-7b16-4373-8618-3decc99bb851 · outbound

This paper cites The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasets.

The Geometry of Harmfulness in LLMs through Subconcept Probing The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasets

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.812006Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a681fd9b-fb00-472c-b52d-078489649203 · outbound

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

The Geometry of Harmfulness in LLMs through Subconcept Probing Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models

Reference 17

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.815767Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 509a0729-c1e0-46ab-a923-9a4e81bc055d · outbound

This paper cites Emergent Linear Representations in World Models of Self-Supervised Sequence Models.

The Geometry of Harmfulness in LLMs through Subconcept Probing Emergent Linear Representations in World Models of Self-Supervised Sequence Models

Reference 18

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.819681Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 70ffcc5d-2c1c-4592-ac8a-bbbb34600921 · outbound

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

The Geometry of Harmfulness in LLMs through Subconcept Probing The Hidden Dimensions of LLM Alignment: A Multi-Dimensional Analysis of Orthogonal Safety Directions

Reference 19

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.822986Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 0a8fb2e9-746c-4bd9-808e-5dc127b0f233 · outbound

This paper cites Interpretable steering of large language models with feature guided activation additions.

The Geometry of Harmfulness in LLMs through Subconcept Probing Interpretable steering of large language models with feature guided activation additions

Reference 20

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T15:01:48.208051Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-09T06:31:02.800959+00:00.

Observation a850eefb-8758-4c6a-934d-26aa0304363f · outbound

This paper cites Linear Representations of Sentiment in Large Language Models.

The Geometry of Harmfulness in LLMs through Subconcept Probing Linear Representations of Sentiment in Large Language Models

Reference 21

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.830115Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4ad54f81-b217-4a3e-b17c-67d477fd1619 · outbound

This paper cites Steering Language Models With Activation Engineering.

The Geometry of Harmfulness in LLMs through Subconcept Probing Steering Language Models With Activation Engineering

Reference 22

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.833410Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 47b0c430-ee7b-49a7-acd6-18f093ce0bce · outbound

This paper cites pyvene: A library for understanding and improving PyTorch models via interventions.

The Geometry of Harmfulness in LLMs through Subconcept Probing pyvene: A library for understanding and improving PyTorch models via interventions

Reference 23

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T15:01:48.198757Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-09T06:31:02.800959+00:00.

Observation 3407a27e-cdb7-4591-9266-9b563e4e3180 · outbound

This paper cites URL https://aclanthology.org/2024.naacl-demo.

The Geometry of Harmfulness in LLMs through Subconcept Probing URL https://aclanthology.org/2024.naacl-demo

Reference 24

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T15:01:48.189182Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-09T06:31:02.800959+00:00.

Observation ea24d6d9-d573-408f-b00e-5409aa37c140 · outbound

This paper cites Qwen2 Technical Report.

The Geometry of Harmfulness in LLMs through Subconcept Probing Qwen2 Technical Report

Reference 25

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.843887Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation f5c4d94f-9caa-4c5e-8b5b-e722938ad976 · outbound

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

The Geometry of Harmfulness in LLMs through Subconcept Probing From Directions to Cones: Exploring Multidimensional Representations of Propositional Facts in LLMs

Reference 26

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.847228Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation e59f07d6-563b-4f03-b42d-8bf5f66c7b67 · outbound

This paper cites Under review.

The Geometry of Harmfulness in LLMs through Subconcept Probing Under review

Reference 27

Resolution: verified fuzzy; raw_fallback, observed 2026-08-06T15:01:48.177822Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-09T06:31:02.800959+00:00.

Observation 323d869c-b384-4e58-b0b8-df3e41f18161 · outbound

This paper cites an unresolved cited work.

The Geometry of Harmfulness in LLMs through Subconcept Probing Unresolved cited work

Reference 28

Resolution: unresolved; raw_fallback, observed 2026-08-06T15:01:48.167501Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-09T06:31:02.800959+00:00.

Observation 4a5a077b-c308-4fea-820a-662d0f93c565 · outbound

This paper cites doi: https:// doi.org/10.1016/S0031-3203(96)00142-2.

The Geometry of Harmfulness in LLMs through Subconcept Probing doi: https:// doi.org/10.1016/S0031-3203(96)00142-2

Reference 1997

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.769206Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 580723f0-2f20-47b0-897a-9660e8d9a8e0 · outbound

This paper cites Refusal Behavior in Large Language Models: A Nonlinear Perspective.

The Geometry of Harmfulness in LLMs through Subconcept Probing Refusal Behavior in Large Language Models: A Nonlinear Perspective

Reference 2002

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.794774Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 15050118-aeec-4901-9351-987148f61d04 · outbound

This paper cites emnlp-main.273.

The Geometry of Harmfulness in LLMs through Subconcept Probing emnlp-main.273

Reference 2020

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.787746Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2ad56729-0f6b-43cc-88de-f5f7a594dbb2 · outbound

This paper cites Toy Models of Superposition.

The Geometry of Harmfulness in LLMs through Subconcept Probing Toy Models of Superposition

Reference 2021

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.780339Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 266f749f-500e-42c1-9e67-1ad2a6323e5f · outbound

This paper cites an unresolved cited work.

The Geometry of Harmfulness in LLMs through Subconcept Probing Unresolved cited work

Reference 2022

Resolution: unresolved; raw_fallback, observed 2026-08-06T15:01:48.218134Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-09T06:31:02.800959+00:00.

Observation 2b54dfec-7589-47ba-96ab-80ad776a304e · outbound

This paper cites Refusal in Language Models Is Mediated by a Single Direction.

The Geometry of Harmfulness in LLMs through Subconcept Probing Refusal in Language Models Is Mediated by a Single Direction

Reference 2023

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.757141Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9ab45db4-4393-4a33-b972-58749b0b829a · outbound

This paper cites On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency, pp.

The Geometry of Harmfulness in LLMs through Subconcept Probing On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency, pp

Reference 2024

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.761453Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 89355c50-affb-4354-b6b5-88a10c87a02d · outbound

This paper cites On the Origins of Linear Representations in Large Language Models.

The Geometry of Harmfulness in LLMs through Subconcept Probing On the Origins of Linear Representations in Large Language Models

Reference 2025

Resolution: unresolved; no resolver link, observed 2026-08-06T15:01:47.798462Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Pith citing papers

Observation 05b043eb-10a1-4ebf-9169-085a654c8545 · inbound

Before the Last Token: Diagnosing Final-Token Safety Probe Failures cites this paper.

Before the Last Token: Diagnosing Final-Token Safety Probe Failures The Geometry of Harmfulness in LLMs through Subconcept Probing

Reference 11

Resolution: verified exact; arxiv_id, observed 2026-05-14T21:38:00.073909Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-09T06:31:02.800959+00:00.

Observation 28a2ee74-b0c1-4dc9-a21e-8d16199d1ba7 · inbound

The Entanglement Wall: Activation-Space Probes as Risk Detectors, Not Context Adjudicators cites this paper.

The Entanglement Wall: Activation-Space Probes as Risk Detectors, Not Context Adjudicators The Geometry of Harmfulness in LLMs through Subconcept Probing

Reference 28

Resolution: unresolved; no resolver link, observed 2026-08-02T07:09:32.736304Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.