Typed states for the displayed outbound observations.
Source: paper_references, paper_reference_links, observed 2026-08-06T17:46:22.465170Z
Paper Citation Record · LEDGER
As of 7 August 2026, this Paper Citation Record lists 78 of 78 outbound references and 0 inbound Pith citation observations for arXiv:2507.10106.
A citation records a reference. It does not transfer a finding from one paper to another.
Typed states for the displayed outbound observations.
Source: paper_references, paper_reference_links, observed 2026-08-06T17:46:22.465170Z
One-hop event checks from named stored sources.
Source: scholarly_work_events, retraction_status_cache, observed 2026-08-06T06:34:29.942622+00:00
Pith citing papers itemized under the disclosed page cap.
Source: paper_references, paper_reference_links
A source-named dated measurement, never combined with another source.
Source: cited_works
78 of 78 outbound references displayed
External citation measurements
No source-named external measurement is stored.
Observation c88b976d-03ed-4382-8076-217bd43b3a34 · outbound
BlueGlass: A Framework for Composite AI Safety write newline
Reference 1
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 724930c3-b3fe-4b85-899b-4cc49bcdc7be · outbound
BlueGlass: A Framework for Composite AI Safety Sanity Checks for Saliency Maps
Reference 2
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation a6f99e68-2dc0-42e8-9c9d-f2aebe854de5 · outbound
BlueGlass: A Framework for Composite AI Safety Understanding intermediate layers using linear classifier probes
Reference 3
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 409004ff-8edb-4db6-9268-463a36c4b7d1 · outbound
BlueGlass: A Framework for Composite AI Safety Physics of Language Models: Part 3.1, Knowledge Storage and Extraction
Reference 4
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation bccc730e-1a91-4613-ade5-2b6a9024088e · outbound
BlueGlass: A Framework for Composite AI Safety Apache parquet
Reference 5
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.
Observation e97cc8f2-9068-49ee-ba09-53893093cb6f · outbound
BlueGlass: A Framework for Composite AI Safety Apache arrow: A cross-language development platform for in-memory data
Reference 6
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.
Observation 1fba99e1-0585-455d-a137-20b30e57c618 · outbound
BlueGlass: A Framework for Composite AI Safety Refusal in Language Models Is Mediated by a Single Direction
Reference 7
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 0dbd63e7-ed16-4b54-a06c-59c0a0d00814 · outbound
BlueGlass: A Framework for Composite AI Safety A Survey of Word Embeddings Evaluation Methods
Reference 8
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.
Observation 342c7307-d71a-46a7-b46b-9edb473814ec · outbound
BlueGlass: A Framework for Composite AI Safety Mechanistic Interpretability for AI Safety -- A Review
Reference 9
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 3065a345-7af3-4c74-b4f5-3f4676bccafe · outbound
BlueGlass: A Framework for Composite AI Safety Unresolved cited work
Reference 10
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.
Observation edab9dcc-3591-4ec2-bc6c-ddaa94c0264e · outbound
BlueGlass: A Framework for Composite AI Safety Unresolved cited work
Reference 11
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 9afa9af6-a364-4c41-95f9-278155481ddd · outbound
BlueGlass: A Framework for Composite AI Safety E., Hume, T., Carter, S., Henighan, T., and Olah, C
Reference 12
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.
Observation e744d89e-f625-48bd-9a2d-944b37cf8872 · outbound
BlueGlass: A Framework for Composite AI Safety BatchTopK Sparse Autoencoders
Reference 13
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 10b64f59-8fbe-4986-b96d-8f7e92bc671d · outbound
BlueGlass: A Framework for Composite AI Safety Learning Multi-Level Features with Matryoshka Sparse Autoencoders
Reference 14
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation e1da9890-ef98-414d-b7c6-8811a04724ea · outbound
BlueGlass: A Framework for Composite AI Safety M., Favero, A., and Wyart, M
Reference 15
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 25553133-d791-41b7-ad7e-a8d99adab7ea · outbound
BlueGlass: A Framework for Composite AI Safety On Evaluating Adversarial Robustness
Reference 16
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation c3a84728-238a-4f7d-b9c9-74df179a3200 · outbound
BlueGlass: A Framework for Composite AI Safety DLST 4: Phase Transitions in Neural Networks
Reference 17
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.
Observation 7a4ad18b-71c6-4173-9af5-1a34a864385e · outbound
BlueGlass: A Framework for Composite AI Safety MMDetection: Open MMLab Detection Toolbox and Benchmark
Reference 18
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation c9ae5232-9ace-4ab1-b128-8b8795276059 · outbound
BlueGlass: A Framework for Composite AI Safety Generative Region-Language Pretraining for Open-Ended Object Detection
Reference 19
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.
Observation 6487b47b-7b5f-46e4-97ad-2bf01af2e79e · outbound
BlueGlass: A Framework for Composite AI Safety Sparse Autoencoders Find Highly Interpretable Features in Language Models
Reference 20
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 22291758-2325-43be-9687-36d8a4a06dc4 · outbound
BlueGlass: A Framework for Composite AI Safety BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Reference 21
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 0f6126c4-260a-4f21-b87e-b92da4ded3e1 · outbound
BlueGlass: A Framework for Composite AI Safety An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
Reference 22
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 45d1f631-5cf5-46aa-bc47-32db5aa3e311 · outbound
BlueGlass: A Framework for Composite AI Safety Transcoders Find Interpretable LLM Feature Circuits
Reference 23
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 9e7baaaa-8f6f-4733-b251-c52a42f4c79c · outbound
BlueGlass: A Framework for Composite AI Safety Garçon , December 2021
Reference 24
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.
Observation a89a8554-ad1e-4b76-9e14-76dea3809264 · outbound
BlueGlass: A Framework for Composite AI Safety NNsight and NDIF: Democratizing Access to Open-Weight Foundation Model Internals
Reference 25
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation b3604c46-0f88-49d6-91a3-b8aa86a0b994 · outbound
BlueGlass: A Framework for Composite AI Safety Scaling and evaluating sparse autoencoders
Reference 26
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation b41e325c-0cb0-4e68-9f84-d300db121f70 · outbound
BlueGlass: A Framework for Composite AI Safety Detecting Strategic Deception Using Linear Probes
Reference 27
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 9861cd60-4169-4728-bd33-757c017c30c0 · outbound
BlueGlass: A Framework for Composite AI Safety VALERIE22 -- A photorealistic, richly metadata annotated dataset of urban environments
Reference 28
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.
Observation ab7b6516-8cad-4b23-85ae-e04ac0f1cbe8 · outbound
BlueGlass: A Framework for Composite AI Safety Safety by Measurement: A Systematic Literature Review of AI Safety Evaluation Methods
Reference 29
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 3254e4a8-ed4d-46f4-a1e4-25c0826b1c07 · outbound
BlueGlass: A Framework for Composite AI Safety LVIS: A Dataset for Large Vocabulary Instance Segmentation
Reference 30
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation c110fa2f-ff69-41fb-a658-47720fa33ba2 · outbound
BlueGlass: A Framework for Composite AI Safety How to use and interpret activation patching
Reference 31
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 4eaa86aa-c39d-49fc-9911-1f448e9bde96 · outbound
BlueGlass: A Framework for Composite AI Safety FunnyBirds: A Synthetic Vision Dataset for a Part-Based Analysis of Explainable AI Methods
Reference 32
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.
Observation e996239e-be49-4477-ae13-bd1613d439b5 · outbound
BlueGlass: A Framework for Composite AI Safety A Comprehensive Survey on Applications of Transformers for Deep Learning Tasks
Reference 33
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.
Observation d0be80d8-2336-43d6-ab5f-4d1182d0b754 · outbound
BlueGlass: A Framework for Composite AI Safety Ultralytics YOLO , January 2023
Reference 34
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.
Observation 0afec4d0-01ee-444b-bdd3-993b6fcbf964 · outbound
BlueGlass: A Framework for Composite AI Safety Are Sparse Autoencoders Useful? A Case Study in Sparse Probing
Reference 35
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation acd04884-4e1a-4a82-ac33-d96a34da1479 · outbound
BlueGlass: A Framework for Composite AI Safety The Open Images Dataset V4: Unified image classification, object detection, and visual relationship detection at scale
Reference 36
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.
Observation 080aa468-8507-4b6b-ba6c-8d6d68a747ef · outbound
BlueGlass: A Framework for Composite AI Safety NV-Embed: Improved Techniques for Training LLMs as Generalist Embedding Models
Reference 37
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation ec3abb6f-c646-4fa8-8b54-1d0af3e8622f · outbound
BlueGlass: A Framework for Composite AI Safety Beyond the Doors of Perception: Vision Transformers Represent Relations Between Objects
Reference 38
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 476a2bf4-6d87-48bb-b1f7-ca4b68b8c97c · outbound
BlueGlass: A Framework for Composite AI Safety Open World Object Detection: A Survey
Reference 39
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.
Observation 9ce1f919-f93d-413d-b860-570ddff36ca1 · outbound
BlueGlass: A Framework for Composite AI Safety Microsoft COCO: Common Objects in Context
Reference 40
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation e21b1cfe-bf83-45d4-89d7-eb5f4d6cc0f4 · outbound
BlueGlass: A Framework for Composite AI Safety Sparse Crosscoders for Cross-Layer Features and Model Diffing
Reference 41
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.
Observation 0c4d431f-f983-4e24-a2a2-08cd898794bb · outbound
BlueGlass: A Framework for Composite AI Safety Unresolved cited work
Reference 42
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.
Observation d8ae6d68-e42b-4faf-98c6-c5f460355fc5 · outbound
BlueGlass: A Framework for Composite AI Safety Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection
Reference 43
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 80c6e2a7-9ab3-4949-9148-078c97e4bee3 · outbound
BlueGlass: A Framework for Composite AI Safety Data Attribution: A Data-Centric Approach for Trustworthy AI Development
Reference 44
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.
Observation 59e4ae61-2b3e-4697-9c56-a0f73d90f512 · outbound
BlueGlass: A Framework for Composite AI Safety A Survey on Vision-Language-Action Models for Embodied AI
Reference 45
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 97c99df3-41c9-419b-a568-d9fde04a3301 · outbound
BlueGlass: A Framework for Composite AI Safety k-Sparse Autoencoders
Reference 46
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 5955998a-17cd-465f-b711-6f1740694cd4 · outbound
BlueGlass: A Framework for Composite AI Safety Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models
Reference 47
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation c3bdf3bf-2306-4363-8e0f-10bb4bcc79f5 · outbound
BlueGlass: A Framework for Composite AI Safety Deep Double Descent: Where Bigger Models and More Data Hurt
Reference 48
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 7293ed22-f9c4-481d-bd3d-47f0abe7e658 · outbound
BlueGlass: A Framework for Composite AI Safety Interpretability will not reliably find deceptive ai
Reference 49
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.
Observation 07e4e179-8f27-4798-98d8-746186b6ce4c · outbound
BlueGlass: A Framework for Composite AI Safety and Bloom, J
Reference 50
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.
Observation f20aa03e-5b6c-4a2e-aa14-417471083741 · outbound
BlueGlass: A Framework for Composite AI Safety Understanding Neural Networks via Feature Visualization: A survey
Reference 51
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.
Observation fd6289ab-5d40-43e8-b39c-7c132a9b7ffe · outbound
BlueGlass: A Framework for Composite AI Safety In-context Learning and Induction Heads
Reference 52
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.
Observation 43551005-0b9f-4ef7-a88f-6dbacf340caa · outbound
BlueGlass: A Framework for Composite AI Safety GPT-4o System Card
Reference 53
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 5ba300c2-cc56-4d1b-95c0-612a9bf8be97 · outbound
BlueGlass: A Framework for Composite AI Safety Mamba-360: Survey of State Space Models as Transformer Alternative for Long Sequence Modelling: Methods, Applications, and Challenges
Reference 54
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 7ebb9618-cde8-44d3-bfe6-58c6473c59c9 · outbound
BlueGlass: A Framework for Composite AI Safety Introducing Gemini 2.0: our new AI model for the agentic era , December 2024
Reference 55
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.
Observation 9951d616-7bf0-416a-a1ca-f0d7631b94d6 · outbound
BlueGlass: A Framework for Composite AI Safety On Evaluating the Durability of Safeguards for Open-Weight LLMs
Reference 56
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 2491fe7e-716e-48d5-ac1a-8d339f982919 · outbound
BlueGlass: A Framework for Composite AI Safety Language Models are Unsupervised Multitask Learners , 2019
Reference 57
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.
Observation b6a4309e-db90-4475-a23b-568dc5649260 · outbound
BlueGlass: A Framework for Composite AI Safety Learning Transferable Visual Models From Natural Language Supervision
Reference 58
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation dfffdb04-f8d6-4c92-b3d4-626fa75efab4 · outbound
BlueGlass: A Framework for Composite AI Safety Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
Reference 59
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 313eb60f-c70a-4a58-bcee-97b520ee1469 · outbound
BlueGlass: A Framework for Composite AI Safety A Phase Transition in Diffusion Models Reveals the Hierarchical Nature of Data
Reference 60
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation a8aef35c-367d-42d3-b983-eabc606ef55c · outbound
BlueGlass: A Framework for Composite AI Safety b1ade series of models , 2024
Reference 61
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.
Observation 857dab4a-4f74-4e98-97d2-0855672292b1 · outbound
BlueGlass: A Framework for Composite AI Safety The 'strong' feature hypothesis could be wrong
Reference 62
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.
Observation 2faf6f59-3585-4f9e-a665-1375f64d4dca · outbound
BlueGlass: A Framework for Composite AI Safety Negative Results for SAEs On Downstream Tasks and Deprioritising SAE Research (GDM Mech Interp Team Progress Update \#2)
Reference 63
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.
Observation 8a3da632-3e0b-442b-85e2-2ef2c99e7e16 · outbound
BlueGlass: A Framework for Composite AI Safety PaliGemma 2: A Family of Versatile VLMs for Transfer
Reference 64
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation a2f6727a-56a5-41a2-a1e0-0e65ba8efac6 · outbound
BlueGlass: A Framework for Composite AI Safety Attribution Patching Outperforms Automated Circuit Discovery
Reference 65
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 4f581f6a-f19e-4337-bba0-09f90abf4149 · outbound
BlueGlass: A Framework for Composite AI Safety Gemini Robotics: Bringing AI into the Physical World
Reference 66
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 374a1aba-2c34-48d2-95ef-6ba1a555a2d8 · outbound
BlueGlass: A Framework for Composite AI Safety L., McDougall, C., MacDiarmid, M., Freeman, C
Reference 67
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.
Observation 0ed0daa9-1d69-47ee-b766-db511d0f2bec · outbound
BlueGlass: A Framework for Composite AI Safety Deep Learning and the Information Bottleneck Principle
Reference 68
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 7ffc898e-b009-4868-9660-4c6490e37a57 · outbound
BlueGlass: A Framework for Composite AI Safety HuggingFace's Transformers: State-of-the-art Natural Language Processing
Reference 69
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation ffd8580b-fa19-40ae-9f32-7c3ce4f756fe · outbound
Reference 70
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.
Observation c56b6820-0adb-4373-9b3d-0f16f9b40ddc · outbound
BlueGlass: A Framework for Composite AI Safety Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks
Reference 71
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.
Observation 86ec7568-ca42-4508-b65a-5b75bd54ac0f · outbound
BlueGlass: A Framework for Composite AI Safety Diffusion models: A comprehensive survey of methods and applications, 2024
Reference 72
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 11e7ef1a-1583-4170-9e1c-6f1df756af55 · outbound
BlueGlass: A Framework for Composite AI Safety BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning
Reference 73
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation c011ffda-9750-4188-9c2a-c87a58e49476 · outbound
BlueGlass: A Framework for Composite AI Safety DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection
Reference 74
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 3db49778-a5be-4ef0-a1c8-f7351c95ef90 · outbound
BlueGlass: A Framework for Composite AI Safety Vision-Language Models for Vision Tasks: A Survey
Reference 75
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 6bc78453-d434-4a79-9b95-1fc9f3fee28c · outbound
BlueGlass: A Framework for Composite AI Safety Vision Language Models in Autonomous Driving: A Survey and Outlook
Reference 76
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 6084b07b-a3bc-4de6-91d2-ce5533769d9d · outbound
BlueGlass: A Framework for Composite AI Safety Object Detection in 20 Years: A Survey
Reference 77
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
Observation 306e0605-2734-4201-9e22-f5f69268bb97 · outbound
BlueGlass: A Framework for Composite AI Safety Grokking phase transitions in learning local rules with gradient descent
Reference 78
Source-reported events for the cited work
Unavailable: canonical work link unavailable.
No inbound Pith citation observations are available.