Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense | Paper Citation Record

Observation e2140333-e41d-4262-babf-344189700f7e · outbound

This paper cites Chat generative pre-trained transformer (chat- gpt).

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Chat generative pre-trained transformer (chat- gpt)

Reference 1

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.445405Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 0cea145c-5b3c-45de-a31e-bd36475a504c · outbound

This paper cites Llama 2: Open Foundation and Fine-Tuned Chat Models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Llama 2: Open Foundation and Fine-Tuned Chat Models

Reference 2

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.449262Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5ba827a8-8fac-4239-962e-599a67040617 · outbound

This paper cites Gemma: Open Models Based on Gemini Research and Technology.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Gemma: Open Models Based on Gemini Research and Technology

Reference 3

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.452541Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 74610e66-12fc-4343-8437-0fcfe916ef52 · outbound

This paper cites Mistral 7B.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Mistral 7B

Reference 4

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.455573Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 01e4ab5a-08e7-4396-9012-3dfd8245a74f · outbound

This paper cites The Falcon Series of Open Language Models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense The Falcon Series of Open Language Models

Reference 5

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.458365Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 64401fd9-c23c-4d59-aa32-5ca14ea0d8f3 · outbound

This paper cites Qwen Technical Report.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Qwen Technical Report

Reference 6

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.461305Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d0ec0fee-50b6-4ebe-92af-222ca99d8f52 · outbound

This paper cites Gaussian Error Linear Units (GELUs).

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Gaussian Error Linear Units (GELUs)

Reference 7

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.464411Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c704735d-6f10-4dcb-b2a8-67c736845040 · outbound

This paper cites Sigmoid- weighted linear units for neural network function ap- proximation in reinforcement learning.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Sigmoid- weighted linear units for neural network function ap- proximation in reinforcement learning

Reference 8

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.467467Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a6aaae0c-d6cd-4a15-bb84-41a7bbc6a0cb · outbound

This paper cites Awq: Activation- aware weight quantization for on-device llm compres- sion and acceleration.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Awq: Activation- aware weight quantization for on-device llm compres- sion and acceleration

Reference 9

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.470224Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 35836eb3-c6d7-437f-8d81-84cfea072082 · outbound

This paper cites GPTQ: Accurate post-training compression for generative pretrained transformers.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense GPTQ: Accurate post-training compression for generative pretrained transformers

Reference 10

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.472818Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 479f9a15-8326-4792-8bfd-ec4a0f8371c4 · outbound

This paper cites Billm: Pushing the limit of post-training quan- tization for llms.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Billm: Pushing the limit of post-training quan- tization for llms

Reference 11

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.475289Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 651d0dfb-4ab2-45ee-9c83-d4503e1b4c98 · outbound

This paper cites Sparsegpt: Massive lan- guage models can be accurately pruned in one-shot.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Sparsegpt: Massive lan- guage models can be accurately pruned in one-shot

Reference 12

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.477774Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2bbb1169-9b05-45db-bc95-7c1ca97e02af · outbound

This paper cites Llm- pruner: On the structural pruning of large language mod- els.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Llm- pruner: On the structural pruning of large language mod- els

Reference 13

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.480322Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d7a430ef-fed4-41b6-b1ba-6de630ba7c3e · outbound

This paper cites Sheared LLaMA: Accelerating Language Model Pre-training via Structured Pruning.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Sheared LLaMA: Accelerating Language Model Pre-training via Structured Pruning

Reference 14

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.482768Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation faed12e6-880d-4a75-9f5a-8bbbb02903de · outbound

This paper cites Structured Pruning of Large Language Models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Structured Pruning of Large Language Models

Reference 15

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.485481Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 058c99be-1b39-4bf5-9a2b-a218ac461f99 · outbound

This paper cites Fast inference from transformers via speculative decoding.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Fast inference from transformers via speculative decoding

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.488230Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 68d3aff6-70f9-42cc-b2b5-c48b44aca84b · outbound

This paper cites Speculative decoding with big little de- coder.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Speculative decoding with big little de- coder

Reference 17

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.490567Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation bdda5e25-d4fe-4fb3-92b7-8fe140a74b12 · outbound

This paper cites Iron: Private infer- ence on transformers.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Iron: Private infer- ence on transformers

Reference 18

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.492867Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a5dafbbe-a6ca-40c3-a41c-8fb73b586f97 · outbound

This paper cites Ciphergpt: Secure two- party gpt inference.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Ciphergpt: Secure two- party gpt inference

Reference 19

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.494884Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation be552ba8-a521-4ba9-9f1b-57b70edcef59 · outbound

This paper cites Bolt: Privacy-preserving, accu- rate and efficient inference for transformers.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Bolt: Privacy-preserving, accu- rate and efficient inference for transformers

Reference 20

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.496815Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 19367cde-fe3b-4f0e-b017-bbd0653347c2 · outbound

This paper cites BumbleBee: Secure Two-party Inference Frame- work for Large Transformers.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense BumbleBee: Secure Two-party Inference Frame- work for Large Transformers

Reference 21

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.208924Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 1b0a96f9-7cf6-43b7-8d4a-bccdc494eec1 · outbound

This paper cites Secure transformer infer- ence made non-interactive.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Secure transformer infer- ence made non-interactive

Reference 22

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.194371Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 4e6321c5-dda7-4236-96aa-7c40970326c2 · outbound

This paper cites Training-Free Activation Sparsity in Large Language Models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Training-Free Activation Sparsity in Large Language Models

Reference 23

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.505073Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 287bf1e2-4352-430b-ac76-5ab02d64ee48 · outbound

This paper cites ReLU$^2$ Wins: Discovering Efficient Activation Functions for Sparse LLMs.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense ReLU$^2$ Wins: Discovering Efficient Activation Functions for Sparse LLMs

Reference 24

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.507833Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation dcc82b43-ba2b-4f9a-b7d9-6cbb855b197e · outbound

This paper cites Relu strikes back: Exploiting activation sparsity in large lan- guage models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Relu strikes back: Exploiting activation sparsity in large lan- guage models

Reference 25

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.187461Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 288148a7-7b86-42e1-af4f-39648d09b0d8 · outbound

This paper cites Smoothquant: Accu- rate and efficient post-training quantization for large language models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Smoothquant: Accu- rate and efficient post-training quantization for large language models

Reference 26

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.180020Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation a1d2126a-2aa5-4d7a-869f-5e4a87678d1a · outbound

This paper cites Omniquant: Omnidirection- ally calibrated quantization for large language models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Omniquant: Omnidirection- ally calibrated quantization for large language models

Reference 27

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.172593Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 8c840334-b818-4007-91e1-3b4854e5963b · outbound

This paper cites Sirnn: A math library for secure rnn inference.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Sirnn: A math library for secure rnn inference

Reference 28

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.165223Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 4f1e32b9-8c20-42b2-84ef-e4f526db4df2 · outbound

This paper cites From chatgpt to threat- gpt: Impact of generative ai in cybersecurity and privacy.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense From chatgpt to threat- gpt: Impact of generative ai in cybersecurity and privacy

Reference 29

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.157872Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 637b992f-b3f1-45cf-ba58-87c2879cdb14 · outbound

This paper cites Exploiting large language models (llms) through decep- tion techniques and persuasion principles.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Exploiting large language models (llms) through decep- tion techniques and persuasion principles

Reference 30

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.150589Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 4e4e6dbc-4e9c-4b45-afbe-ce3af4e2d85b · outbound

This paper cites PsySafe: A Comprehensive Framework for Psychological-based Attack, Defense, and Evaluation of Multi-agent System Safety.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense PsySafe: A Comprehensive Framework for Psychological-based Attack, Defense, and Evaluation of Multi-agent System Safety

Reference 31

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.525244Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1ef5ecec-76c2-49c2-9151-df6e15127c28 · outbound

This paper cites Training language models to follow instructions with human feedback.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Training language models to follow instructions with human feedback

Reference 32

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.527855Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c76fa3c1-23e2-4af4-95c2-8bb3a4510ec2 · outbound

This paper cites Safe RLHF: Safe Reinforcement Learning from Human Feedback.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Safe RLHF: Safe Reinforcement Learning from Human Feedback

Reference 33

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.530426Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 990fe8fb-ecd8-4a21-a561-8001f009b155 · outbound

This paper cites Di- rect preference optimization: Your language model is secretly a reward model.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Di- rect preference optimization: Your language model is secretly a reward model

Reference 34

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.139508Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 28292dea-c199-4d8d-b0c6-068dbfc45920 · outbound

This paper cites Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!

Reference 35

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.535809Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 21dcb9ae-d6d4-444f-96b0-4f402b7abfbe · outbound

This paper cites Safe LoRA: the Silver Lining of Reducing Safety Risks when Fine-tuning Large Language Models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Safe LoRA: the Silver Lining of Reducing Safety Risks when Fine-tuning Large Language Models

Reference 36

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.538531Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 648e5c71-b1cd-46a7-b4ce-4f903832b0df · outbound

This paper cites GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers

Reference 37

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.541244Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 554ca386-87ae-4910-a4bb-96a9a97df6b0 · outbound

This paper cites A simple and effective pruning approach for large lan- guage models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense A simple and effective pruning approach for large lan- guage models

Reference 38

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.132421Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation eb45a763-0ab2-44c5-be77-10d227b55ad4 · outbound

This paper cites Dynamic context pruning for efficient and interpretable autoregressive transformers.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Dynamic context pruning for efficient and interpretable autoregressive transformers

Reference 39

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.125481Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 7ce3b523-7318-4acf-8037-686931756f86 · outbound

This paper cites Adapting language models to compress contexts.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Adapting language models to compress contexts

Reference 40

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.118461Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation c805132b-81c1-4aa9-86dc-05ae00074791 · outbound

This paper cites {InfiniGen}: Efficient generative inference of large language models with dynamic {KV} cache manage- ment.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense {InfiniGen}: Efficient generative inference of large language models with dynamic {KV} cache manage- ment

Reference 41

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.111328Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 71451ffd-5c83-4a1e-a185-f5420ad3677c · outbound

This paper cites {IMPRESS}: An {Importance- Informed}{Multi-Tier} prefix {KV} storage system for large language model inference.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense {IMPRESS}: An {Importance- Informed}{Multi-Tier} prefix {KV} storage system for large language model inference

Reference 42

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.104171Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 3a1c35c4-aa16-4794-9217-f105d484939a · outbound

This paper cites Scissorhands: Exploiting the persistence of importance hypothesis for llm kv cache compression at test time.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Scissorhands: Exploiting the persistence of importance hypothesis for llm kv cache compression at test time

Reference 43

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.098118Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation e7788879-bce4-4f23-b5e5-cc32f49dd2cb · outbound

This paper cites Kvquant: Towards 10 million con- text length llm inference with kv cache quantization.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Kvquant: Towards 10 million con- text length llm inference with kv cache quantization

Reference 44

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.091674Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 10aad630-7787-44c9-855c-ac005805c07b · outbound

This paper cites Sequoia: Scalable and robust speculative de- coding.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Sequoia: Scalable and robust speculative de- coding

Reference 45

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.085231Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 48c61911-7bba-476a-bea3-44a973b2df16 · outbound

This paper cites Decoding compressed trust: Scrutinizing the trustworthiness of efficient llms under compression.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Decoding compressed trust: Scrutinizing the trustworthiness of efficient llms under compression

Reference 46

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.078252Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 45a28834-227a-4d5b-8201-4e64193306de · outbound

This paper cites Harmlevelbench: Evaluating harm-level compliance and the impact of quantization on model alignment.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Harmlevelbench: Evaluating harm-level compliance and the impact of quantization on model alignment

Reference 47

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.070834Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 5b8a9931-e89a-415e-892e-742fb4724cb2 · outbound

This paper cites Exploiting llm quantization.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Exploiting llm quantization

Reference 48

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.063577Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation ceffec12-a271-4d19-b0cb-f420f35f503f · outbound

This paper cites Zeroquant: Efficient and affordable post-training quantization for large-scale transformers.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Zeroquant: Efficient and affordable post-training quantization for large-scale transformers

Reference 49

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.056232Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 58b43f76-c1a2-4ce1-b0e4-34a321323825 · outbound

This paper cites Qllm: Accurate and efficient low-bitwidth quantization for large language models.In- ternational Conference on Machine Learning, 2024.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Qllm: Accurate and efficient low-bitwidth quantization for large language models.In- ternational Conference on Machine Learning, 2024

Reference 50

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.048307Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 4b9e6550-0094-4b81-ad21-6164bfb5beee · outbound

This paper cites Introducing meta llama 3: The most capable openly available llm to date.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Introducing meta llama 3: The most capable openly available llm to date

Reference 51

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.575830Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 036314c3-df4a-428e-85ef-917363701540 · outbound

This paper cites Mpcformer: fast, performant and private transformer inference with mpc.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Mpcformer: fast, performant and private transformer inference with mpc

Reference 52

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.037095Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation bdb1cc3d-e089-4a1e-9e20-bf96c91f299b · outbound

This paper cites Encryption-Friendly LLM Architecture.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Encryption-Friendly LLM Architecture

Reference 53

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.579915Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9f2ac483-69c7-42ab-bbfb-31c9b162cb94 · outbound

This paper cites Deja vu: Con- textual sparsity for efficient llms at inference time.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Deja vu: Con- textual sparsity for efficient llms at inference time

Reference 54

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.029642Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 6da7ef52-b474-4671-86df-2cb90b6c996c · outbound

This paper cites Cats: Context-aware thresh- olding for sparsity in large language models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Cats: Context-aware thresh- olding for sparsity in large language models

Reference 55

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.022246Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation d9906681-c789-4184-9a25-d9900b8443bc · outbound

This paper cites Judging llm-as- a-judge with mt-bench and chatbot arena.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Judging llm-as- a-judge with mt-bench and chatbot arena

Reference 56

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.586311Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 04a0dd48-3eb6-413b-94ef-63469149a2a4 · outbound

This paper cites Assessing the brittleness of safety alignment via pruning and low-rank modifica- tions.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Assessing the brittleness of safety alignment via pruning and low-rank modifica- tions

Reference 57

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.010797Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation f9eaa58c-f081-48d1-9de3-5221b7b1162b · outbound

This paper cites Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey

Reference 58

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.590442Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b32f45a2-16e6-49de-8886-851d2c32f982 · outbound

This paper cites Lisa: Lazy safety alignment for large language models against harmful fine-tuning at- tack.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Lisa: Lazy safety alignment for large language models against harmful fine-tuning at- tack

Reference 59

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:40.003409Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation a5fb86ba-5cd5-4115-adb7-184980feab14 · outbound

This paper cites Universal and Transferable Adversarial Attacks on Aligned Language Models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Universal and Transferable Adversarial Attacks on Aligned Language Models

Reference 60

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.595425Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a2012904-6ca2-4b52-8886-23136f88c45a · outbound

This paper cites GPTFUZZER: Red Teaming Large Language Models with Auto-Generated Jailbreak Prompts.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense GPTFUZZER: Red Teaming Large Language Models with Auto-Generated Jailbreak Prompts

Reference 61

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.598235Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation e98b3c49-8494-4c98-9d5a-0b2b8460beac · outbound

This paper cites Detoxifying Large Language Models via Knowledge Editing.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Detoxifying Large Language Models via Knowledge Editing

Reference 62

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.600920Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 9ed5edc6-4186-4910-8e00-71df8537532c · outbound

This paper cites Recovering the Pre-Fine-Tuning Weights of Generative Models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Recovering the Pre-Fine-Tuning Weights of Generative Models

Reference 63

Resolution: verified exact; local_arxiv, observed 2026-08-09T17:37:39.818906Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation dcecaae4-378e-4c13-a3a7-3cc5ef393661 · outbound

This paper cites Self-Distillation Bridges Distribution Gap in Language Model Fine-Tuning.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Self-Distillation Bridges Distribution Gap in Language Model Fine-Tuning

Reference 64

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.606386Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 0d6d80ec-20af-4486-bbc5-73332f878ca0 · outbound

This paper cites Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Keeping LLMs Aligned After Fine-tuning: The Crucial Role of Prompt Templates

Reference 65

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.609496Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c1a338de-b2c4-4f79-8bcc-dacded386d85 · outbound

This paper cites SEAL: Safety-enhanced Aligned LLM Fine-tuning via Bilevel Data Selection.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense SEAL: Safety-enhanced Aligned LLM Fine-tuning via Bilevel Data Selection

Reference 66

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.612801Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6a15da23-7dd0-4138-8b6c-b2475b3ffc78 · outbound

This paper cites Compressing LLMs: The Truth is Rarely Pure and Never Simple.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Compressing LLMs: The Truth is Rarely Pure and Never Simple

Reference 67

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.615850Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 960c1af9-ac41-4784-877c-0b3489ae88f3 · outbound

This paper cites Outlier Weighed Layerwise Sparsity (OWL): A Missing Secret Sauce for Pruning LLMs to High Sparsity.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Outlier Weighed Layerwise Sparsity (OWL): A Missing Secret Sauce for Pruning LLMs to High Sparsity

Reference 68

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.618406Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d45dd858-0691-4600-b9dd-4eca0d8c4165 · outbound

This paper cites Discovering sparsity allo- cation for layer-wise pruning of large language models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Discovering sparsity allo- cation for layer-wise pruning of large language models

Reference 69

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:39.996108Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 63d92d98-cf18-45a9-a45d-6a71f0bcdb09 · outbound

This paper cites Beyond Perplexity: Multi-dimensional Safety Evaluation of LLM Compression.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Beyond Perplexity: Multi-dimensional Safety Evaluation of LLM Compression

Reference 70

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.623449Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a13e492f-863c-42fb-816e-a0528327c55a · outbound

This paper cites Navigating Extremes: Dynamic Sparsity in Large Output Spaces.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Navigating Extremes: Dynamic Sparsity in Large Output Spaces

Reference 71

Resolution: verified exact; local_arxiv, observed 2026-08-09T17:37:39.772417Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 840f2a7a-b166-4dbd-8102-4b404d28ec5f · outbound

This paper cites Catastrophic Jailbreak of Open-source LLMs via Exploiting Generation.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Catastrophic Jailbreak of Open-source LLMs via Exploiting Generation

Reference 72

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.628959Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 67235692-f973-4818-a960-1f910571583d · outbound

This paper cites TrustLLM: Trustworthiness in Large Language Models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense TrustLLM: Trustworthiness in Large Language Models

Reference 73

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.631778Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation aa89ebd6-d450-44d1-9013-7aa020b7f05f · outbound

This paper cites HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal

Reference 74

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.634473Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 778cdd0d-f1e7-4867-b52e-1cc7557956df · outbound

This paper cites Autodan: Generating stealthy jailbreak prompts on aligned large language models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Autodan: Generating stealthy jailbreak prompts on aligned large language models

Reference 75

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:39.988585Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 7cdadd0f-7f26-40e0-83d7-617cc7654d8e · outbound

This paper cites Automatically auditing large language mod- els via discrete optimization.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Automatically auditing large language mod- els via discrete optimization

Reference 76

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:39.981980Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 894c6b5b-92f3-4d3d-9ee8-b8e810b3b87f · outbound

This paper cites Making them ask and an- swer: Jailbreaking large language models in few queries via disguise and reconstruction.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Making them ask and an- swer: Jailbreaking large language models in few queries via disguise and reconstruction

Reference 77

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:39.975523Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 5df229dc-92c6-4790-b0ac-a7dcdadbf21b · outbound

This paper cites Beavertails: Towards im- proved safety alignment of llm via a human-preference dataset.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Beavertails: Towards im- proved safety alignment of llm via a human-preference dataset

Reference 78

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:39.968993Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation 334cc4f1-54fd-45de-949c-8bb81b5e6779 · outbound

This paper cites Pointer Sentinel Mixture Models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Pointer Sentinel Mixture Models

Reference 79

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.645158Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 8e3347ca-54f3-4957-8d95-4b6af5c0d62e · outbound

This paper cites Measuring Massive Multitask Language Understanding.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Measuring Massive Multitask Language Understanding

Reference 80

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.647318Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 00071c23-c419-4149-b7ba-be718d7fb249 · outbound

This paper cites Qlora: Efficient finetuning of quan- tized llms.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Qlora: Efficient finetuning of quan- tized llms

Reference 81

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:39.961340Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation eb141789-1fd9-4c8f-895b-2aebe0be15ac · outbound

This paper cites Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone

Reference 82

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.651866Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ccb460b1-b160-421d-a89a-934d83896715 · outbound

This paper cites Mixtral of Experts.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Mixtral of Experts

Reference 83

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.654525Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5bd9bd37-2bfb-4474-83a5-fff8d3c9e48c · outbound

This paper cites Zephyr: Direct Distillation of LM Alignment.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Zephyr: Direct Distillation of LM Alignment

Reference 84

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.657148Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation faabd1bf-74b9-4db4-8168-81affad895e7 · outbound

This paper cites Qwen2.5 Technical Report.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Qwen2.5 Technical Report

Reference 85

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.660075Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5a5aa9ea-1b8c-4af4-8fb0-b5de3faa4aee · outbound

This paper cites Hashimoto.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Hashimoto

Reference 86

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.662669Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 8efe3443-9481-48b6-a96d-ff673f06b509 · outbound

This paper cites Language models are few-shot learn- ers.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Language models are few-shot learn- ers

Reference 87

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.665104Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4a7600c1-c273-4d04-9415-e4c8c40dab33 · outbound

This paper cites Exploring the limits of transfer learning with a unified text-to-text transformer.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Exploring the limits of transfer learning with a unified text-to-text transformer

Reference 88

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.667643Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 553ca044-2223-4d81-820c-df4a5472aaea · outbound

This paper cites Shaping the Safety Boundaries: Understanding and Defending Against Jailbreaks in Large Language Models.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Shaping the Safety Boundaries: Understanding and Defending Against Jailbreaks in Large Language Models

Reference 89

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.670171Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 8404c6c5-62b5-44a5-b9bd-9a2607544e79 · outbound

This paper cites Towards Deep Learning Models Resistant to Adversarial Attacks.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Towards Deep Learning Models Resistant to Adversarial Attacks

Reference 90

Resolution: unresolved; no resolver link, observed 2026-08-09T17:37:39.672901Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5f45c01d-dfe1-44ed-96e7-a634fb74bb6f · outbound

This paper cites Stanford alpaca: An instruction- following llama model, 2023.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Stanford alpaca: An instruction- following llama model, 2023

Reference 91

Resolution: verified fuzzy; raw_fallback, observed 2026-08-09T17:37:39.942588Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Observation fa618968-8898-4dca-a5ae-b76e4f6f171f · outbound

This paper cites an unresolved cited work.

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense Unresolved cited work

Reference 2025

Resolution: unresolved; raw_fallback, observed 2026-08-09T17:37:40.201564Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-23T06:30:58.430688+00:00.

Coverage vector

Reference resolution

Outbound references

Pith citing papers