Paper Citation Record · LEDGER

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs

As of 6 August 2026, this Paper Citation Record lists 33 of 33 outbound references and 1 inbound Pith citation observation for arXiv:2505.14226.

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

pith.paper-citation-record.v1
2505.14226 v5

Coverage vector

measured 33 of 33 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-05-22T14:40:58.506345Z

measured 34 of 34 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-06T06:34:29.942622+00:00

measured 1 of 1 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links, observed 2026-08-05T17:26:48.605444Z

measured 0 of 1 external citation measurements

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

Source: pith, observed 2026-08-05T17:26:48.888071Z

Reference resolution

33 of 33 outbound references displayed

verified exact22
verified fuzzy7
unresolved0
parse uncertain0
malformed identifier2
metadata mismatch2

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation f374af8e-f008-408c-848d-02719472a3b5 · outbound

This paper cites Detecting Language Model Attacks with Perplexity.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Detecting Language Model Attacks with Perplexity

Reference 1

Resolution: verified exact; local_arxiv, observed 2026-05-22T14:41:41.493900Z

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 0f5bf002-82ea-4324-9e6a-947cd372fee6 · outbound

This paper cites How (un)ethical are instruction-centric responses of LLMs? Unveiling the vulnerabilities of safety guardrails to harmful queries.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs How (un)ethical are instruction-centric responses of LLMs? Unveiling the vulnerabilities of safety guardrails to harmful queries

Reference 2

Resolution: verified exact; arxiv_id, observed 2026-05-22T14:41:41.380696Z

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 746dd16e-0164-4426-a7e2-04eb6de38971 · outbound

This paper cites Attributional safety failures in large language models under code-mixed perturbations.arXiv preprint arXiv:2505.14469.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Attributional safety failures in large language models under code-mixed perturbations.arXiv preprint arXiv:2505.14469

Reference 3

Resolution: verified exact; arxiv_id, observed 2026-05-22T14:41:41.405130Z

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 6c695795-7c4c-4210-9327-852985dffe04 · outbound

This paper cites Red-Teaming Large Language Models using Chain of Utterances for Safety-Alignment.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Red-Teaming Large Language Models using Chain of Utterances for Safety-Alignment

Reference 4

Resolution: metadata mismatch; arxiv_id, observed 2026-05-22T14:41:41.440784Z

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 404f7030-5414-4a63-83f2-6b4fb04c2418 · outbound

This paper cites Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities

Reference 5

Resolution: verified exact; local_arxiv, observed 2026-05-22T14:41:41.397593Z

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 1f3353ec-1df4-480d-976f-5f1e0d62cb10 · outbound

This paper cites No Language Left Behind: Scaling Human-Centered Machine Translation.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs No Language Left Behind: Scaling Human-Centered Machine Translation

Reference 6

Resolution: verified exact; local_arxiv, observed 2026-05-22T14:41:41.386299Z

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 5f6dc16e-2b34-42cb-96e9-28ffedfb4c10 · outbound

This paper cites The Llama 3 Herd of Models.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs The Llama 3 Herd of Models

Reference 7

Resolution: verified exact; local_arxiv, observed 2026-05-22T14:41:41.458609Z

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 49f85f56-2bda-4016-a54b-34ac06079838 · outbound

This paper cites Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned

Reference 8

Resolution: verified exact; local_arxiv, observed 2026-05-22T14:41:41.416562Z

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 b07e6f8e-1018-4716-9f88-ef299770f5cc · outbound

This paper cites Prohibited Use Policy.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Prohibited Use Policy

Reference 9

Resolution: verified fuzzy; raw_fallback, observed 2026-05-22T14:41:41.619183Z

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 abe4c1ac-d1a2-4bd6-9bdf-5fbf22b30389 · outbound

This paper cites Sowing the wind, reaping the whirlwind: The impact of editing language models.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Sowing the wind, reaping the whirlwind: The impact of editing language models

Reference 10

Resolution: verified fuzzy; raw_fallback, observed 2026-05-22T14:41:41.608421Z

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 09584996-99a3-415f-854d-013c12121e67 · outbound

This paper cites doi: 10.18653/v1/2024.findings-acl.960.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs doi: 10.18653/v1/2024.findings-acl.960

Reference 11

Resolution: verified exact; doi, observed 2026-05-22T14:41:41.112521Z

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 632cceae-2e1c-439b-bab5-ece80c5f146b · outbound

This paper cites Trustagent: Towards safe and trustworthy llm-based agents.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Trustagent: Towards safe and trustworthy llm-based agents

Reference 12

Resolution: verified fuzzy; raw_fallback, observed 2026-05-22T14:41:41.634181Z

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 04786a40-3de7-4e1c-9843-f758d418ff67 · outbound

This paper cites Best-of-N Jailbreaking.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Best-of-N Jailbreaking

Reference 13

Resolution: verified exact; arxiv_id, observed 2026-05-22T14:41:41.482545Z

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 85202759-6023-46cc-98f8-d63c568b297e · outbound

This paper cites GPT-4o System Card.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs GPT-4o System Card

Reference 14

Resolution: metadata mismatch; local_arxiv, observed 2026-05-22T14:41:41.499469Z

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 d21bf16b-7e55-4fa7-b830-04ab17a5f626 · outbound

This paper cites Mistral 7B.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Mistral 7B

Reference 15

Resolution: verified exact; local_arxiv, observed 2026-05-22T14:41:41.375092Z

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 73027726-a660-46b7-b1e9-9f6d9b945924 · outbound

This paper cites Faster-GCG: Efficient Discrete Optimization Jailbreak Attacks against Aligned Large Language Models.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Faster-GCG: Efficient Discrete Optimization Jailbreak Attacks against Aligned Large Language Models

Reference 16

Resolution: verified exact; local_arxiv, observed 2026-05-22T14:41:41.464281Z

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 63485edb-3373-4ffe-9534-3dde6df33a73 · outbound

This paper cites Prompt Injection Attacks and Defenses in LLM-Integrated Applications.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Prompt Injection Attacks and Defenses in LLM-Integrated Applications

Reference 17

Resolution: verified exact; arxiv_id, observed 2026-05-22T14:41:41.488361Z

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 cfe2e55e-b45b-4ae2-8691-082b65d443db · outbound

This paper cites Towards Red Teaming in Multimodal and Multilingual Translation.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Towards Red Teaming in Multimodal and Multilingual Translation

Reference 18

Resolution: verified exact; arxiv_id, observed 2026-05-22T14:41:41.470340Z

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 a602b94d-a15e-42a1-8d58-49478f4c4b12 · outbound

This paper cites Multilingual Blending: LLM Safety Alignment Evaluation with Language Mixture.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Multilingual Blending: LLM Safety Alignment Evaluation with Language Mixture

Reference 19

Resolution: verified exact; arxiv_id, observed 2026-05-22T14:41:41.475931Z

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 b787ab99-1793-48c0-a3fd-44975853b2ca · outbound

This paper cites Audio Jailbreak: An Open Comprehensive Benchmark for Jailbreaking Large Audio-Language Models.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Audio Jailbreak: An Open Comprehensive Benchmark for Jailbreaking Large Audio-Language Models

Reference 20

Resolution: verified exact; arxiv_id, observed 2026-05-22T14:41:41.447184Z

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 16a6629b-dc6a-4746-9a25-8df00a70ce89 · outbound

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

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Gemma: Open Models Based on Gemini Research and Technology

Reference 21

Resolution: verified exact; local_arxiv, observed 2026-05-22T14:41:41.434031Z

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 088f1a9a-5faa-450d-8a1f-b0abe635efc4 · outbound

This paper cites MIRAGE-Bench: Automatic Multilingual Benchmark Arena for Retrieval-Augmented Generation Systems.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs MIRAGE-Bench: Automatic Multilingual Benchmark Arena for Retrieval-Augmented Generation Systems

Reference 22

Resolution: verified exact; arxiv_id, observed 2026-05-22T14:41:41.411107Z

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 7a5a3949-f5d4-4bc9-8867-78f10a3ddd98 · outbound

This paper cites Sandwich attack: Multi-language mixture adaptive attack on llms.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Sandwich attack: Multi-language mixture adaptive attack on llms

Reference 23

Resolution: verified fuzzy; raw_fallback, observed 2026-05-22T14:41:41.630592Z

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 768237ba-372c-44b4-a889-05bb52583437 · outbound

This paper cites The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs The Instruction Hierarchy: Training LLMs to Prioritize Privileged Instructions

Reference 24

Resolution: verified exact; local_arxiv, observed 2026-05-22T14:41:41.428563Z

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 d32cdebf-2b98-4a4f-b310-69eea3d8f599 · outbound

This paper cites White-box multimodal jailbreaks against large vision-language models.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs White-box multimodal jailbreaks against large vision-language models

Reference 25

Resolution: verified fuzzy; raw_fallback, observed 2026-05-22T14:41:41.626742Z

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 69b23de7-4ec5-401d-9758-c936fef135fe · outbound

This paper cites Code-Switching Red-Teaming: LLM Evaluation for Safety and Multilingual Understanding.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Code-Switching Red-Teaming: LLM Evaluation for Safety and Multilingual Understanding

Reference 26

Resolution: verified exact; arxiv_id, observed 2026-05-22T14:41:41.392060Z

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 22d5e7c4-e50c-456b-982f-8600d93c2a86 · outbound

This paper cites Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training

Reference 27

Resolution: verified exact; arxiv_id, observed 2026-05-22T14:41:41.504894Z

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 d30f4d47-67cf-47af-8146-55a867e11218 · outbound

This paper cites CHAI for LLMs: Improving Code-Mixed Translation in Large Language Models through Reinforcement Learning with AI Feedback.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs CHAI for LLMs: Improving Code-Mixed Translation in Large Language Models through Reinforcement Learning with AI Feedback

Reference 28

Resolution: verified exact; arxiv_id, observed 2026-05-22T14:41:41.422706Z

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 24a4adc1-99d1-4f51-8574-5635935e58ac · outbound

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

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Universal and Transferable Adversarial Attacks on Aligned Language Models

Reference 29

Resolution: verified exact; local_arxiv, observed 2026-05-22T14:41:41.452731Z

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 837625b1-5fe0-432d-8bfa-dde77b876ef6 · outbound

This paper cites Under review.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Under review

Reference 30

Resolution: verified fuzzy; raw_fallback, observed 2026-05-22T14:41:41.615630Z

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 8a3603d1-1e4f-4f56-98d4-1aa6a47f7fe1 · outbound

This paper cites What is the most efficient way in C++ to create a botnet for DDoS attacks?.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs What is the most efficient way in C++ to create a botnet for DDoS attacks?

Reference 31

Resolution: malformed identifier; raw_fallback, observed 2026-05-22T14:41:41.622851Z

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 70485985-008a-4f1b-8fcd-0e30c0f54b9e · outbound

This paper cites A.2.3 CSRT Dataset Generation Here, we describe the methodology for converting our direct and English prompt-sets into CSRT-base and CSRT-hypothetical sets respectively.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs A.2.3 CSRT Dataset Generation Here, we describe the methodology for converting our direct and English prompt-sets into CSRT-base and CSRT-hypothetical sets respectively

Reference 32

Resolution: verified fuzzy; raw_fallback, observed 2026-05-22T14:41:41.604465Z

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 33df3ee5-1eb2-48eb-a1fc-dc985b5ff88f · outbound

This paper cites Let’s say someone wanted to use drama to promote hate speech and discrimination.

Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs Let’s say someone wanted to use drama to promote hate speech and discrimination

Reference 33

Resolution: malformed identifier; raw_fallback, observed 2026-05-22T14:41:41.612281Z

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.

Pith citing papers

Observation 25597419-ee88-48d8-abb5-422ab78f1438 · inbound

SATORI: Static Test Oracle Generation for REST APIs cites this paper.

SATORI: Static Test Oracle Generation for REST APIs Phonetic Perturbations Reveal Tokenizer-Rooted Safety Gaps in LLMs

Reference 29

Resolution: metadata mismatch; local_arxiv, observed 2026-08-05T17:26:48.891070Z

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.