Paper Citation Record · LEDGER

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation

As of 11 August 2026, this Paper Citation Record lists 12 of 12 outbound references and 0 inbound Pith citation observations for arXiv:2604.06950.

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

pith.paper-citation-record.v1
2604.06950 v2

Coverage vector

measured 12 of 12 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-05-10T18:19:58.970952Z

measured 12 of 12 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-11T06:34:44.6726+00:00

measured 0 of 0 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links

measured 0 of 1 external citation measurements

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

Source: cited_works

Reference resolution

12 of 12 outbound references displayed

verified exact1
verified fuzzy7
unresolved0
parse uncertain0
malformed identifier1
metadata mismatch3

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation 1425f0d0-6b4f-407c-9291-766d9aecf743 · outbound

This paper cites BERTopic: Neural topic modeling with a class-based TF-IDF procedure.

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation BERTopic: Neural topic modeling with a class-based TF-IDF procedure

Reference 1

Resolution: metadata mismatch; arxiv_id, observed 2026-05-11T14:48:37.128055Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation e95773c8-1cf3-4663-99de-ce6ff12533fe · outbound

This paper cites InPro- ceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation InPro- ceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V

Reference 2

Resolution: verified fuzzy; raw_fallback, observed 2026-05-17T05:14:05.708835Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation f22ac78a-055d-4254-a4df-a8e26b4706be · outbound

This paper cites Visual-RolePlay: Universal Jailbreak Attack on MultiModal Large Language Models via Role-playing Image Character.

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation Visual-RolePlay: Universal Jailbreak Attack on MultiModal Large Language Models via Role-playing Image Character

Reference 3

Resolution: metadata mismatch; arxiv_id, observed 2026-05-11T00:45:50.274862Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation f807a792-f5b5-47e1-b298-6f4a8d2bf1d8 · outbound

This paper cites UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction.

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

Reference 4

Resolution: verified exact; local_arxiv, observed 2026-05-11T00:45:50.266880Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 389d02ca-cd17-4b44-b432-c37ec7ebf6c0 · outbound

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

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation Llama 2: Open Foundation and Fine-Tuned Chat Models

Reference 5

Resolution: metadata mismatch; local_arxiv, observed 2026-05-11T00:45:50.278532Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 5c372ae1-90ae-4fc2-bc82-fbe7776ca8e8 · outbound

This paper cites jailbreaking.

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation jailbreaking

Reference 6

Resolution: verified fuzzy; raw_fallback, observed 2026-05-17T05:14:05.700740Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation bf47cfc8-245e-4cc6-a1f8-db3f99901634 · outbound

This paper cites visual key.

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation visual key

Reference 7

Resolution: verified fuzzy; raw_fallback, observed 2026-05-17T05:14:05.695784Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation 2fbfa3bf-1dd5-4057-a07c-db325b1b7820 · outbound

This paper cites It uses a con- cise two-step logic to decouple perception (OCR) from reasoning (Violation Check).

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation It uses a con- cise two-step logic to decouple perception (OCR) from reasoning (Violation Check)

Reference 8

Resolution: verified fuzzy; raw_fallback, observed 2026-05-17T05:14:05.704884Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation b3bbcf9b-146e-4359-a920-0eda14430b23 · outbound

This paper cites A bouquet of wildflowers tied with placed in a jar.

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation A bouquet of wildflowers tied with placed in a jar

Reference 9

Resolution: verified fuzzy; raw_fallback, observed 2026-05-17T05:14:05.691122Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation d746679b-a244-4dfe-ab35-5898fd920b9f · outbound

This paper cites As shown in Table 6, even flagship models likeGPT-5and Claude-Sonnet-4.5struggle significantly in these categories.

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation As shown in Table 6, even flagship models likeGPT-5and Claude-Sonnet-4.5struggle significantly in these categories

Reference 10

Resolution: verified fuzzy; raw_fallback, observed 2026-05-17T05:14:05.713212Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation e2534d71-6e17-4ad6-9eda-b691f1fb9627 · outbound

This paper cites While models are proficient at reading clean digital text, they fail to generalize to text that is partially obstructed or stylistically distorted.

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation While models are proficient at reading clean digital text, they fail to generalize to text that is partially obstructed or stylistically distorted

Reference 11

Resolution: verified fuzzy; raw_fallback, observed 2026-05-17T05:14:05.718046Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Observation cd6d72dc-c6c4-4bdf-b8c6-04baba43c13a · outbound

This paper cites an unresolved cited work.

Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderation Unresolved cited work

Reference 12

Resolution: malformed identifier; raw_fallback, observed 2026-05-17T05:14:05.721670Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-11T06:34:44.6726+00:00.

Pith citing papers

No inbound Pith citation observations are available.