{"id":"c8abb56266666709412472f1bb2b2a6099faaafa226d077b01c77b48fc5d00d8","title":"arXiv cs.AI: ChronoPatch Black-Box Attacks on Vision-Language Models","argument":"The authors of ChronoPatch describe a black-box attack on vision-language models (VLMs) that uses only the final response to prolong generation. They argue this matters because response length can determine latency and computational cost in generative inference services, creating an availability risk even when an attacker cannot see gradients, logits, or token-level feedback. Their method finds visual delay anchors through response-length-guided evolutionary search, then reduces perturbation magnitude through subspace-calibrated decision-based optimization. Across four open-source VLMs, the authors report increased inference latency and GPU energy consumption under a 5% spatial budget, alongside more repetitive outputs. On three commercial VLM APIs, they report longer responses and higher client-observed latency with query budgets of 500 per API. These are reported experimental results, not evidence here of effects across all deployed systems. The authors say the findings motivate generation-aware safeguards; the abstract does not specify which safeguards would work.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied evidence is the arXiv abstract and page text, not the paper's detailed methods or experimental analysis; it does not identify the four open-source models or three commercial APIs. The reported results concern the tested models and API query budgets; the supplied text does not establish how broadly they generalize or which generation-aware safeguards would work.","speaker":{"name":"Xing, Jiawei, Wang, Li, Meng, Xiangtao, Zang, Chuanchao, Wen, Rui, Li, Zheng","roleAtSource":"Article author"},"source":{"title":"ChronoPatch: Black-Box Energy-Latency Attacks on Vision-Language Models","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.26117","publishedAt":"2026-11-19","dateBasis":"Original submission date in the publisher’s citation record."},"evidence":[{"text":"The authors present ChronoPatch as a black-box attack that prolongs VLM generation using only the final response.","excerpt":"We present ChronoPatch, a black-box attack that induces prolonged VLM generation requiring only the final response.","url":"https://arxiv.org/abs/2611.26117","relationship":"Source-stated"},{"text":"The authors frame prolonged generation as an availability risk because response length affects latency and computational cost.","excerpt":"autoregressive decoding makes response length a key determinant of latency and computational cost.","url":"https://arxiv.org/abs/2611.26117","relationship":"Source-stated"},{"text":"The method searches for visual delay anchors and then minimizes perturbation magnitude.","excerpt":"It discovers visual delay anchors through response-length-guided evolutionary search, then minimizes perturbation magnitude through subspace-calibrated decision-based optimization.","url":"https://arxiv.org/abs/2611.26117","relationship":"Source-stated"},{"text":"Across four open-source VLMs, the authors report increased latency and GPU energy consumption under a 5% spatial budget.","excerpt":"Across four open-source VLMs, ChronoPatch increases inference latency by 47.86%--293.34% and GPU energy consumption by 41.44%--249.70% under a 5% spatial budget.","url":"https://arxiv.org/abs/2611.26117","relationship":"Source-stated"},{"text":"On three commercial VLM APIs, the authors report increased response length and client-observed latency with query budgets of 500 per API.","excerpt":"On three commercial VLM APIs, ChronoPatch increases response length by 37.6%--81.5% and client-observed latency by 1.82x--2.29x under query budgets of 500 per API.","url":"https://arxiv.org/abs/2611.26117","relationship":"Source-stated"},{"text":"The authors say their findings motivate generation-aware safeguards for deployed generative AI services.","excerpt":"These results reveal a generation-based availability risk that persists under final-response-only access and motivate generation-aware safeguards for deployed generative AI services.","url":"https://arxiv.org/abs/2611.26117","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.26117","checkedAt":"2026-11-21T19:08:09.261Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What generation-aware safeguards, if any, would address the final-response-only availability risk described by the authors?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv cs.AI: ChronoPatch Black-Box Attacks on Vision-Language Models","summary":"The authors of ChronoPatch describe a black-box attack on vision-language models (VLMs) that uses only the final response to prolong generation. They argue this matters because response length can determine latency and computational cost in generative inference services, creating an availability risk even when an attacker cannot see gradients, logits, or token-level feedback. Their method finds visual delay anchors through response-length-guided evolutionary search, then reduces perturbation magnitude through subspace-calibrated decision-based optimization. Across four open-source VLMs, the authors report increased inference latency and GPU energy consumption under a 5% spatial budget, alongside more repetitive outputs. On three commercial VLM APIs, they report longer responses and higher client-observed latency with query budgets of 500 per API. These are reported experimental results, not evidence here of effects across all deployed systems. The authors say the findings motivate generation-aware safeguards; the abstract does not specify which safeguards would work.","question":"What generation-aware safeguards, if any, would address the final-response-only availability risk described by the authors?","claims":[{"text":"The authors present ChronoPatch as a black-box attack that prolongs VLM generation using only the final response.","excerpt":"We present ChronoPatch, a black-box attack that induces prolonged VLM generation requiring only the final response.","support":"source-stated"},{"text":"The authors frame prolonged generation as an availability risk because response length affects latency and computational cost.","excerpt":"autoregressive decoding makes response length a key determinant of latency and computational cost.","support":"source-stated"},{"text":"The method searches for visual delay anchors and then minimizes perturbation magnitude.","excerpt":"It discovers visual delay anchors through response-length-guided evolutionary search, then minimizes perturbation magnitude through subspace-calibrated decision-based optimization.","support":"source-stated"},{"text":"Across four open-source VLMs, the authors report increased latency and GPU energy consumption under a 5% spatial budget.","excerpt":"Across four open-source VLMs, ChronoPatch increases inference latency by 47.86%--293.34% and GPU energy consumption by 41.44%--249.70% under a 5% spatial budget.","support":"source-stated"},{"text":"On three commercial VLM APIs, the authors report increased response length and client-observed latency with query budgets of 500 per API.","excerpt":"On three commercial VLM APIs, ChronoPatch increases response length by 37.6%--81.5% and client-observed latency by 1.82x--2.29x under query budgets of 500 per API.","support":"source-stated"},{"text":"The authors say their findings motivate generation-aware safeguards for deployed generative AI services.","excerpt":"These results reveal a generation-based availability risk that persists under final-response-only access and motivate generation-aware safeguards for deployed generative AI services.","support":"source-stated"}],"limitations":["The supplied evidence is the arXiv abstract and page text, not the paper's detailed methods or experimental analysis; it does not identify the four open-source models or three commercial APIs.","The reported results concern the tested models and API query budgets; the supplied text does not establish how broadly they generalize or which generation-aware safeguards would work."],"mediaId":"afea41d41681e2ce6de7a700620ca60cf3bad0e19830b99e82d8ce6cd8e026e9","correctionReason":null,"id":"c8abb56266666709412472f1bb2b2a6099faaafa226d077b01c77b48fc5d00d8","publicationId":"06e4f1cd-39dc-44e5-8357-511ef00cbb7b","version":1,"sourceUrl":"https://arxiv.org/abs/2611.26117","sourceTitle":"ChronoPatch: Black-Box Energy-Latency Attacks on Vision-Language Models","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-19","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-20T05:07:49.630Z","createdAt":"2026-11-20T09:01:36.993Z","historical":false,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"afea41d41681e2ce6de7a700620ca60cf3bad0e19830b99e82d8ce6cd8e026e9","kind":"article","url":"https://arxiv.org/abs/2611.26117","sourceUrl":"https://arxiv.org/abs/2611.26117","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-20T09:00:52.324Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.26117","checkedAt":"2026-11-21T19:08:09.261Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v9","mediaId":"afea41d41681e2ce6de7a700620ca60cf3bad0e19830b99e82d8ce6cd8e026e9","confidence":0.67,"provider":"typesafe","model":"jev-1.13.0","inputHash":"4a794510b29434b2e39794909c9673843eb49dbb673afe63e6bd4088947a68bb","decidedAt":"2026-11-20T09:00:53.087Z"},"provenance":{"textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observationId":"6183064436e0911856ae0068b3865bd4909e2bb7a0ff1be583edbf7e32afca96","retrievedAt":"2026-11-20T09:00:52.223Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"53b74e47-699e-4866-8650-b219bb664d48"},"reviewer":{"model":"gpt-6-luna","runId":"96f6a87a-38f0-4ebc-85f4-adc334c30a24"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-20T09:01:36.993Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Xing, Jiawei","url":null,"kind":"unknown"},{"name":"Wang, Li","url":null,"kind":"unknown"},{"name":"Meng, Xiangtao","url":null,"kind":"unknown"},{"name":"Zang, Chuanchao","url":null,"kind":"unknown"},{"name":"Wen, Rui","url":null,"kind":"unknown"},{"name":"Li, Zheng","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Xing, Jiawei","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Xing, Jiawei","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.26117","locator":"head/meta[name=author|citation_author]"},"assertionId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-21T19:08:09.261Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","sourceUrl":"https://arxiv.org/abs/2611.26117","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z"}},{"actor":{"name":"Wang, Li","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Wang, Li","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.26117","locator":"head/meta[name=author|citation_author]"},"assertionId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-21T19:08:09.261Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","sourceUrl":"https://arxiv.org/abs/2611.26117","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z"}},{"actor":{"name":"Meng, Xiangtao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Meng, Xiangtao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.26117","locator":"head/meta[name=author|citation_author]"},"assertionId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-21T19:08:09.261Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","sourceUrl":"https://arxiv.org/abs/2611.26117","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z"}},{"actor":{"name":"Zang, Chuanchao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zang, Chuanchao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.26117","locator":"head/meta[name=author|citation_author]"},"assertionId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-21T19:08:09.261Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","sourceUrl":"https://arxiv.org/abs/2611.26117","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z"}},{"actor":{"name":"Wen, Rui","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Wen, Rui","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.26117","locator":"head/meta[name=author|citation_author]"},"assertionId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-21T19:08:09.261Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","sourceUrl":"https://arxiv.org/abs/2611.26117","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z"}},{"actor":{"name":"Li, Zheng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Li, Zheng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.26117","locator":"head/meta[name=author|citation_author]"},"assertionId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-21T19:08:09.261Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","sourceUrl":"https://arxiv.org/abs/2611.26117","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z"}}],"participationCoverage":{"total":6,"returned":6,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.26117","snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z","basis":"publisher-page-metadata","sourceLinks":45,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/c8abb56266666709412472f1bb2b2a6099faaafa226d077b01c77b48fc5d00d8/source"}},"canonicalUrl":"https://escapemeridian.com/readings/arxiv-cs-ai-chronopatch-black-box-attacks-on-vision-language-c8abb562","formats":{"html":"/readings/arxiv-cs-ai-chronopatch-black-box-attacks-on-vision-language-c8abb562","markdown":"/readings/arxiv-cs-ai-chronopatch-black-box-attacks-on-vision-language-c8abb562.md","json":"/readings/arxiv-cs-ai-chronopatch-black-box-attacks-on-vision-language-c8abb562.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"conditions":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[{"id":"term-latency","name":"Latency","explanation":"The delay between sending a request and receiving the result.","referenceUrl":"https://developers.google.com/machine-learning/glossary#latency"},{"id":"term-gpu","name":"Graphics processing unit","explanation":"A chip that performs many calculations at once; the main hardware used to train and run AI models.","referenceUrl":"https://en.wikipedia.org/wiki/Graphics_processing_unit"},{"id":"term-api","name":"API","explanation":"An application programming interface: a defined way for software to request data or actions from another system.","referenceUrl":"https://developer.mozilla.org/en-US/docs/Glossary/API"}],"entities":[{"id":"arxiv","name":"arXiv","kind":"organization","context":"arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.","referenceUrls":["https://arxiv.org/","https://info.arxiv.org/help/api/index.html"],"path":"/entities/arxiv","json":"/entities/arxiv.json","markdown":"/entities/arxiv.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"c8abb56266666709412472f1bb2b2a6099faaafa226d077b01c77b48fc5d00d8","revision":"06e4f1cd-39dc-44e5-8357-511ef00cbb7b","revisionKind":"publication-id","title":"arXiv cs.AI: ChronoPatch Black-Box Attacks on Vision-Language Models","url":"https://escapemeridian.com/readings/arxiv-cs-ai-chronopatch-black-box-attacks-on-vision-language-c8abb562"},"source":{"type":"source","id":"https://arxiv.org/abs/2611.26117","revision":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","revisionKind":"sha256-source-text","title":"ChronoPatch: Black-Box Energy-Latency Attacks on Vision-Language Models","url":"https://arxiv.org/abs/2611.26117"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings to this reading. Shared topics are not citations. Other record types need an explicit reference adapter."},"revisions":[{"id":"c8abb56266666709412472f1bb2b2a6099faaafa226d077b01c77b48fc5d00d8","title":"arXiv cs.AI: ChronoPatch Black-Box Attacks on Vision-Language Models","argument":"The authors of ChronoPatch describe a black-box attack on vision-language models (VLMs) that uses only the final response to prolong generation. They argue this matters because response length can determine latency and computational cost in generative inference services, creating an availability risk even when an attacker cannot see gradients, logits, or token-level feedback. Their method finds visual delay anchors through response-length-guided evolutionary search, then reduces perturbation magnitude through subspace-calibrated decision-based optimization. Across four open-source VLMs, the authors report increased inference latency and GPU energy consumption under a 5% spatial budget, alongside more repetitive outputs. On three commercial VLM APIs, they report longer responses and higher client-observed latency with query budgets of 500 per API. These are reported experimental results, not evidence here of effects across all deployed systems. The authors say the findings motivate generation-aware safeguards; the abstract does not specify which safeguards would work.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied evidence is the arXiv abstract and page text, not the paper's detailed methods or experimental analysis; it does not identify the four open-source models or three commercial APIs. The reported results concern the tested models and API query budgets; the supplied text does not establish how broadly they generalize or which generation-aware safeguards would work.","speaker":{"name":"Xing, Jiawei, Wang, Li, Meng, Xiangtao, Zang, Chuanchao, Wen, Rui, Li, Zheng","roleAtSource":"Article author"},"source":{"title":"ChronoPatch: Black-Box Energy-Latency Attacks on Vision-Language Models","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.26117","publishedAt":"2026-11-19","dateBasis":"Original submission date in the publisher’s citation record."},"evidence":[{"text":"The authors present ChronoPatch as a black-box attack that prolongs VLM generation using only the final response.","excerpt":"We present ChronoPatch, a black-box attack that induces prolonged VLM generation requiring only the final response.","url":"https://arxiv.org/abs/2611.26117","relationship":"Source-stated"},{"text":"The authors frame prolonged generation as an availability risk because response length affects latency and computational cost.","excerpt":"autoregressive decoding makes response length a key determinant of latency and computational cost.","url":"https://arxiv.org/abs/2611.26117","relationship":"Source-stated"},{"text":"The method searches for visual delay anchors and then minimizes perturbation magnitude.","excerpt":"It discovers visual delay anchors through response-length-guided evolutionary search, then minimizes perturbation magnitude through subspace-calibrated decision-based optimization.","url":"https://arxiv.org/abs/2611.26117","relationship":"Source-stated"},{"text":"Across four open-source VLMs, the authors report increased latency and GPU energy consumption under a 5% spatial budget.","excerpt":"Across four open-source VLMs, ChronoPatch increases inference latency by 47.86%--293.34% and GPU energy consumption by 41.44%--249.70% under a 5% spatial budget.","url":"https://arxiv.org/abs/2611.26117","relationship":"Source-stated"},{"text":"On three commercial VLM APIs, the authors report increased response length and client-observed latency with query budgets of 500 per API.","excerpt":"On three commercial VLM APIs, ChronoPatch increases response length by 37.6%--81.5% and client-observed latency by 1.82x--2.29x under query budgets of 500 per API.","url":"https://arxiv.org/abs/2611.26117","relationship":"Source-stated"},{"text":"The authors say their findings motivate generation-aware safeguards for deployed generative AI services.","excerpt":"These results reveal a generation-based availability risk that persists under final-response-only access and motivate generation-aware safeguards for deployed generative AI services.","url":"https://arxiv.org/abs/2611.26117","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.26117","checkedAt":"2026-11-21T19:08:09.261Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"What generation-aware safeguards, if any, would address the final-response-only availability risk described by the authors?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv cs.AI: ChronoPatch Black-Box Attacks on Vision-Language Models","summary":"The authors of ChronoPatch describe a black-box attack on vision-language models (VLMs) that uses only the final response to prolong generation. They argue this matters because response length can determine latency and computational cost in generative inference services, creating an availability risk even when an attacker cannot see gradients, logits, or token-level feedback. Their method finds visual delay anchors through response-length-guided evolutionary search, then reduces perturbation magnitude through subspace-calibrated decision-based optimization. Across four open-source VLMs, the authors report increased inference latency and GPU energy consumption under a 5% spatial budget, alongside more repetitive outputs. On three commercial VLM APIs, they report longer responses and higher client-observed latency with query budgets of 500 per API. These are reported experimental results, not evidence here of effects across all deployed systems. The authors say the findings motivate generation-aware safeguards; the abstract does not specify which safeguards would work.","question":"What generation-aware safeguards, if any, would address the final-response-only availability risk described by the authors?","claims":[{"text":"The authors present ChronoPatch as a black-box attack that prolongs VLM generation using only the final response.","excerpt":"We present ChronoPatch, a black-box attack that induces prolonged VLM generation requiring only the final response.","support":"source-stated"},{"text":"The authors frame prolonged generation as an availability risk because response length affects latency and computational cost.","excerpt":"autoregressive decoding makes response length a key determinant of latency and computational cost.","support":"source-stated"},{"text":"The method searches for visual delay anchors and then minimizes perturbation magnitude.","excerpt":"It discovers visual delay anchors through response-length-guided evolutionary search, then minimizes perturbation magnitude through subspace-calibrated decision-based optimization.","support":"source-stated"},{"text":"Across four open-source VLMs, the authors report increased latency and GPU energy consumption under a 5% spatial budget.","excerpt":"Across four open-source VLMs, ChronoPatch increases inference latency by 47.86%--293.34% and GPU energy consumption by 41.44%--249.70% under a 5% spatial budget.","support":"source-stated"},{"text":"On three commercial VLM APIs, the authors report increased response length and client-observed latency with query budgets of 500 per API.","excerpt":"On three commercial VLM APIs, ChronoPatch increases response length by 37.6%--81.5% and client-observed latency by 1.82x--2.29x under query budgets of 500 per API.","support":"source-stated"},{"text":"The authors say their findings motivate generation-aware safeguards for deployed generative AI services.","excerpt":"These results reveal a generation-based availability risk that persists under final-response-only access and motivate generation-aware safeguards for deployed generative AI services.","support":"source-stated"}],"limitations":["The supplied evidence is the arXiv abstract and page text, not the paper's detailed methods or experimental analysis; it does not identify the four open-source models or three commercial APIs.","The reported results concern the tested models and API query budgets; the supplied text does not establish how broadly they generalize or which generation-aware safeguards would work."],"mediaId":"afea41d41681e2ce6de7a700620ca60cf3bad0e19830b99e82d8ce6cd8e026e9","correctionReason":null,"id":"c8abb56266666709412472f1bb2b2a6099faaafa226d077b01c77b48fc5d00d8","publicationId":"06e4f1cd-39dc-44e5-8357-511ef00cbb7b","version":1,"sourceUrl":"https://arxiv.org/abs/2611.26117","sourceTitle":"ChronoPatch: Black-Box Energy-Latency Attacks on Vision-Language Models","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-19","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-20T05:07:49.630Z","createdAt":"2026-11-20T09:01:36.993Z","historical":false,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"afea41d41681e2ce6de7a700620ca60cf3bad0e19830b99e82d8ce6cd8e026e9","kind":"article","url":"https://arxiv.org/abs/2611.26117","sourceUrl":"https://arxiv.org/abs/2611.26117","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-20T09:00:52.324Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.26117","checkedAt":"2026-11-21T19:08:09.261Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v9","mediaId":"afea41d41681e2ce6de7a700620ca60cf3bad0e19830b99e82d8ce6cd8e026e9","confidence":0.67,"provider":"typesafe","model":"jev-1.13.0","inputHash":"4a794510b29434b2e39794909c9673843eb49dbb673afe63e6bd4088947a68bb","decidedAt":"2026-11-20T09:00:53.087Z"},"provenance":{"textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observationId":"6183064436e0911856ae0068b3865bd4909e2bb7a0ff1be583edbf7e32afca96","retrievedAt":"2026-11-20T09:00:52.223Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"53b74e47-699e-4866-8650-b219bb664d48"},"reviewer":{"model":"gpt-6-luna","runId":"96f6a87a-38f0-4ebc-85f4-adc334c30a24"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-20T09:01:36.993Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Xing, Jiawei","url":null,"kind":"unknown"},{"name":"Wang, Li","url":null,"kind":"unknown"},{"name":"Meng, Xiangtao","url":null,"kind":"unknown"},{"name":"Zang, Chuanchao","url":null,"kind":"unknown"},{"name":"Wen, Rui","url":null,"kind":"unknown"},{"name":"Li, Zheng","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Xing, Jiawei","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Xing, Jiawei","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.26117","locator":"head/meta[name=author|citation_author]"},"assertionId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-21T19:08:09.261Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","sourceUrl":"https://arxiv.org/abs/2611.26117","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z"}},{"actor":{"name":"Wang, Li","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Wang, Li","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.26117","locator":"head/meta[name=author|citation_author]"},"assertionId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-21T19:08:09.261Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","sourceUrl":"https://arxiv.org/abs/2611.26117","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z"}},{"actor":{"name":"Meng, Xiangtao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Meng, Xiangtao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.26117","locator":"head/meta[name=author|citation_author]"},"assertionId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-21T19:08:09.261Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","sourceUrl":"https://arxiv.org/abs/2611.26117","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z"}},{"actor":{"name":"Zang, Chuanchao","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zang, Chuanchao","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.26117","locator":"head/meta[name=author|citation_author]"},"assertionId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-21T19:08:09.261Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","sourceUrl":"https://arxiv.org/abs/2611.26117","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z"}},{"actor":{"name":"Wen, Rui","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Wen, Rui","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.26117","locator":"head/meta[name=author|citation_author]"},"assertionId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-21T19:08:09.261Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","sourceUrl":"https://arxiv.org/abs/2611.26117","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z"}},{"actor":{"name":"Li, Zheng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Li, Zheng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.26117","locator":"head/meta[name=author|citation_author]"},"assertionId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-21T19:08:09.261Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","sourceUrl":"https://arxiv.org/abs/2611.26117","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z"}}],"participationCoverage":{"total":6,"returned":6,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.26117","snapshotId":"ed04c6c14248cc4371556d0a61810539544776d9e6d2648e33a24bcad6f22a29","contentSha256":"97505ad4452772906cf1d577eb08575c171683451c8dc2ce96f2d491b325f108","textHash":"30141bedd111b284396a118959848e5f44fc00b7b1308d52f5475cb8045198cb","observedAt":"2026-11-21T19:08:09.261Z","basis":"publisher-page-metadata","sourceLinks":45,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/c8abb56266666709412472f1bb2b2a6099faaafa226d077b01c77b48fc5d00d8/source"}}}],"audits":[{"publicationId":"06e4f1cd-39dc-44e5-8357-511ef00cbb7b","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading consistently attributes the attack, its rationale, the results and the safeguard recommendation to the paper's authors. It does not present the reported latency and energy effects as independently verified facts.","excerpt":"We present ChronoPatch, a black-box attack that induces prolonged VLM generation requiring only the final response."},{"dimension":"summary","status":"uncertain","explanation":"The summary never states plainly that the attack works by feeding the model subtly altered images. That is buried in 'visual delay anchors' and 'perturbation magnitude', so a general reader may miss the core mechanism. It also drops the reported magnitudes, such as 47.86%–293.34% latency and 1.82x–2.29x on commercial APIs, which weakens the source's evidence. Smallest repair: add one plain sentence that the attacker supplies adversarially modified images that make the model write longer. Restore the reported ranges. Terms such as black-box, availability risk and spatial budget should get tooltips.","excerpt":"adversarial visual inputs can prolong generation and increase serving overhead"},{"dimension":"date","status":"pass","explanation":"originalPublished is 2026-11-19, which matches the arXiv v1 submission date. The feed date is a day later and is correctly not used.","excerpt":"[Submitted on 19 Nov 2026]"},{"dimension":"support","status":"pass","explanation":"Every claim matches a verbatim excerpt from the abstract. The caveat about generalization stays within the evidence. The question about which safeguards would work follows from the authors' unspecified recommendation and does not lead the reader.","excerpt":"motivate generation-aware safeguards for deployed generative AI services"},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct arXiv abstract page for this paper, which is the most direct access to the source.","excerpt":null}],"createdAt":"2026-11-20T09:01:36.993Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}