Silent Sabotage During Fine-Tuning: Few-Shot Rationale Poisoning of Compact Medical LLMs

arXiv 2603.02262•85b9aaf5b04ef7cecc88e48264988a379ccca4b680ce4534700b532afd38d05a
Android PendingIntentC2PA integrity clashDocker/OCIFHELLM agent securityLLM poisoningSDK authentication bypassSFT poisoningTraceGuardVA-DAR wallet recoveryZeroDay evaluationZeroDayBenchblockchain communication vulnerabilitiescomposable attestationcontainer securityfederated learningmedical AI safetymobile supply chainpost-quantum cryptographyreasoning backdoorssandbox escapespeaker impersonationsupply-chain securityvoice spoofingwatermarking

Paper metadata

arXiv ID
2603.02262
Version
Not specified by this published record
Category
Computer Science — Cryptography and Security (cs.CR)

The PDF link points to arxiv.org. Baitaphish does not expose a private stored PDF.

Evidence and limitations

Source ID
arxiv_cs_cr
Record identifier
85b9aaf5b04ef7cecc88e48264988a379ccca4b680ce4534700b532afd38d05a
Enrichment time
2026-03-04T19:45:48Z
AI-assisted enrichment
Yes

This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.