LF Security Corpus: Tokenizer & Processor Security
Tokenizer templates, role boundaries, processor classes and Unicode mutation challenges.
UpdatedNote LF-CH-TOKX-0012: Processor custom class reference.
LayerFault/processor-multimodal-sidecar-indirection
UpdatedNote LF-CH-TOKX-0013: Processor multimodal sidecar indirection.
LayerFault/tokenizer-added-token-shadowing
UpdatedNote LF-CH-TOKX-0011: Tokenizer added token shadowing.
LayerFault/tokenizer-bos-eos-conflict
UpdatedNote LF-CH-TOKX-0002: Tokenizer bos eos conflict.
LayerFault/tokenizer-clean-template-control
UpdatedNote LF-CH-TOKX-0001: Tokenizer clean template control.
LayerFault/tokenizer-dynamic-include-template
UpdatedNote LF-CH-TOKX-0008: Tokenizer dynamic include template.
LayerFault/tokenizer-encoded-template-marker
UpdatedNote LF-CH-TOKX-0009: Tokenizer encoded template marker.
LayerFault/tokenizer-hidden-prompt-plain
UpdatedNote LF-CH-TOKX-0003: Tokenizer hidden prompt plain.
LayerFault/tokenizer-hidden-prompt-zero-width
UpdatedNote LF-CH-TOKX-0004: Tokenizer hidden prompt zero width.
LayerFault/tokenizer-role-boundary-confusion
UpdatedNote LF-CH-TOKX-0006: Tokenizer role boundary confusion.
LayerFault/tokenizer-role-remap
UpdatedNote LF-CH-TOKX-0005: Tokenizer role remap.
LayerFault/tokenizer-security-words-control
UpdatedNote LF-CH-TOKX-0014: Tokenizer security words control.
LayerFault/tokenizer-special-token-reassignment
UpdatedNote LF-CH-TOKX-0010: Tokenizer special token reassignment.
LayerFault/tokenizer-tool-template-introspection
UpdatedNote LF-CH-TOKX-0007: Tokenizer tool template introspection.