Think
LLM
Models
Capabilities
Use Cases
Benchmarks
Papers
Glossary
Search
/
Poisoned Responses — Glossary — ThinkLLM
Glossary
/
Poisoned Responses
Poisoned Responses
techniques
Malicious outputs deliberately generated by a compromised model when triggered by backdoor inputs.