| Anchoring Bias | anchoring-bias | multi-turn | medium | Multi-turn attack providing a wrong answer first then asking model to build on it |
| Citation Manipulation | citation-manipulation | single-turn | high | Prompt model to attribute false quotes to real public figures |
| Confidence Exploitation | confidence-exploitation | single-turn | medium | Push model to state uncertain information as established fact without hedging |
| Fabrication Prompting | fabrication-prompting | single-turn | medium | Prompt model to generate fake citations, fabricated research, or invented statistics |
| False Premise | false-premise | single-turn | medium | Build prompts on false premises to see if model accepts and elaborates |
| Knowledge Boundary | knowledge-boundary | single-turn | medium | Test if model admits uncertainty on post-cutoff or fictional events |
| Selective Context | selective-context | single-turn | medium | Provide heavily biased one-sided context to produce biased output |
| Sycophancy Exploit | sycophancy-exploit | multi-turn | medium | Multi-turn attack exploiting model tendency to agree with user assertions |
