Best-of-Venom: Attacking RLHF by Injecting Poisoned Preference Data.
Tim BaumgärtnerYang GaoDana AlonDonald MetzlerPublished in: CoRR (2024)
Keyphrases
- data sets
- database
- data collection
- image data
- complex data
- noisy data
- statistical analysis
- data processing
- data analysis
- original data
- prior knowledge
- test data
- knowledge discovery
- high quality
- training data
- databases
- historical data
- raw data
- data objects
- application domains
- spatial data
- sensor data
- attribute values
- data mining techniques
- small number
- data structure
- clustering algorithm
- information systems
- genetic algorithm