Esta página ha sido traducida por una máquina. Otras páginas pueden seguir apareciendo en inglés. View in English

Preste más atención a la robustez de los LLM en el aviso adversario para la minería de datos de instrucciones

  • 0National Key Laboratory of Parallel and Distributed Computing, College of Computer Science and Technology, National University of Defense Technology, Hunan Changsha, 410073, China.

Resumen

Este resumen es generado por máquina.

Este estudio introduce nuevos métodos para mejorar los grandes modelos lingüísticos (LLM) mediante la extracción de datos de instrucción de alta calidad. Desarrollamos técnicas para identificar muestras de instrucción desafiantes, mejorando la robustez de LLM contra las solicitudes adversas.

Área De La Ciencia

  • Inteligencia artificial
  • Procesamiento del lenguaje natural
  • Aprendizaje automático

Sus Antecedentes

  • El ajuste de instrucciones es clave para adaptar los comportamientos del modelo de lenguaje grande (LLM).
  • Se puede lograr un alto rendimiento con datos de instrucción limitados y de alta calidad.
  • La Dificultad de Seguimiento de Instrucciones (IFD) extrae datos cuando los LLM no siguen las instrucciones.

Objetivo Del Estudio

  • Investigar cómo la robustez de LLM a las solicitudes adversarias influye en la selección de datos de instrucción de alta calidad.
  • Proponer un nuevo marco para la minería de datos de instrucción de alta calidad para la sintonización de instrucciones.

Principales Métodos

  • Datos de instrucciones adversas generados mediante instrucciones de ataque.
  • Se introdujo la métrica de Dificultad de Seguimiento de Instrucciones Adversaria (AIFD) utilizando pares de muestras.
  • Desarrollo de la coherencia de incorporación de salida de instrucciones adversas (AIOEC) utilizando solo indicaciones para la minería de datos en línea.

Principales Resultados

  • Los resultados experimentales demuestran la eficacia de los métodos propuestos por el DFAI y la AIOEC.
  • El estudio pone de relieve la importancia de la robustez del LLM frente a las solicitudes adversas en la minería de datos.
  • Ambos métodos identificaron con éxito datos de instrucción de alta calidad para la afinación.

Conclusiones

  • La robustez de LLM a las solicitudes adversarias es crucial para la minería de datos de instrucción efectiva.
  • Los métodos propuestos por el AIFD y la AIOEC ofrecen mejoras significativas en el ajuste de instrucciones.
  • La consideración de la robustez adversarial mejora la calidad y la utilidad de los datos de instrucción minados.

Videos de Conceptos Relacionados

Language Development 01:22

444

Children master language quickly and with relative ease, supported by both biological predisposition and reinforcement. B. F. Skinner (1957) proposed that language is learned through reinforcement, while Noam Chomsky (1965) argued that language acquisition mechanisms are biologically determined.
The critical period for language acquisition suggests that the ability to acquire language is at its peak early in life. As people age, this proficiency decreases. Language development begins very...

Purposive Learning 01:22

204

E. C. Tolman emphasized the purposiveness of behavior — the idea that much of our behavior is goal-directed. For instance, employees who aim for a promotion work diligently to meet their targets. Tolman argued that when classical conditioning and operant conditioning occur, the organism acquires certain expectations. In classical conditioning, a child might fear a dog because they expect it to bite. In operant conditioning, a person might consistently work overtime because they expect a...

Survival Tree 01:19

159

Survival trees are a non-parametric method used in survival analysis to model the relationship between a set of covariates and the time until an event of interest occurs, often referred to as the "time-to-event" or "survival time." This method is particularly useful when dealing with censored data, where the event has not occurred for some individuals by the end of the study period, or when the exact time of the event is unknown.
 Building a Survival Tree
Constructing a...

Generalization, Discrimination, and Extinction 01:24

784

Generalization, discrimination, and extinction are key concepts in operant conditioning that influence how behaviors are learned and maintained.
Generalization occurs when a behavior reinforced in one context is performed in similar situations. For instance, a student who studies diligently for calculus and receives excellent grades might apply the same study habits to psychology and history, expecting similar results. Generalization shows how learning in one setting can influence behavior in...

Language and Cognition 01:27

438

Language serves as a bridge between ideas and communication, influencing how individuals perceive and interact with the world. Psychologists have long debated whether language shapes thought or vice versa. This discussion gained grip with Edward Sapir and Benjamin Lee Whorf in the 1940s, who proposed that language determines thought, a concept known as linguistic determinism. They suggested that the vocabulary and structure of a language influence how its speakers think and perceive reality.

Observational Learning 01:12

310

Albert Bandura's observational learning, also known as imitation or modeling, occurs when a person observes and imitates another's behavior. It is a quicker process than operant conditioning. A well-known example is the Bobo doll study, where children who saw an adult acting aggressively towards the doll were more likely to act aggressively when left alone, compared to those who observed a nonaggressive adult. Many psychologists view observational learning as a form of latent learning...