Teaching Models to Balance Resisting and Accepting Persuasion
This addresses a safety and robustness issue for users of large language models by making them more resilient to harmful persuasion while still open to helpful corrections, though it is an incremental step in model training methods.
The paper tackles the problem of large language models being vulnerable to adversarial persuasion while also needing to accept beneficial persuasion, and introduces Persuasion-Training (PBT) to balance these aspects, resulting in improved resistance to misinformation, better performance on holistic persuasion data, and more stable multi-agent debates with less order dependence.
Large language models (LLMs) are susceptible to persuasion, which can pose risks when models are faced with an adversarial interlocutor. We take a first step towards defending models against persuasion while also arguing that defense against adversarial (i.e. negative) persuasion is only half of the equation: models should also be able to accept beneficial (i.e. positive) persuasion to improve their answers. We show that optimizing models for only one side results in poor performance on the other. In order to balance positive and negative persuasion, we introduce Persuasion-Training (or PBT), which leverages multi-agent recursive dialogue trees to create data and trains models via preference optimization to accept persuasion when appropriate. PBT allows us to use data generated from dialogues between smaller 7-8B models for training much larger 70B models. Moreover, PBT consistently improves resistance to misinformation and resilience to being challenged while also resulting in the best overall performance on holistic data containing both positive and negative persuasion. Crucially, we show that PBT models are better teammates in multi-agent debates across two domains (trivia and commonsense QA). We find that without PBT, pairs of stronger and weaker models have unstable performance, with the order in which the models present their answers determining whether the team obtains the stronger or weaker model's performance. PBT leads to better and more stable results and less order dependence, with the stronger model consistently pulling the weaker one up.