Large language models (LLMs) have made significant advancements, but their increasing capabilities present serious risks of misuse, particularly in open-weight models where direct access to the model’s parameters is possible. Current safeguards, designed for closed-weight API models, are inadequate for open-weight models, as minimal fine-tuning can bypass these protections. Preserving the integrity of open-weight LLMs before deployment has thus become a critical challenge. We argue that these vulnerabilities stem from the overemphasis on maximizing the LLM’s log-likelihood during training, which amplifies data biases, especially with large datasets. To address these issues, we introduce Kahneman and Tversky’s Prospect Theoretic Integrity Preserving Alignment (KT-IPA), a framework that prioritizes maximizing generative utility rather than a singular optimization metric. This approach strengthens LLMs against misuse and weaponization while maintaining high performance, even after extensive fine-tuning. Our results demonstrate that integrating prospect theory into LLM training enhances robustness, security, and responsible innovation in this rapidly evolving field. Our codes are available on https://anonymous.4open.science/r/KT-IPA-40B7

Accurate and personalized product recommendation is central to user satisfaction in e-commerce. However, a persistent language gap often exists between user queries and product titles or descriptions. While traditional user behavior-based recommenders and LLM-based Retrieval-Augmented Generation systems typically optimize for maximum likelihood objectives, they may struggle to bridge this gap or capture users’ true intent. In this paper, we propose a strategy based on Prospect Theoretic Self-Alignment, that reframes LLM-based recommendations as a utility-driven process. Given a user query and a set of candidate products, our model acts as a seller who anticipates latent user needs and generates product descriptions tailored to the user’s perspective. Simultaneously, it simulates user decision-making utility to assess whether the generated content would lead to a purchase. This self-alignment is achieved through a training strategy grounded in Kahneman & Tversky’s prospect theory, ensuring that recommendations are optimized for perceived user value rather than likelihood alone. Experiments on real-world product data demonstrate substantial improvements in intent alignment and recommendation quality, validating the effectiveness of our approach in producing personalized and decision-aware recommendations.

The burgeoning generative capabilities of large language models (LLMs) have raised growing concerns about abuse, demanding automatic machine-generated text detectors. DetectGPT, a zero-shot metric-based detector, first introduces perturbation and shows great performance improvement. However, in DetectGPT, the random perturbation strategy could introduce noise, and logit regression depends on the threshold, harming the generalizability and applicability of individual or small-batch inputs. Hence, we propose a novel fine-tuned detector, PECOLA, bridging metric-based and fine-tuned methods by contrastive learning on selective perturbation. Selective strategy retains important tokens during perturbation and weights for multi-pair contrastive learning. The experiments show that PECOLA outperforms the state-of-the-art (SOTA) by 1.20% in accuracy on average on four public datasets. And we further analyze the effectiveness, robustness, and generalization of the method.