paper-with-me

홈 › Papers

KTO: Model Alignment as Prospect Theoretic Optimization

2024-02-02 · Kawin Ethayarajh, Winnie Xu, Niklas Muennighoff, Dan Jurafsky, Douwe Kiela

Kahneman & Tversky's $\textit{prospect theory}$ tells us that humans perceive random variables in a biased but well-defined manner (1992); for example, humans are famously loss-averse. We show that objectives for aligning LLMs with human feedback implicitly incorporate many of these biases -- the success of these objectives (e.g., DPO) over cross-entropy minimization can partly be ascribed to them belonging to a family of loss functions that we call $\textit{human-aware losses}$ (HALOs). However, the utility functions these methods attribute to humans still differ from those in the prospect theory literature. Using a Kahneman-Tversky model of human utility, we propose a HALO that directly maximizes the utility of generations instead of maximizing the log-likelihood of preferences, as current methods do. We call this approach KTO, and it matches or exceeds the performance of preference-based methods at scales from 1B to 30B, despite only learning from a binary signal of whether an output is desirable. More broadly, our work suggests that there is no one HALO that is universally superior; the best loss depends on the inductive biases most appropriate for a given setting, an oft-overlooked consideration.

📄 PDF Abstract BibTeX arXiv:2402.01306

Code (5)

contextualai/halos 공식 구현 pytorch
openbmb/eurus pytorch
reml-group/deliberation-on-priors pytorch
tidedra/vl-rlhf pytorch
uppaal/detox-edit pytorch

Tasks

Attributemodel

Similar Papers 제목 키워드 기반

Aligning Diffusion Language Models via Unpaired Preference Optimization

2025-10-26 · Vaibhav Jindal, Hejian Sang, Chun-Mao Lai, Yanning Chen 외 arxiv

Diffusion language models (dLLMs) are an emerging alternative to autoregressive (AR) generators, but aligning them to human preferences is challenging because sequence log-likelihoods are intractable and pairwise prefere…

Review and Prospect of Algebraic Research in Equivalent Framework between Statistical Mechanics and Machine Learning Theory

2024-05-31 · Sumio Watanabe

Mathematical equivalence between statistical mechanics and machine learning theory has been known since the 20th century, and research based on this equivalence has provided novel methodologies in both theoretical physic…

Learning Theory

Prospective Learning: Learning for a Dynamic Future

2024-10-31 · Ashwin De Silva, Rahul Ramesh, Rubing Yang, Siyu Yu 외

In real-world applications, the distribution of the data, and our goals, evolve over time. The prevailing theoretical framework for studying machine learning, namely probably approximately correct (PAC) learning, largely…

PAC learning

Limitations of Information-Theoretic Generalization Bounds for Gradient Descent Methods in Stochastic Convex Optimization

2022-12-27 · Mahdi Haghifam, Borja Rodríguez-Gálvez, Ragnar Thobaben, Mikael Skoglund 외

To date, no "information-theoretic" frameworks for reasoning about generalization error have been shown to establish minimax rates for gradient descent in the setting of stochastic convex optimization. In this work, we c…

Generalization Bounds

Cumulative Prospect Theory Meets Reinforcement Learning: Prediction and Control

2015-06-08 · Prashanth L. A., Cheng Jie, Michael Fu, Steve Marcus 외

Cumulative prospect theory (CPT) is known to model human decisions well, with substantial empirical evidence supporting this claim. CPT works by distorting probabilities and is more general than the classic expected util…

Predictionreinforcement-learningReinforcement LearningReinforcement Learning (RL)+1