Conditional Attribute Estimation with Autoregressive Sequence Models

TL;DR

This paper introduces Conditional Attribute Transformers, a method that jointly estimates token probabilities and sequence attributes, enabling efficient attribute control, counterfactual analysis, and improved sequence generation.

Contribution

The paper presents a novel framework that combines attribute estimation with autoregressive models, improving global property control without sequence modification.

Findings

Achieves state-of-the-art performance on sparse reward tasks.

Estimates attribute probabilities significantly faster than sampling.

Enhances next-token prediction with larger models.

Abstract

Generative models are often trained with a next-token prediction objective, yet many downstream applications require the ability to estimate or control sequence-level properties. Next-token prediction can lead to overfitting of local patterns during training, underfitting of global structure, and requires significant downstream modifications or expensive sampling to guide or predict the global attributes of generated samples at inference time. Here, we introduce Conditional Attribute Transformers, a novel method for jointly estimating the next-token probability and the value of an attribute conditional on each potential next token selection. This framework enables three critical capabilities within a single forward pass, without modification of the input sequence: (1) per-token credit assignment across an entire sequence, by identifying how each token in a sequence is associated with an attribute's value; (2) counterfactual analysis, by quantifying attribute differences conditional on alternative next token choices; (3) steerable generation, by decoding sequences based on a combination of next-token and attribute likelihoods. Our approach achieves state of the art performance on sparse reward tasks, improves next-token prediction at sufficient model sizes, estimates attribute probabilities orders of magnitude faster than sampling, and can guide decoding of autoregressive sequence models on a range of language tasks.