OASST-ETC Dataset: Alignment Signals from Eye-tracking Analysis of LLM Responses
This provides a novel resource for researchers studying human cognitive processing in LLM evaluation, though it is incremental as it suggests directions rather than implementing new alignment methods.
The paper tackles the challenge of aligning LLMs with human preferences by introducing OASST-ETC, an eye-tracking dataset from 24 participants that reveals distinct reading patterns between preferred and non-preferred LLM responses, showing stronger correlations with transformer attention patterns in preferred responses.
While Large Language Models (LLMs) have significantly advanced natural language processing, aligning them with human preferences remains an open challenge. Although current alignment methods rely primarily on explicit feedback, eye-tracking (ET) data offers insights into real-time cognitive processing during reading. In this paper, we present OASST-ETC, a novel eye-tracking corpus capturing reading patterns from 24 participants, while evaluating LLM-generated responses from the OASST1 dataset. Our analysis reveals distinct reading patterns between preferred and non-preferred responses, which we compare with synthetic eye-tracking data. Furthermore, we examine the correlation between human reading measures and attention patterns from various transformer-based models, discovering stronger correlations in preferred responses. This work introduces a unique resource for studying human cognitive processing in LLM evaluation and suggests promising directions for incorporating eye-tracking data into alignment methods. The dataset and analysis code are publicly available.