OASST-ETC Dataset: Alignment Signals from Eye-tracking Analysis of LLM Responses

Abstract

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.

This paper has been accepted to ACM ETRA 2025 and published on PACMHCI

References in corpus (12)

• Llama 2: Open Foundation and Fine-Tuned Chat Models
• RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI Feedback
• Textbooks Are All You Need II: phi-1.5 technical report
• RRHF: Rank Responses to Align Language Models with Human Feedback without tears
• Advancing NLP with Cognitive Language Processing Signals
• RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment
• EyeTrans: Merging Human and Machine Attention for Neural Code Summarization
• Reporting Eye-Tracking Data Quality: Towards a New Standard
• Advancing LLM Reasoning Generalists with Preference Trees
• RLCD: Reinforcement Learning from Contrastive Distillation for Language Model Alignment
• HRLAIF: Improvements in Helpfulness and Harmlessness in Open-domain Reinforcement Learning From AI Feedback
• Seeing Eye to AI: Human Alignment via Gaze-Based Response Rewards for Large Language Models