onPanda: Efficient Annotation of On-Policy Alignment Data for LLMs and Agents via Token-Level Correction
We present onPanda, an interactive tool for efficiently annotating LLM alignment data and agent trajectories. onPanda adopts token-level correction as its core interaction: while reading a model response, the annotator locates the first inappropriate token and either picks a substitute from the model's candidate tokens or types the correct text via free-form editing. The system
Reading in Plain English mode — toggle below to switch
We present onPanda, an interactive tool for efficiently annotating LLM alignment data and agent trajectories. onPanda adopts token-level correction as its core interaction: while reading a model response, the annotator locates the first inappropriate token and either picks a substitute from the model's candidate tokens or types the correct text via free-form editing. The system
Efficiently annotating alignment data is crucial for improving LLM and agent performance and safety. This tool significantly reduces the time and cost associated with generating high-quality human feedback, directly impacting the development cycle of advanced AI systems.