Large language models (LLMs) have been applied to a wide range of tasks, including text summarization, web navigation, and chat- bots. They have benefitted from supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF) following an un- supervised pretraining. These datasets can be difficult to collect, limited in scope, and vary in sample quality. Additionally, datasets can vary extensively in supervision format, from numer- ical to binary as well as multi-dimensional with many different values. We present a framework for fine-tuning LLMs using heterogeneous feed- back, which has two main components. First, we combine the heterogeneous feedback data into a single supervision format, compatible with methods like SFT and RLHF. Next, given this unified feedback dataset, we extract a high- quality and diverse subset to obtain perfor- mance increases potentially exceeding the full dataset. We conduct extensive experiments to understand the effectiveness of these tech- niques for incorporating heterogeneous feed- back, and demonstrate improvements from us- ing a high-quality and diverse subset of the data. We find that our framework is able to improve models in multiple areas simultaneously, such as in instruction following and bias reduction.

Agents powered by large language models (LLMs) inherit important limitations, such as the restricted context length, dependency on human-engineered exemplars (e.g., for task decomposition), and insufficient generalization. To address these challenges, we propose RaDA, a novel planning method for Web agents that does not require manual exemplars, efficiently leverages the LLMs’ context, and enhances generalization. RaDA disentangles planning into two stages: for a new given task, during Retrieval-augmented Task Decomposition (RaD), it decomposes tasks into high-level subtasks; next, during Retrieval-augmented Action Generation (RaA), it traverses the trajectory obtained with RaD to iteratively synthesize actions based on dynamically retrieved exemplars. We compare RaDA with strong baselines covering a broad space of design choices, using both GPT-3.5 and GPT-4 as backbones; and we find consistent improvements over previous SOTA in two challenging benchmarks, CompWoB and Mind2Web, covering settings with different complexities. We show the contributions of RaDA via ablation studies and qualitative analysis; and we discuss the structural benefits of our more compositional design.