Generalizable Dense Reward for Long-Horizon Robotic Tasks

TL;DR

VLLR introduces a dense reward framework combining language and vision models for improved long-horizon robotic task performance, reducing manual reward engineering and enhancing success rates across diverse tasks.

Contribution

The paper presents VLLR, a novel dense reward method leveraging LLMs and VLMs for task decomposition and progress estimation, improving RL finetuning efficiency and success on complex tasks.

Findings

VLLR achieves up to 56% success rate gains on the CHORES benchmark.

VLLR improves in-distribution task success by up to 5% over state-of-the-art methods.

VLLR enhances out-of-distribution task success by up to 10%.

Abstract

Existing robotic foundation policies are trained primarily via large-scale imitation learning. While such models demonstrate strong capabilities, they often struggle with long-horizon tasks due to distribution shift and error accumulation. While reinforcement learning (RL) can finetune these models, it cannot work well across diverse tasks without manual reward engineering. We propose VLLR, a dense reward framework combining (1) an extrinsic reward from Large Language Models (LLMs) and Vision-Language Models (VLMs) for task progress recognition, and (2) an intrinsic reward based on policy self-certainty. VLLR uses LLMs to decompose tasks into verifiable subtasks and then VLMs to estimate progress to initialize the value function for a brief warm-up phase, avoiding prohibitive inference cost during full training; and self-certainty provides per-step intrinsic guidance throughout PPO finetuning. Ablation studies reveal complementary benefits: VLM-based value initialization primarily improves task completion efficiency, while self-certainty primarily enhances success rates, particularly on out-of-distribution tasks. On the CHORES benchmark covering mobile manipulation and navigation, VLLR achieves up to 56% absolute success rate gains over the pretrained policy, up to 5% gains over state-of-the-art RL finetuning methods on in-distribution tasks, and up to $10\%$ gains on out-of-distribution tasks, all without manual reward engineering. Additional visualizations can be found in https://silongyong.github.io/vllr_project_page/