Generalizable Dense Reward for Long-Horizon Robotic Tasks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yong, Silong, Sheng, Stephen, Qi, Carl, Wang, Xiaojie, Sheehan, Evan, Shivaprasad, Anurag, Xie, Yaqi, Sycara, Katia, Dattatreya, Yesh
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915904091062272
author Yong, Silong
Sheng, Stephen
Qi, Carl
Wang, Xiaojie
Sheehan, Evan
Shivaprasad, Anurag
Xie, Yaqi
Sycara, Katia
Dattatreya, Yesh
author_facet Yong, Silong
Sheng, Stephen
Qi, Carl
Wang, Xiaojie
Sheehan, Evan
Shivaprasad, Anurag
Xie, Yaqi
Sycara, Katia
Dattatreya, Yesh
contents 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/
format Preprint
id arxiv_https___arxiv_org_abs_2604_00055
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Generalizable Dense Reward for Long-Horizon Robotic Tasks
Yong, Silong
Sheng, Stephen
Qi, Carl
Wang, Xiaojie
Sheehan, Evan
Shivaprasad, Anurag
Xie, Yaqi
Sycara, Katia
Dattatreya, Yesh
Robotics
Computer Vision and Pattern Recognition
Machine Learning
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/
title Generalizable Dense Reward for Long-Horizon Robotic Tasks
topic Robotics
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2604.00055