Saved in:
Bibliographic Details
Main Authors: Roy, Simon, Barbeau, Samuel, Beltrame, Giovanni, Desrosiers, Christian, Thome, Nicolas
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.20675
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909974912827392
author Roy, Simon
Barbeau, Samuel
Beltrame, Giovanni
Desrosiers, Christian
Thome, Nicolas
author_facet Roy, Simon
Barbeau, Samuel
Beltrame, Giovanni
Desrosiers, Christian
Thome, Nicolas
contents Learning generalizable reward functions is a core challenge in embodied intelligence. Recent work leverages contrastive vision language models (VLMs) to obtain dense, domain-agnostic rewards without human supervision. These methods adapt VLMs into reward models through increasingly complex learning objectives, yet meaningful comparison remains difficult due to differences in training data, architectures, and evaluation settings. In this work, we isolate the impact of the learning objective by evaluating recent VLM-based reward models under a unified framework with identical backbones, finetuning data, and evaluation environments. Using Meta-World tasks, we assess modeling accuracy by measuring consistency with ground truth reward and correlation with expert progress. Remarkably, we show that a simple triplet loss outperforms state-of-the-art methods, suggesting that much of the improvements in recent approaches could be attributed to differences in data and architectures.
format Preprint
id arxiv_https___arxiv_org_abs_2512_20675
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Revisiting the Learning Objectives of Vision-Language Reward Models
Roy, Simon
Barbeau, Samuel
Beltrame, Giovanni
Desrosiers, Christian
Thome, Nicolas
Machine Learning
Artificial Intelligence
Learning generalizable reward functions is a core challenge in embodied intelligence. Recent work leverages contrastive vision language models (VLMs) to obtain dense, domain-agnostic rewards without human supervision. These methods adapt VLMs into reward models through increasingly complex learning objectives, yet meaningful comparison remains difficult due to differences in training data, architectures, and evaluation settings. In this work, we isolate the impact of the learning objective by evaluating recent VLM-based reward models under a unified framework with identical backbones, finetuning data, and evaluation environments. Using Meta-World tasks, we assess modeling accuracy by measuring consistency with ground truth reward and correlation with expert progress. Remarkably, we show that a simple triplet loss outperforms state-of-the-art methods, suggesting that much of the improvements in recent approaches could be attributed to differences in data and architectures.
title Revisiting the Learning Objectives of Vision-Language Reward Models
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2512.20675