AI Paper Atlas

Research map / Vision-language and multimodal / Robot action and navigation

QUAR-VLA: Vision-Language-Action Model for Quadruped Robots

Pengxiang Ding, Han Zhao, Wenjie Zhang, Wenxuan Song, Min Zhang, Siteng Huang, Ningxi Yang, Donglin Wang

ECCV 2024 · Vision-language and multimodal › Robot action and navigation · approach: Vision-language-action foundation models · 41 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use a VLA with a pretrained multimodal LLM instead of a visual encoder plus policy headQUART vs VC-1/CLIP · QUARD Crawlsuccess rate: 0 -> 0.32Improved
add more simulated data to co-training with 3K real samples (ablation)QUART · Real-world Go tosuccess (trials): 3/20 -> 13/20Improved
replace MLP policy head with decoder-only transformer on a VLM baseline (ablation)Aligned VLM+P · QUARD seen taskssuccess rate: Go to 0.43 -> 0.15; Crawl 0 -> 0No clear effect
swap Fuyu-8B backbone for LLaVa-7B (ablation)QUART · QUARD seen taskssuccess rate: Go to 0.60 -> 0.52; Unload 0.12 -> 0.16Mixed
multi-task training instead of single-task training (ablation)QUART · QUARD seen taskssuccess rate: Go avoid 0.19 -> 0.53; Crawl 0.25 -> 0.32Improved
multi-task instead of single-task training for baseline on Distinguish (ablation)VC-1 · QUARD Distinguishsuccess rate: 0.70 -> 0.46Hurt
use pretrained MLLM for unseen verbal instructionsQUART · QUARD unseen verbalsuccess rate: VC-1 0.19 -> 0.33; R3M 0 -> 0.33Improved
use unaligned vision encoder (R3M) without language alignmentR3M baseline · QUARD seen taskssuccess rate: Go to 0.58 on distinguish, 0 elsewhereHurt
switch navigation planner from D* to A* for data collectionQUARD simulation data · data qualitypath qualityImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy