AI Paper Atlas

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

DISCO: Embodied Navigation and Interaction via Differentiable Scene Semantics and Dual-level Control

Xinyu Xu, Shengcheng Luo, Yanchao Yang, Yong-Lu Li, Cewu Lu

ECCV 2024 · Vision-language and multimodal › Robot action and navigation · approach: Embodied multimodal LLM agents · 7 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
replace cell-based scene map with differentiable scene representation (ablation)DISCO · ALFRED valid unseenSR: 42.7 -> 55.0Improved
add interactive affordances (openable, affordance-augmented localization) (ablation)DISCO · ALFRED valid unseenSR: 51.3 -> 55.0Improved
add navigability affordance instead of point-obstacle navigation (ablation)DISCO · ALFRED valid unseenSR: 46.0 -> 55.0Improved
add map-based coarse control (ablation)DISCO · ALFRED valid unseenSR: 9.1 -> 55.0Improved
add neural fine control after coarse navigation (ablation)DISCO · ALFRED valid unseenSR: 51.7 -> 55.0Improved
add low-level step-by-step instructions (ablation)DISCO · ALFRED valid unseenSR: 55.0 -> 59.1Improved
use ground-truth language parsing (ablation)DISCO · ALFRED valid unseenSR: 55.0 -> 64.1Improved
use ground-truth perception (ablation)DISCO · ALFRED valid unseenSR: 55.0 -> 66.8Improved
use ground-truth perception and language together (ablation)DISCO · ALFRED valid unseenSR: 55.0 -> 79.6Improved
use DISCO without step-by-step instructions versus prior methods that use themALFRED agents · ALFRED test unseenSR: 46.1 (CAPEAM with instr.) -> 54.7Improved
use DISCO instead of PrompterALFRED agents · ALFRED test unseenPLWSR (efficiency): 20.8 -> 36.5Improved

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