AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

DualWorldBench: Can Agents Plan Deliveries across Symbolic and Grounded Worlds?

Yang Chen, Bin Wen, Danyang Peng, Hong-Jie You, Ziqiao Shang, Lan-Zhe Guo

NeurIPS 2026 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Embodied agent learning · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Introduces DualWorldBench, a benchmark for evaluating whether multimodal agents can plan multi-goal deliveries under temporal, physical, and multimodal constraints across symbolic and grounded worlds.

Problem

Existing evaluations of multimodal agents rely on simplified tasks with limited combinatorial structure, so it is unclear whether agents can integrate perception and reasoning for multi-goal planning under temporal, physical, and multimodal constraints.

Approach

A unified benchmark pairing symbolic and grounded versions of delivery planning tasks to test combinatorial multi-goal planning with perception and reasoning integration.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use frontier proprietary LLM instead of open-source LLMLLM planners · TextWorld (4 settings)average reward: 166.3 -> 385.0Improved
add online task release (L2) with heterogeneous tasksopen-source LLMs · TextWorld M-L2reward vs Offline DP: 23.8% of Offline-DP rewardHurt
add online task release (L2)proprietary LLMs · TextWorld A-L2reward vs Offline DP: 96.1% of Offline-DP reward (278.5 vs. 289.7)Hurt
increase task count to 15Llama-3-8B · TextWorldreward: 0.00 at 15 tasksHurt
use image-only input instead of textweaker VLMs (Claude-Haiku-4.5, GPT-5.4-mini) · SimWorldfeasibility: 0.905 -> 0.175 (Haiku)Hurt
use image-only input instead of textGPT-5.6-Sol · SimWorldfeasibility: 0.968 -> 0.989No clear effect
use multimodal instead of text inputQwen3-VL-8B · SimWorldcompletion: 0.544 -> 0.974Improved
increase task count from 5 to 10Claude-Opus-5 multimodal · SimWorldfeasibility: 0.730 -> 0.397Hurt
increase task countSimWorld models · SimWorldnormalized costHurt
evaluate perception/topology vs planningGPT-5.4-mini · DualWorld-VQAaccuracy: 55.5% / 55.9% -> 6.3%Hurt

Experimental setup

Benchmarks: DualWorldBench

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