Driving · 2026-10-22

CausGround: Causal Interaction Grounding for Adaptive Embodied Planning

Yijun Shen, Diqi He, Jinbang Guo, Wanyue Zhang, Jiaming Mi, Huan Yang, Zhongyuan Wang, Zhiwei Yu, Kai Zhang

arXiv:2610.29702PDF

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

CausGround is a non-parametric causal interaction grounding framework for embodied planning. It extracts action preconditions and effects from experience into a factorized world-state representation (object attributes + inter-object relations). At planning time it grounds a task-relevant causal graph to enable backward reasoning over preconditions and forward reasoning over effects, improving planning and interaction reasoning across VirtualHome, RoboCasa, SWITCH, and real-robot manipulation without parameter updates.

Why it matters

The authors target reliable physical interaction with MLLMs by explicitly modeling when actions are executable (preconditions) and how they change the world (effects) in an environment-dependent way. By turning interaction experience into reusable action-centered causal knowledge and grounding it at inference time, the approach aims to improve long-horizon planning and action reasoning without costly environment-specific training or parameter updates.

Method

  • Construct a factorized world-state representation (object attributes as key-value factors and typed inter-object relations) so observations and causal knowledge share a common schema for matching preconditions/effects.
  • Extract reusable action-centered causal knowledge by identifying precondition factors that distinguish successful from failed executions and effect state changes that consistently follow execution; organize generic vs instance-specific knowledge.
  • At task time, retrieve and ground task-relevant causal dependencies into a local causal graph; support backward reasoning over preconditions and forward reasoning over effects for multi-step planning with a frozen planner.

Limitation

The paper text does not state a specific limitation of the method.

Abstract (from arXiv)

Multimodal large language models (MLLMs) show promising reasoning capabilities, but reliable physical interaction requires planning that accounts for when actions are executable and how they change the world. These preconditions and effects are environment-dependent and often not captured by MLLMs, making adaptation challenging. Yet existing approaches either require costly environment-specific training or rely on unstructured interaction experience that is difficult to reuse. Therefore, we introduce CausGround, a non-parametric causal interaction grounding framework that distills interaction experience into reusable action-centered causal knowledge. CausGround identifies action preconditions and effects and organizes them in a factorized world-state representation capturing object attributes and inter-object relations. At planning time, task-relevant causal dependencies are grounded in current observations to support backward reasoning over preconditions and forward reasoning over effects for action selection and multi-step planning. We evaluate CausGround on simulated high-level planning, interaction reasoning, and real-world robot manipulation. On adapted VirtualHome and RoboCasa benchmarks, CausGround improves vision-based high-level planning across MLLM backbones, increasing average task success from 24.9% to 53.3% on VirtualHome and closed-loop planning success from 43.0% to 57.0% on RoboCasa. On SWITCH, average interaction reasoning accuracy increases from 25.8% to 31.9%, indicating improved action selection, recovery, and verification. Moreover, real-world experiments show that CausGround reasons backward from MLLM-interpreted goals to derive executable high-level action sequences for sequential manipulation. These results demonstrate that reusable action preconditions and effects support long-horizon embodied planning across diverse environments without parameter updates.

Related papers

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-10-21 · Indexed by Pub-AI: 2026-10-22 UTC

preprint

CausGround: Causal Interaction Grounding for Adaptive Embodied Planning

Yijun Shen, Diqi He, Jinbang Guo, Wanyue Zhang, Jiaming Mi, Huan Yang, Zhongyuan Wang, Zhiwei Yu, Kai Zhang

Preprint posted: 2026-10-21 · Original record ↗

arxiv
2610.29702
Explicit source versions (1)
  • v1 · 2026-10-21
  1. Preprint available

    Source date: 2026-10-21 · First detected: 2026-10-25 UTC · Historical baseline

    Source evidence (1)
Abstract · arxiv

Multimodal large language models (MLLMs) show promising reasoning capabilities, but reliable physical interaction requires planning that accounts for when actions are executable and how they change the world. These preconditions and effects are environment-dependent and often not captured by MLLMs, making adaptation challenging. Yet existing approaches either require costly environment-specific training or rely on unstructured interaction experience that is difficult to reuse. Therefore, we introduce CausGround, a non-parametric causal interaction grounding framework that distills interaction experience into reusable action-centered causal knowledge. CausGround identifies action preconditions and effects and organizes them in a factorized world-state representation capturing object attributes and inter-object relations. At planning time, task-relevant causal dependencies are grounded in current observations to support backward reasoning over preconditions and forward reasoning over effects for action selection and multi-step planning. We evaluate CausGround on simulated high-level planning, interaction reasoning, and real-world robot manipulation. On adapted VirtualHome and RoboCasa benchmarks, CausGround improves vision-based high-level planning across MLLM backbones, increasing average task success from 24.9% to 53.3% on VirtualHome and closed-loop planning success from 43.0% to 57.0% on RoboCasa. On SWITCH, average interaction reasoning accuracy increases from 25.8% to 31.9%, indicating improved action selection, recovery, and verification. Moreover, real-world experiments show that CausGround reasons backward from MLLM-interpreted goals to derive executable high-level action sequences for sequential manipulation. These results demonstrate that reusable action preconditions and effects support long-horizon embodied planning across diverse environments without parameter updates.