CoSPlan: Corrective Sequential Planning via Scene Graph Incremental Updates

TL;DR

This paper introduces CoSPlan, a benchmark for evaluating vision-language models in error-prone sequential planning tasks, and proposes Scene Graph Incremental updates (SGI), a training-free method that improves reasoning performance by adding intermediate steps.

Contribution

The paper presents a new benchmark for vision-language models in sequential planning with errors and introduces SGI, a novel, training-free approach that enhances reasoning accuracy.

Findings

VLMs struggle with error detection and correction in CoSPlan.

SGI improves VLM performance by an average of 5.2%.

SGI generalizes to traditional planning and VQA tasks.

Abstract

Large-scale Vision-Language Models (VLMs) exhibit impressive complex reasoning capabilities but remain largely unexplored in visual sequential planning, i.e., executing multi-step actions towards a goal. Additionally, practical sequential planning often involves non-optimal (erroneous) steps, challenging VLMs to detect and correct such steps. We propose Corrective Sequential Planning Benchmark (CoSPlan) to evaluate VLMs in error-prone, vision-based sequential planning tasks across 4 domains: maze navigation, block rearrangement, image reconstruction,and object reorganization. CoSPlan assesses two key abilities: Error Detection (identifying non-optimal action) and Step Completion (correcting and completing action sequences to reach the goal). Despite using state-of-the-art reasoning techniques such as Chain-of-Thought and Scene Graphs, VLMs (e.g. Intern-VLM and Qwen2) struggle on CoSPlan, failing to leverage contextual cues to reach goals. Addressing this, we propose a novel training-free method, Scene Graph Incremental updates (SGI), which introduces intermediate reasoning steps between the initial and goal states. SGI helps VLMs reason about sequences, yielding an average performance gain of 5.2%. In addition to enhancing reliability in corrective sequential planning, SGI generalizes to traditional planning tasks such as Plan-Bench and VQA. Project Page : https://shroglck.github.io/cos_plan/