# STAR: Mitigating Cascading Errors in Spatial Reasoning via Turn-point Alignment and Segment-level DPO

arXiv: https://arxiv.org/abs/2604.00558 · HTML: https://mlsanity.com/paper/2604.00558/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning.
- Builds on 18 of its 21 references that are in the map.
- First posted on arXiv 1 Apr 2026.

## Placed in

- LLM reasoning / [Spatial and embodied reasoning](https://mlsanity.com/lr/spatial-and-embodied-reasoning/)
- LLM alignment / RLHF / preference optimisation / [Diffusion and multimodal model alignment](https://mlsanity.com/alignment/diffusion-and-multimodal-model-alignment/)

## Builds on (most cited first, 18 of 18)

- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training](https://mlsanity.com/paper/2501.17161/) (2025)
- [Visual Sketchpad: Sketching as a Visual Chain of Thought for Multimodal Language Models](https://mlsanity.com/paper/2406.09403/) (2024)
- [Imagine while Reasoning in Space: Multimodal Visualization-of-Thought](https://mlsanity.com/paper/2501.07542/) (2025)
- [SpartQA: : A Textual Question Answering Benchmark for Spatial Reasoning](https://mlsanity.com/paper/2104.05832/) (2021)
- [StepGame: A New Benchmark for Robust Multi-Hop Spatial Reasoning in Texts](https://mlsanity.com/paper/2204.08292/) (2022)
- [Evaluating Spatial Understanding of Large Language Models](https://mlsanity.com/paper/2310.14540/) (2023)
- [Mind's Eye of LLMs: Visualization-of-Thought Elicits Spatial Reasoning in Large Language Models](https://mlsanity.com/paper/2404.03622/) (2024)
- [Coupling Large Language Models with Logic Programming for Robust and General Reasoning from Text](https://mlsanity.com/paper/2307.07696/) (2023)
- [VideoGen-of-Thought: Step-by-step generating multi-shot video with minimal manual intervention](https://mlsanity.com/paper/2503.15138/) (2025)
- [AlphaMaze: Enhancing Large Language Models' Spatial Intelligence via GRPO](https://mlsanity.com/paper/2502.14669/) (2025)
- [Mind with Eyes: from Language Reasoning to Multimodal Reasoning](https://mlsanity.com/paper/2503.18071/) (2025)
- [DependEval: Benchmarking LLMs for Repository Dependency Understanding](https://mlsanity.com/paper/2503.06689/) (2025)
- [Recent Advances in Multimodal Affective Computing: An NLP Perspective](https://mlsanity.com/paper/2409.07388/) (2024)
- [Text-Visual Semantic Constrained AI-Generated Image Quality Assessment](https://mlsanity.com/paper/2507.10432/) (2025)
- [EvoEmpirBench: Dynamic Spatial Reasoning with Agent-ExpVer](https://mlsanity.com/paper/2509.12718/) (2025)

## Built on by (most cited first, 4 of 4)

- [Affordance Agent Harness: Verification-Gated Skill Orchestration](https://mlsanity.com/paper/2605.00663/) (2026)
- [Find, Fix, Reason: Context Repair for Video Reasoning](https://mlsanity.com/paper/2604.16243/) (2026)
- [No Place to Hide: Benchmarking Video Hallucination with Background-Controlled Pairs](https://mlsanity.com/paper/2606.31933/) (2026)
- [Are Frontier VLM Agents Ready to Be Robot Generalists? An Empirical Study with the Embodied Agent Arena](https://mlsanity.com/paper/2610.00854/) (2026)

## Abstract (arXiv)

Structured spatial navigation is a core benchmark for Large Language Models (LLMs) spatial reasoning. Existing paradigms like Visualization-of-Thought (VoT) are prone to cascading errors in complex topologies. To solve this, we propose STAR, a two-stage framework grounded on topological anchors, and introduce the RedMaze-23K dataset with human-inspired turnpoint annotations. The first stage uses supervised fine-tuning to help models internalize spatial semantics and prune redundant paths. The second adopts Spatial-aware Segment-level Direct Preference Optimization (SDPO) to refine self-correction in long-horizon navigation. Experiments show STAR achieves state-of-the-art performance among open-source models: its 32B variant outperforms DeepSeek-V3 (29.27% vs. 25.00%) and reaches 82.4% of GPT-4's performance.