# Knowing the Answer Isn't Enough: Fixing Reasoning Path Failures in LVLMs

arXiv: https://arxiv.org/abs/2512.06258 · HTML: https://mlsanity.com/paper/2512.06258/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 36 of its 56 references that are in the map.
- First posted on arXiv 6 Dec 2025.

## Placed in

- LLM reasoning / [Multimodal reasoning](https://mlsanity.com/lr/multimodal-reasoning/)
- Multimodal LLMs / vision-language models / [Embodied and 3D vision-language models](https://mlsanity.com/mllm/embodied-and-3d-vision-language-models/)

## Builds on (most cited first, 30 of 36)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [GPT-4o System Card](https://mlsanity.com/paper/2410.21276/) (2024)
- [BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation](https://mlsanity.com/paper/2201.12086/) (2022)
- [LLaVA-OneVision: Easy Visual Task Transfer](https://mlsanity.com/paper/2408.03326/) (2024)
- [MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts](https://mlsanity.com/paper/2310.02255/) (2023)
- [ChartQA: A Benchmark for Question Answering about Charts with Visual and Logical Reasoning](https://mlsanity.com/paper/2203.10244/) (2022)
- [Are We on the Right Way for Evaluating Large Vision-Language Models?](https://mlsanity.com/paper/2403.20330/) (2024)
- [Gemini: A Family of Highly Capable Multimodal Models](https://mlsanity.com/paper/2312.11805/) (2023)
- [Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Models](https://mlsanity.com/paper/2503.06749/) (2025)
- [R1-Onevision: Advancing Generalized Multimodal Reasoning through Cross-Modal Formalization](https://mlsanity.com/paper/2503.10615/) (2025)
- [Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Models](https://mlsanity.com/paper/2501.09686/) (2025)
- [Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey](https://mlsanity.com/paper/2503.12605/) (2025)
- [VisualPRM: An Effective Process Reward Model for Multimodal Reasoning](https://mlsanity.com/paper/2503.10291/) (2025)
- [Math-LLaVA: Bootstrapping Mathematical Reasoning for Multimodal Large Language Models](https://mlsanity.com/paper/2406.17294/) (2024)
- [Boosting the Generalization and Reasoning of Vision Language Models with Curriculum Reinforcement Learning](https://mlsanity.com/paper/2503.07065/) (2025)
- [MultiMath: Bridging Visual and Mathematical Reasoning for Large Language Models](https://mlsanity.com/paper/2409.00147/) (2024)
- [Unlocking Multimodal Mathematical Reasoning via Process Reward Model](https://mlsanity.com/paper/2501.04686/) (2025)
- [Multi-SpatialMLLM: Multi-Frame Spatial Understanding with Multi-Modal Large Language Models](https://mlsanity.com/paper/2505.17015/) (2025)
- [ReAgent-V: A Reward-Driven Multi-Agent Framework for Video Understanding](https://mlsanity.com/paper/2506.01300/) (2025)
- [SophiaVL-R1: Reinforcing MLLMs Reasoning with Thinking Reward](https://mlsanity.com/paper/2505.17018/) (2025)
- [Online DPO: Online Direct Preference Optimization with Fast-Slow Chasing](https://mlsanity.com/paper/2406.05534/) (2024)
- [SpatialLLM: A Compound 3D-Informed Design towards Spatially-Intelligent Large Multimodal Models](https://mlsanity.com/paper/2505.00788/) (2025)
- [Look Back to Reason Forward: Revisitable Memory for Long-Context LLM Agents](https://mlsanity.com/paper/2509.23040/) (2025)
- [Position: Multimodal Large Language Models Can Significantly Advance Scientific Reasoning](https://mlsanity.com/paper/2502.02871/) (2025)
- [Stop Reasoning! When Multimodal LLM with Chain-of-Thought Reasoning Meets Adversarial Image](https://mlsanity.com/paper/2402.14899/) (2024)
- [MathGLM-Vision: Solving Mathematical Problems with Multi-Modal Large Language Model](https://mlsanity.com/paper/2409.13729/) (2024)
- [AtomThink: Multimodal Slow Thinking with Atomic Step Reasoning](https://mlsanity.com/paper/2411.11930/) (2024)
- [A Systematic Literature Review of Retrieval-Augmented Generation: Techniques, Metrics, and Challenges](https://mlsanity.com/paper/2508.06401/) (2025)

## Built on by (most cited first, 5 of 5)

- [AdaTooler-V: Adaptive Tool-Use for Images and Videos](https://mlsanity.com/paper/2512.16918/) (2025)
- [What Does Vision Tool-Use Reinforcement Learning Really Learn? Disentangling Tool-Induced and Intrinsic Effects for Crop-and-Zoom](https://mlsanity.com/paper/2602.01334/) (2026)
- [VLA-Thinker: Boosting Vision-Language-Action Models through Thinking-with-Image Reasoning](https://mlsanity.com/paper/2603.14523/) (2026)
- [V-Retrver: Evidence-Driven Agentic Reasoning for Universal Multimodal Retrieval](https://mlsanity.com/paper/2602.06034/) (2026)
- [When Thinking Hurts: Epistemic Signals in the Reasoning Chains of Visual Language Models](https://mlsanity.com/paper/2607.08059/) (2026)

## Abstract (arXiv)

We reveal a critical yet underexplored flaw in Large Vision-Language Models (LVLMs): even when these models know the correct answer, they frequently arrive there through incorrect reasoning paths. The core issue is not a lack of knowledge, but a path selection bias within the vast reasoning search space. Although LVLMs are often capable of sampling correct solution trajectories, they disproportionately favor unstable or logically inconsistent ones, leading to erratic and unreliable outcomes. The substantial disparity between Pass@K (with large K) and Pass@1 across numerous models provides compelling evidence that such failures primarily stem from misreasoning rather than ignorance. To systematically investigate and address this issue, we propose PSO (Path-Select Optimization), a two-stage post-training framework designed to enhance both the reasoning performance and stability of existing LVLMs. In the first stage, we employ Group Relative Policy Optimization (GRPO) with template and answer-based rewards to cultivate structured, step-by-step reasoning. In the second stage, we conduct online preference optimization, where the model samples reasoning paths from GRPO-generated data, self-evaluates them, and aligns itself toward the preferred trajectories. Incorrect or suboptimal paths are concurrently stored in a Negative Replay Memory (NRM) as hard negatives, which are periodically revisited to prevent the model from repeating prior mistakes and to facilitate continual reasoning refinement. Extensive experiments show that PSO effectively prunes invalid reasoning paths, substantially enhances reasoning accuracy (with 7.4% improvements on average), and yields more stable and consistent chains of thought. Our code will be available at https://github.com/aiming-lab/PSO.