# Compile Scene Graphs with Reinforcement Learning

arXiv: https://arxiv.org/abs/2504.13617 · HTML: https://mlsanity.com/paper/2504.13617/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- Placed in VLM benchmarks and visual reasoning, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 37 of its 45 references that are in the map.
- First posted on arXiv 18 Apr 2025.

## Placed in

- Multimodal LLMs / vision-language models / [VLM benchmarks and visual reasoning](https://mlsanity.com/mllm/vlm-benchmarks-and-visual-reasoning/)
- Deep reinforcement learning methods / [RL training of language and agent models](https://mlsanity.com/deeprl/rl-training-of-language-and-agent-models/)

## Builds on (most cited first, 30 of 37)

- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks](https://mlsanity.com/paper/1506.01497/) (2015)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [End-to-End Object Detection with Transformers](https://mlsanity.com/paper/2005.12872/) (2020)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [GPT-4o System Card](https://mlsanity.com/paper/2410.21276/) (2024)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context](https://mlsanity.com/paper/2403.05530/) (2024)
- [Scene Graph Generation by Iterative Message Passing](https://mlsanity.com/paper/1701.02426/) (2017)
- [Neural Motifs: Scene Graph Parsing with Global Context](https://mlsanity.com/paper/1711.06640/) (2017)
- [ConceptGraphs: Open-Vocabulary 3D Scene Graphs for Perception and Planning](https://mlsanity.com/paper/2309.16650/) (2023)
- [Unbiased Scene Graph Generation from Biased Training](https://mlsanity.com/paper/2002.11949/) (2020)
- [Learning to Compose Dynamic Tree Structures for Visual Contexts](https://mlsanity.com/paper/1812.01880/) (2018)
- [GPS-Net: Graph Property Sensing Network for Scene Graph Generation](https://mlsanity.com/paper/2003.12962/) (2020)
- [Knowledge-Embedded Routing Network for Scene Graph Generation](https://mlsanity.com/paper/1903.03326/) (2019)
- [Scene Graph Generation from Objects, Phrases and Region Captions](https://mlsanity.com/paper/1707.09700/) (2017)
- [Panoptic Scene Graph Generation](https://mlsanity.com/paper/2207.11247/) (2022)
- [SG-Nav: Online 3D Scene Graph Prompting for LLM-based Zero-shot Object Navigation](https://mlsanity.com/paper/2410.08189/) (2024)
- [SGTR: End-to-end Scene Graph Generation with Transformer](https://mlsanity.com/paper/2112.12970/) (2021)
- [RelTR: Relation Transformer for Scene Graph Generation](https://mlsanity.com/paper/2201.11460/) (2022)
- [The All-Seeing Project V2: Towards General Relation Comprehension of the Open World](https://mlsanity.com/paper/2402.19474/) (2024)
- [Hierarchical Planning for Long-Horizon Manipulation with Geometric and Symbolic Scene Graphs](https://mlsanity.com/paper/2012.07277/) (2020)
- [Revisiting the Test-Time Scaling of o1-like Models: Do they Truly Possess Test-Time Scaling Capabilities?](https://mlsanity.com/paper/2502.12215/) (2025)
- [Learning to Generate Scene Graph from Natural Language Supervision](https://mlsanity.com/paper/2109.02227/) (2021)
- [Open-Vocabulary Functional 3D Scene Graphs for Real-World Indoor Spaces](https://mlsanity.com/paper/2503.19199/) (2025)
- [4D-OR: Semantic Scene Graphs for OR Domain Modeling](https://mlsanity.com/paper/2203.11937/) (2022)
- [From Pixels to Graphs: Open-Vocabulary Scene Graph Generation with Vision-Language Models](https://mlsanity.com/paper/2404.00906/) (2024)

## Built on by (most cited first, 11 of 11)

- [The Landscape of Agentic Reinforcement Learning for LLMs: A Survey](https://mlsanity.com/paper/2509.02547/) (2025)
- [3D-R1: Enhancing Reasoning in 3D VLMs for Unified Scene Understanding](https://mlsanity.com/paper/2507.23478/) (2025)
- [Reinforcement Learning Meets Large Language Models: A Survey of Advancements and Applications Across the LLM Lifecycle](https://mlsanity.com/paper/2509.16679/) (2025)
- [WeThink: Toward General-purpose Vision-Language Reasoning via Reinforcement Learning](https://mlsanity.com/paper/2506.07905/) (2025)
- [SpatialThinker: Reinforcing Scene Graph-Grounded Spatial Reasoning via Dense Rewards](https://mlsanity.com/paper/2511.07403/) (2025)
- [Reinforcement Fine-Tuning Powers Reasoning Capability of Multimodal Large Language Models](https://mlsanity.com/paper/2505.18536/) (2025)
- [Relation-R1: Progressively Cognitive Chain-of-Thought Guided Reinforcement Learning for Unified Relation Comprehension](https://mlsanity.com/paper/2504.14642/) (2025)
- [Video-LMM Post-Training: A Deep Dive into Video Reasoning with Large Multimodal Models](https://mlsanity.com/paper/2510.05034/) (2025)
- [SGG-Rrm 3: From Next-Token Prediction to End-to-End Unbiased Scene Graph Generation](https://mlsanity.com/paper/2603.07961/) (2026)
- [SenBen: Sensitive Scene Graphs for Explainable Content Moderation](https://mlsanity.com/paper/2604.08819/) (2026)
- [OracleAnalyser: Analysing Implicit Semantics of Oracle Bone Scripts through MLLMs with Post-training](https://mlsanity.com/paper/2606.25906/) (2026)

## Abstract (arXiv)

Next-token prediction is the fundamental principle for training large language models (LLMs), and reinforcement learning (RL) further enhances their reasoning performance. As an effective way to model language, image, video, and other modalities, the use of LLMs for end-to-end extraction of structured visual representations, such as scene graphs, remains underexplored. It requires the model to accurately produce a set of objects and relationship triplets, rather than generating text token by token. To achieve this, we introduce R1-SGG, a multimodal LLM (M-LLM) initially trained via supervised fine-tuning (SFT) on the scene graph dataset and subsequently refined using reinforcement learning to enhance its ability to generate scene graphs in an end-to-end manner. The SFT follows a conventional prompt-response paradigm, while RL requires the design of effective reward signals. We design a set of graph-centric rewards, including three recall-based variants -- Hard Recall, Hard Recall+Relax, and Soft Recall -- which evaluate semantic and spatial alignment between predictions and ground truth at the object and relation levels. A format consistency reward further ensures that outputs follow the expected structural schema. Extensive experiments on the VG150 and PSG benchmarks show that R1-SGG substantially reduces failure rates and achieves strong performance in Recall and mean Recall, surpassing traditional SGG models and existing multimodal language models. Our code is available at https://github.com/gpt4vision/R1-SGG