Research map / Vision-language and multimodal / Multimodal LLMs
NAVER: A Neuro-Symbolic Compositional Automaton for Visual Grounding with Explicit Logic Reasoning
Zhixi Cai, Fucai Ke, Simindokht Jahangard, Maria Garcia de la Banda, Reza Haffari, Peter J. Stuckey, Hamid Rezatofighi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add DFA self-correction transitions (ablation) | NAVER · RefCOCO testA | accuracy: 57.4 -> 78.1 | Improved |
| add DFA self-correction transitions (ablation) | NAVER · Ref-Adv test | accuracy: 43.6 -> 75.4 | Improved |
| add logic generation and reasoning (ablation) | NAVER · RefCOCO+ testA | accuracy: 51.3 -> 62.0 | Improved |
| add answerer validation (ablation) | NAVER · RefCOCO testA | accuracy: 51.7 -> 57.4 | Improved |
| add answerer validation to DFA + logic (ablation) | NAVER · RefCOCOg test | accuracy: 75.8 -> 91.6 | Improved |
| add captioner before entity category extraction (ablation) | NAVER · RefCOCOg test | accuracy: 80.2 -> 91.6 | Improved |
| add captioner before entity category extraction (ablation) | NAVER · RefCOCO testA | accuracy: 91.6 -> 96.2 | Improved |
| use NAVER with GLIP-L instead of GLIP-L alone | GLIP-L · RefCOCO testA | accuracy: 55.0 -> 70.1 | Improved |
| wrap Florence2-L in NAVER | Florence2-L · RefCOCO testA | accuracy: 95.1 -> 96.2 | Improved |
| use NAVER instead of HYDRA (same models) | HYDRA · RefCOCOg test | runtime failure rate (%): 10.3 -> 00.3 | Improved |
| swap VLM among BLIP2-XXL, xGen-MM, LLaVA1.5-7B, InternVL2-8B (ablation) | NAVER · RefCOCO testA | accuracy: 94.5 -> 96.2 | No clear effect |
| swap LLM from GPT-4o Mini to DeepSeek-R1 70B (ablation) | NAVER · RefCOCOg detection | accuracy: 91.6 -> 85.8 | Hurt |
Later papers that compared with this method
1 later accepted paper reports NAVER as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- NePTune: A Neuro-Pythonic Framework for Tunable Compositional Reasoning on Vision-Language (ICLR 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- GLaMM — GLaMM: Pixel Grounding Large Multimodal Model (CVPR 2024) · better on 1 of 1 shared result
- ViperGPT — ViperGPT: Visual Inference via Python Execution for Reasoning (ICCV 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Investigating Compositional Challenges in Vision-Language Models for Visual Grounding (CVPR 2024 · similarity 0.90)
- NePTune: A Neuro-Pythonic Framework for Tunable Compositional Reasoning on Vision-Language (ICLR 2026 · similarity 0.90)
- Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning (AAAI 2026 · similarity 0.89)
- What Makes a Maze Look Like a Maze? (ICLR 2025 · similarity 0.89)
- No Labels, No Problem: Training Visual Reasoners with Multimodal Verifiers (ICLR 2026 · similarity 0.89)
- Multi-step Visual Reasoning with Visual Tokens Scaling and Verification (NeurIPS 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
