Research map / Vision-language and multimodal / Human and embodied modeling
Beyond Unidirectional Bias: Reciprocal Perspective Calibration in Scene Graph Generation
Haifeng Zhao, Wenbo Zhao, Xuemei Luo, Lei-Lei Ma, Dengdi Sun
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 Additive Injection (RPCAdd) (ablation) | CLIPM · VG PredCls | M@100: 42.5 -> 44.2 | Improved |
| add Stochastic Replacement (RPCRep) (ablation) | CLIPM · VG PredCls | M@100: 42.5 -> 43.9 | Improved |
| add hypernym-guided prompt (HGP) (ablation) | CLIPM · VG PredCls | M@100: 42.5 -> 44.7 | Improved |
| add RPCAdd on top of HGP (ablation) | CLIPM+HGP · VG PredCls | M@100: 44.7 -> 45.7 | Improved |
| add RPCRep on top of HGP (ablation) | CLIPM+HGP · VG PredCls | M@100: 44.7 -> 45.2 | Improved |
| restrict Add to TAIL and SEMANTIC predicates instead of all (ablation) | CLIPM+RPCAdd · VG PredCls | M@100: 43.2 -> 44.2 | Improved |
| restrict Rep to HEAD and GEOMETRIC predicates instead of all (ablation) | CLIPM+RPCRep · VG PredCls | M@100: 42.8 -> 43.9 | Improved |
| add thinking strategy with MPIR constraints to LLM prompts (ablation) | CLIPM+HGP (Gemini-pro) · VG PredCls | M@100: 44.2 -> 44.7 | Improved |
| swap LLM used for hypernym mapping (ablation) | CLIPM+HGP · VG PredCls | M@100: 43.1 / 43.8 / 43.9 / 44.1 / 44.2 | No clear effect |
| add RPC*Add to PE-Net | PE-Net · VG PredCls | R@100 and mR@100: R@100 61.9 -> 59.6; mR@100 37.0 -> 42.9 | Mixed |
| add RPC*Add to Motifs | Motifs · VG PredCls (inverse test set) | MI@100: 4.4 -> 7.7 | Improved |
| add RPCAdd+HGP to CLIPM | CLIPM · VG PredCls (inverse test set) | MI@100: 32.8 -> 36.4 | Improved |
Closest papers by meaning
- Human-Inspired Scene Understanding: A Grounded Cognition Method for Unbiased Scene Graph Generation (TPAMI 2025 · similarity 0.88)
- Universal Scene Graph Generation (CVPR 2025 · similarity 0.87)
- Bridging Visual and Textual Semantics: Towards Consistency for Unbiased Scene Graph Generation (TPAMI 2024 · similarity 0.87)
- Relation-aware Hierarchical Prompt for Open-vocabulary Scene Graph Generation (AAAI 2025 · similarity 0.86)
- IR3D-Bench: Evaluating Vision-Language Model Scene Understanding as Agentic Inverse Rendering (NeurIPS 2025 · similarity 0.86)
- Compositional Scene Understanding through Inverse Generative Modeling (ICML 2025 · similarity 0.85)
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.
