Research map / Image generation / Autoregressive and latent generation
Comprehension-Guided Referring Expressions
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| rerank sampled expressions using comprehension scores rather than choosing by perplexity | Referring expression generator · RefCOCO Test A | comprehension accuracy: 78.38% -> 97.23% | Helped |
| rerank sampled expressions using comprehension scores rather than choosing by perplexity | Referring expression generator · RefCOCO Test B | comprehension accuracy: 73.08% -> 94.96% | Helped |
| rerank sampled expressions using comprehension scores rather than choosing by perplexity | Referring expression generator · RefCOCO+ Test A | comprehension accuracy: 62.45% -> 77.32% | Helped |
| rerank sampled expressions using comprehension scores rather than choosing by perplexity | Referring expression generator · RefCOCO+ Test B | comprehension accuracy: 47.86% -> 67.65% | Helped |
| add comprehension-loss guidance during training | Referring expression generator · RefCOCO and RefCOCO+ | comprehension accuracy: 80.14%; 75.44%; 68.54%; 55.87% | Mixed |
| use modified scheduled sampling | Referring expression generator · RefCOCO and RefCOCO+ | comprehension accuracy: 79.94%; 75.93%; 69.41%; 55.59% | Mixed |
| use stochastic mixed incremental training | Referring expression generator · RefCOCO and RefCOCO+ | comprehension accuracy: 79.99%; 75.60%; 69.05%; 54.71% | Mixed |
| initialize comprehension-model word embeddings with pretrained word2vec (ablation) | Comprehension model · RefCLEF Test | accuracy: 31.25% -> 31.85% | Helped |
| initialize comprehension-model word embeddings with pretrained word2vec (ablation) | Comprehension model · RefCOCOg Val, GT | accuracy: 63.39% -> 65.36% | Helped |
| use generate-and-rerank instead of greedy single-output generation | Referring expression generation · RefCOCO+ Test B | human evaluation hit rate: 25% -> 47% | Mixed |
| use comprehension-guided generation rather than MMI | Referring expression generation · RefCOCO Test A and Test B | human evaluation hit rate: 53% -> 66%; 61% -> 75% | Helped |
| use training-by-proxy SMIXEC rather than MMI | Referring expression generation · RefCOCO+ Test B | human evaluation hit rate: 35% -> 25% | Hurt |
Closest papers by meaning
- Revisiting Counterfactual Problems in Referring Expression Comprehension (CVPR 2024 · similarity 0.88)
- Ref-Adv: Exploring MLLM Visual Reasoning in Referring Expression Tasks (ICLR 2026 · similarity 0.88)
- DiffusionREC: Diffusion Model with Adaptive Condition for Referring Expression Comprehension (AAAI 2025 · similarity 0.87)
- RefDetector: A Simple Yet Effective Matching-based Method for Referring Expression Comprehension (AAAI 2025 · similarity 0.87)
- Hierarchical Alignment-enhanced Adaptive Grounding Network for Generalized Referring Expression Comprehension (AAAI 2025 · similarity 0.87)
- Rethinking Two-Stage Referring Expression Comprehension: A Novel Grounding and Segmentation Method Modulated by Point (AAAI 2024 · similarity 0.86)
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.