Research map / Vision-language and multimodal / Visual question answering
Grounding Language Models for Visual Entity Recognition
Zilin Xiao, Ming Gong, Paola Cascante-Bonilla, Xingyao Zhang, Jie Wu, Vicente Ordonez
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 |
|---|---|---|---|
| use retrieval-augmented constrained decoding MLLM (AutoVER) vs PaLI-17B | AutoVER-7B · OVEN-Wiki val Entity seen | accuracy: 30.6 -> 61.5 | Improved |
| add retrieval to constrained decoding (ablation) | AutoVER-7B-0.1 · OVEN-Wiki val Entity unseen | accuracy: 0.6 -> 19.0 | Improved |
| add retrieval to constrained decoding (ablation) | AutoVER-7B-0.1 · OVEN-Wiki val Entity seen | accuracy: 50.7 -> 48.9 | Hurt |
| add constrained decoding over the trie (ablation) | AutoVER-7B-0.1 · OVEN-Wiki val Entity seen | accuracy: 46.8 -> 50.7 | Improved |
| use LoRA (rank 128) instead of full tuning (ablation) | AutoVER-7B-0.1 · OVEN-Wiki val Entity seen | accuracy: 48.9 -> 43.5 | Hurt |
| replace generative framework with a [CLS] classifier over 20,549 entities (ablation) | AutoVER-7B-0.1 · OVEN-Wiki val Entity split | harmonic mean: 27.4 -> 0.2 | Hurt |
| scale base MLLM from 7B to 13B | AutoVER · OVEN-Wiki val overall | harmonic mean: 36.8 -> 39.2 | Improved |
| use AutoVER vs zero-shot GPT-4V on Query unseen | AutoVER-13B · OVEN-Wiki val Query unseen | accuracy: 32.3 vs 52.7 | Hurt |
| apply retrieval-constrained decoding zero-shot to A-OKVQA-Ent vs fine-tuned LLaVA-v1.5 | AutoVER-7B · A-OKVQA-Ent multi-choice | overall accuracy: 62.8 vs 72.8 | Hurt |
| apply AutoVER vs fine-tuned LLaVA-v1.5 under entity match | AutoVER-7B · A-OKVQA-Ent entity match | overall hit rate: 46.2 -> 55.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PaLI-17B — PaLI: A Jointly-Scaled Multilingual Language-Image Model (ICLR 2023) · better on 5 of 5 shared results
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
- Web-Scale Visual Entity Recognition: An LLM-Driven Data Approach (NeurIPS 2024 · similarity 0.88)
- EvdCLIP: Improving Vision-Language Retrieval with Entity Visual Descriptions from Large Language Models (AAAI 2025 · similarity 0.88)
- Open-ended VQA benchmarking of Vision-Language models by exploiting Classification datasets and their semantic hierarchy (ICLR 2024 · similarity 0.87)
- Revisiting the Role of Language Priors in Vision-Language Models (ICML 2024 · similarity 0.87)
- A Generative Approach for Wikipedia-Scale Visual Entity Recognition (CVPR 2024 · similarity 0.87)
- Rephrase, Augment, Reason: Visual Grounding of Questions for Vision-Language Models (ICLR 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.
