AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Teaching VLMs to Localize Specific Objects from In-context Examples

Sivan Doveh, Nimrod Shabtay, Eli Schwartz, Hilde Kuehne, Raja Giryes, Rogerio Feris, Leonid Karlinsky, James Glass, Assaf Arbelle, Shimon Ullman, M. Jehanzeb Mirza

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: VLM capability analysis · 10 citations (OpenAlex)

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 changedWhereMeasuredEffect
add pseudo-name regularization (replace true class names with pseudo-names) to the fine-tuning data (ablation)IPLoc 7B · Few-shot personalized localization averagemIoU (%): 30.98 -> 43.31Improved
add pseudo-name regularization to the fine-tuning data (ablation)IPLoc 72B · Few-shot personalized localization averagemIoU (%): 47.98 -> 55.79Improved
fine-tune on all three tracking datasets (LASOT, GOT, TAO) instead of LASOT only (ablation)IPLoc 7B · ICL-LASOTmIoU (%): 30.30 -> 55.36Improved
fine-tune on TAO only versus the full three-dataset mix (ablation)IPLoc 7B · ICL-LASOTmIoU (%): 55.36 -> 14.06Hurt
sample few-shot conversations from the same video instead of arbitrary videos (ablation)IPLoc 7B · PDMmIoU (%): 28.94 -> 50.05Improved
sample few-shot conversations from the same video instead of arbitrary videos (ablation)IPLoc 7B · ICL-LASOTmIoU (%): 24.14 -> 39.13Improved
fine-tune with LoRA on the localization data (generalization retention check)Qwen2-VL-72B · GQAaccuracy: 64.66 -> 63.37Hurt
fine-tune with LoRA on the localization data (generalization retention check)Qwen2-VL-72B · SEEDaccuracy: 77.64 -> 77.66No clear effect
fine-tune with IPLoc data (pseudo-name test, names swapped for pseudo names)Qwen2-VL-7B · LASOT test set with pseudo namesmIoU (%): 16.55 -> 44.76Improved
apply the IPLoc fine-tuning recipe to LLaVA-OVLLaVA-OV-7B · Few-shot personalized localization averagemIoU (%): 20.65 -> 27.25Improved
scale base model from 7B to 72BQwen2-VL · Few-shot personalized localization averagemIoU (%): 31.77 -> 34.21Mixed
use a more explicit prompt instead of the original training prompt (ablation)Qwen2-VL-72B · Few-shot personalized localization averageIoU (%): 7.69 -> 34.21Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy