The paper presents CycleGRPO, a reinforcement learning framework that unifies region understanding and localization for multimodal large language models (MLLMs). By treating the MLLM as both actor and critic, the method generates region captions and immediately grounds them back into spatial coordinates, using a token‑level cycle‑consistency reward that obviates the need for textual ground truths. Experiments on SAMTok demonstrate that CycleGRPO can bootstrap region captioning, VQA, grounded dialogue, and referring segmentation simultaneously, achieving consistent performance gains without task‑specific fine‑tuning.
arXiv:2610.29552v1 Announce Type: new Abstract: Fine-grained multimodal intelligence requires models to both understand a designated region and localize it from language. Existing methods learn these...
The paper introduces PM4Bench, a multimodal, multilingual, multi-task benchmark built on a strictly parallel 10‑language corpus, allowing fair cross‑lingual comparison of Large Vision‑Language Models (LVLMs). It also proposes a vision setting that embeds textual inputs directly into images to better mimic real deployment scenarios. Experiments show OCR performance drives cross‑lingual gaps, leading to an OCR‑centric GRPO training strategy that improves multilingual VQA and reduces disparities without costly supervision.
PixVL introduces a self‑supervised post‑training framework for pixel‑level multimodal large language models, addressing the scarcity of mask‑text pairs and optimization interference between region segmentation and region understanding. It employs a unified Mask‑Text Consistency Cycle that generates and self‑verifies regional descriptions using confuser‑aware semantic verification and cross‑view verification, while a quality‑coupled bidirectional learning strategy turns segmentation and understanding into mutual generators and verifiers. Experiments show that PixVL improves performance on both region understanding and segmentation tasks.
Multimodal large language models (MLLMs) have made strong progress on visual question answering and image captioning, yet they still produce fluent claims about objects, attributes, or relations that...
The paper introduces FindIt, the first comprehensive benchmark for evaluating the promptable localization abilities of generalist multimodal large language models (MLLMs). It covers four core task categories—object detection, referring expression detection, instance-level detection, and video-based detection—and provides a unified framework that standardizes inputs, enforces parsable bounding box outputs, and defines transparent evaluation protocols. Using this benchmark, the authors assess a range of open-source and proprietary MLLMs, revealing both their strengths and limitations, particularly their sensitivity to formatting constraints and difficulty generalizing to minor variations.
SAVOR is a training framework for multimodal large language models that adds token and answer confidence to the output schema, optimises a Group Relative Policy Optimisation objective to penalise calibration error and poor abstention, and uses the learned confidence at inference to revisit visual evidence only when uncertain. Experiments on POPE, HallusionBench, AMBER, and MMHal-Bench with InternVL3-8B and Qwen3-VL-8B backbones show that SAVOR reduces hallucination while maintaining general capability on MME and MMBench, achieving lower Expected Calibration Error than DPO and decoding baselines.
Intelligent systems that act in the world require image understanding that is both comprehensive and spatially grounded. Current vision-language models (VLMs) can generate fluent and detailed image ca...
arXiv:2606. 00987v1 Announce Type: cross Abstract: Large Vision-Language Models (LVLMs) have shown strong visual understanding and language-guided grounding abilities, yet their capacity for multi-temporal visual reasoning remains underexplored.
arXiv:2605. 23500v2 Announce Type: replace-cross Abstract: Segmentation is a fundamental task in computer vision, underpinning pixel-level scene understanding and serving as a cornerstone for applications ranging from autonomous perception to medical image analysis.
The paper introduces Vision‑RL2, a region‑level reinforcement learning approach that optimizes a lightweight proposal network for fine‑grained multimodal large language model (MLLM) perception. By treating coherent image regions as actions and scoring them with a frozen MLLM reader, the method selectively focuses visual resolution on evidence, reducing token usage while improving accuracy across multiple benchmarks and backbones. The approach eliminates the need for region annotations, response sampling, or reasoning trajectories, and the refined proposals enable sparse encoding that magnifies relevant evidence.
arXiv:2608.22429v1 Announce Type: new Abstract: Multimodal Large Language Models (MLLMs) capable of thinking with images often rely on external tools for fine-grained perception. However, this relian...
PANORAMA introduces a new panoptic grounded captioning framework that jointly generates detailed image captions and associates each phrase with precise pixel-level masks. The authors create PanoCaps, a human‑annotated benchmark with dense captions and near‑complete pixel coverage, and propose a phrase‑mask matching protocol with a generalized Panoptic Quality metric. PANORAMA conditions a pretrained segmenter on contextualized phrase representations, learns to select appropriate masks, and achieves state‑of‑the‑art grounding performance on PanoCaps and other pixel‑level tasks.