2026-08-25 AI 리서치 브리핑

Reinforcement Learning (RL) has demonstrated significant gains in image captioning, yet it is still limited in encouraging Large Vision-Language Models (LVLMs) to explore novel reasoning strategies. This limitation leads to a performance gap between RL and Supervised Fine-Tuning (SFT). In this paper, we argue that multi-modal retrieval can serve as an effective reasoning signal for caption refinement. Based on this insight, we present the Retrieval-Guided Refinement for Image Captioning (Re$^3$Cap), a retrieval-guided reasoning strategy that enhances image captioning without requiring additional annotations. Instantiated by Caption Refinement Suggester (CRS) and Caption Quality Assessor (CQA), this strategy identifies hallucinations and omissions in image captions, leading to more accurate and detailed descriptions. Extensive experiments demonstrate the superiority of our method in image captioning, even compared with Supervised Fine-Tuning. Especially, Re$^3$Cap outperforms GRPO with an average improvement of 8.64% in relation reasoning on the COCO-LN500 benchmark.

InfinityEdit: Infinite Video Editing with a Lightweight Edit-Ignition Adapter에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

Foundation models for medical image segmentation, like prompt-based MedSAM, generalize well across domains and modalities, often in zero or few-shot setups. However, their performance depends on the quality of prompts and the adaptation of the models to custom datasets. This work systematically examines how MedSAM generalizes across diverse medical imaging benchmarks, with six adaptation strategies: full-model and encoder-only LoRA, shallow and deep visual prompt tuning (VPT), and decoder-only and full fine-tuning. Models are trained on the International Skin Imaging Collaboration Challenge (ISIC 2018) dataset and evaluated under clean and increasingly noisy prompts on IN and Out-of-Distribution (OOD) datasets: close-OOD PH2 (dermoscopy), far-OOD BUSI (Breast Ultrasound Images Dataset) and CBIS-DDSM (Curated Breast Imaging Subset of the Digital Database for Screening Mammography). We show that adaptation improves performance on IN and close-OOD data but often reduces performance on far-OOD data. Full fine-tuning provides the best tradeoff, while encoder-only LoRA is the strongest parameter-efficient alternative, outperforming standard LoRA and VPT under far-OOD shifts. Using Centered Kernel Alignment (CKA), we show that far-OOD degradation is strongly associated with drift in decoder representations, whereas encoder similarity alone does not explain robustness. This suggests encoder-only LoRA provides stronger robustness than standard LoRA by adapting the encoder to distribution shift in visual features, while preserving the decoder pathway. We further show that random 0-100 pixel jitter on prompts produces more robust and better performing models. We thus conclude that robust MedSAM adaptation requires the combined consideration of prompt noise exposure, domain shift, and representation preservation. We release our code: https://github.com/ImSounic/medsam-vpt

Recent omni-modal large language models (Omni-LLMs) show great potential as real-time video assistants, which continuously perceive environments and guide users to achieve specific goals. Unlike traditional passive video understanding, interactive assistants should actively combine visual states, user goals, and prior knowledge to provide effective help. Evaluating this is rather challenging, as the model's unpredictable response dynamically changes the user's subsequent actions, which static offline datasets cannot accommodate. To address this bottleneck, we introduce OmniAssistBench. To solve the issue of diverging interaction paths where the same user goal can be achieved through various methods, we provide models with predefined priors derived from the source video, requiring them to guide users along the exact same routes. Since real interaction videos are rare, we construct the dataset by reverse-engineering existing Internet videos. We deduce logical user goals and segment the videos into multi-turn clips to simulate continuous interactions. This rigorous pipeline required over 1000 expert person-hours to build the dataset. Results show that the proprietary Gemini-3-Pro reaches 66.4 out of the max point of 100, while the open-source Qwen3-Omni-Instruct achieves 51.2. Although current models generally understand user inputs, they frequently provide incorrect or incomplete answers. Specifically, they struggle with visual prompts (e.g., hand gestures), fail to maintain historical context during multi-turn interactions, and fail to delay response until the target event. Results indicate substantial room for improvement before models can become reliable assistants.

Robotic strawberry harvesting requires precise 6D pose estimation; however, collecting 6D pose ground truth in real agricultural fields is inherently challenging. Existing strawberry 6D pose estimation studies have therefore relied mainly on synthetic data, leaving their in-field performance unquantified. In this work, we obtain ground truth indirectly, by recovering camera poses via PnP, reconstructing each scene at metric scale, and annotating a single 3D bounding box per strawberry that is propagated across all frames, yielding, to the best of our knowledge, the first real world 6D pose ground truth dataset of red-stage strawberries collected at an actual strawberry farm (12,040 images). We also introduce a synthetic dataset rendered in NVIDIA Isaac Sim, featuring scene-level realism and domainrandomization. Despite this improved simulation setup, models trained on synthetic data alone fail to transfer to in-field images, while introducing a small amount of real data improves both translation and rotation accuracy across all backbones. Under the monocular RGB setting evaluated here, rotation is largely recovered once real data is used, and depth is what limits pose accuracy. These baselines across backbone encoders serve as a reference for future work. The real-world dataset is publicly available at https://huggingface.co/datasets/WoojungSon/F ieldStraw6D, and the data-collection pipeline is available at https://github.com/wjson2435/FieldStraw6D-pipeline.

Let's Scale Step by Step: Compute-Efficient Hyperparameter Transfer for Large-Scale Mixture-of-Experts에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

Graph Engineering in the Era of LLM Agents: From Individual Intelligence to System Intelligence에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

OmniAssistBench: Assistant-style Interaction Benchmark for Omni-LLMs에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

ParaTempo: Efficient Parallel Reasoning via Temporal Confidence에 관한 최근 업데이트입니다. 자세한 내용은 원문 링크에서 확인할 수 있습니다.

Deep-learning models of anatomy can be numerically plausible yet anatomically impossible, and they generalize poorly when data are scarce. We introduce Anatomy-Informed Neural Networks (AINN), in which soft anatomic priors enter as penalty terms in the loss (e.g., a branching penalty that treats a renal transplant artery off the iliac instead of the aorta as unexpected rather than impossible), in direct analogy to a physics-informed neural network, and hard anatomic priors (e.g., continuity of the vessel) are built into the architecture and state representation, making such invalid predictions impossible by construction wherever the prior admits architectural enforcement. We develop it on a clinical test case with limited data: how the aortoiliac tree deforms when a stiff wire is introduced endoluminally. This is important to contemporary aortic surgery and will matter to autonomous endovascular navigation. We lift the vessel centerline and the wire path from R^3 to curves of frames in the Lie group SE(3), and couple a Cosserat-rod wire to a tortuosity-modulated, anatomically anchored vessel through a unilateral lumen-contact inequality. The prediction is a constrained minimizer of the coupled elastic energy, with contact forces as its Lagrange multipliers. Supervision is a Wasserstein-2 optimal-transport loss between the predicted projection through the C-arm geometry and the observed angiogram, so a 2D angiogram can train a 3D prediction. The kinematics, loss and projection are verified against known ground truth; the mechanics solver only against its own optimality conditions, and predicted displacement is not yet mesh-converged. Here, no network is trained. Future work will transfer this in silico model to real CT scans and test whether it improves predictive accuracy and reduces the training data required.