Caption
-
SpectralCanary: Embarrassingly Simple Low-Frequency Traces for T2I Dataset-Use Auditing
2026-10-22 17:59:03Sy-Tuyen Ho, Anirudh Satheesh, Soham Chawathe, Huy Nghiem, Yongkyun Lee, Andrew Adams Schoen, Furong HuangAbstract
Auditing whether a text-to-image (T2I) model was fine-tuned on a particular image collection is difficult: a model can absorb collection-level visual characteristics without reproducing training images. Proactive dataset-use auditing plants a shared signal in a dataset before release, then later checks whether that signal appears in a suspect model's generated outputs. Existing approaches often trade off verifiability, fidelity, robustness, and deployability. We introduce $\textbf{SpectralCanary}$, which uses a smooth low-pass mask to blend a fixed canary image into every collection image while leaving captions unchanged and requiring neither auxiliary generation nor per-image optimization. A lightweight verifier detects the transferred signal in suspect-model outputs. Across eight tests on the Pokemon dataset and three realistic data transformations, spanning two architectures families, SpectralCanary achieves the highest average per-generation TPR@$1\%$ FPR ($91.1\%$) and fixed threshold model-level accuracy using 30 generations ($94.8\%$) among the evaluated methods. Across five datasets spanning artistic, medical, satellite, and fashion domains with SD3.5, SpectralCanary averages $92.4\%$ per-generation TPR@$1\%$ FPR while preserving the released images (SSIM $0.989$, LPIPS $0.022$) with a near-zero mean change FID (mean $\Delta$FID $0.05$) across datasets. Matched-energy ablations on Pokemon and ROCOv2 support our design choice, low-frequency placement transfers more reliably compared to mid/high-frequency placement. Under the evaluated adaptive attacks, the most effective removal requires aggressive modification and incurs more than $25$ $\Delta$FID points. These results identify low-frequency visual signals as effective carriers for collection-level T2I audit traces.
Abstract (translated)
URL
https://arxiv.org/abs/2610.31659
PDF
-
Retrieval-Augmented Online Textual Memory for Episodic Memory Video Question Answering
2026-10-22 13:49:30Raffaele Cali, Giuseppe Lando, Rosario Forte, Antonino FurnariAbstract
Long egocentric videos pose a central challenge for multimodal language models: the visual stream must be compressed into a memory representation that can be searched efficiently while still preserving the evidence needed for question answering. Building on an online textual-memory framework in which first-person videos are converted into temporally grounded captions, this work systematically studies the role of retrieval in episodic-memory video question answering. We evaluate how different retrieval units, query formulations, context budgets, and reranking strategies affect multiple-choice answering on QAEgo4D-Closed, isolating retrieval as the main variable while keeping the underlying generated memory fixed. Our best configuration, Parent, matches fine-grained caption subchunks, then provides their full parent captions to the reasoner. It achieves 56.4 % accuracy, matching strong full-memory and retrieval baselines while exposing only a small fraction of the textual memory at question time. More importantly, our analysis shows that better temporal coverage does not necessarily translate into better answers: increasing the retrieval budget improves temporal Hit@K and Recall@K, but can reduce accuracy by introducing visually plausible distractors and uncertain caption evidence. These findings suggest that retrieval-augmented egocentric memory should not be optimized only for evidence recall, but also for the reliability, specificity, and answer-discriminative value of the retrieved context. The study motivates more structured episodic memories that move beyond raw caption retrieval toward evidence-aware representations for long-form egocentric reasoning.
Abstract (translated)
URL
https://arxiv.org/abs/2610.31284
PDF
-
ReCoG: Relational Concept Graph for Training-Free Personalization
2026-10-22 13:48:41Changseon Yu, Hyeokjun KweonAbstract
Training-free personalization adapts a frozen vision-language model (VLM) to recognize user-defined concepts from only a few exemplars by retrieving concept entries at inference time. While recent retriever-based methods enrich each concept with descriptive evidence, they largely treat concepts in isolation, so the retrieved information can be plausible yet non-discriminative, especially when the personalized concept set contains highly confusable, fine-grained categories. We argue that personalization is inherently relational: correct recognition depends on the specific distinctions between a concept and its nearest alternatives. To this end, we propose ReCoG, a Relational Concept Graph for retrieval-augmented personalization. ReCoG represents concepts as nodes and augments them with directed edges that explicitly capture discriminative cues between concept pairs, elicited by a frozen VLM during database construction. At inference, ReCoG retrieves a shortlist and resolves ambiguity via explicit pairwise comparisons conditioned on the corresponding relational edges, selecting the concept that consistently wins against competing candidates. We also introduce FINGER benchmark to test personalized recognition on confusable concept sets. Across FINGER, ReCoG substantially outperforms prior methods, with the largest gains in the most confusable regimes. ReCoG also achieves consistently best performance on established personalization benchmarks including captioning and VQA, confirming that relational evidence is broadly beneficial beyond fine-grained settings. Project Page: this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2610.31282
PDF
-
Predicting Music Valence and Arousal via Textual and Multimodal In-Context Learning in Large Language Models
2026-10-22 12:49:14Chi-En Dai, Chih-Pin Tan, Yi-Hsuan YangAbstract
While arousal can be reliably predicted from musical audio, emotional valence remains difficult to model due to its subjective and context-dependent nature. In this work, we investigate the potential of in-context learning (ICL) with large language models to advance continuous valence and arousal prediction. Rather than predicting emotion scores directly, our framework leverages contextual demonstrations from an auxiliary dataset to refine the initial predictions produced by a small, task-specific base model. We systematically evaluate a text-based ICL variant using lyrics and generated audio captions, alongside a multimodal variant that incorporates raw audio signals. Using the entire MER60 dataset strictly as a held-out test set, our experiments demonstrate substantial performance gains, particularly for valence. Text-based ICL improves the valence $R^2$ from 0.110 to 0.427 with 80 demonstration shots, while the multimodal variant achieves 0.346 with only 10 shots. These findings suggest that training-free ICL offers an effective inference-time strategy for current language models to predict continuous music emotion.
Abstract (translated)
URL
https://arxiv.org/abs/2610.31202
PDF
-
Where Do Long Captions Fail? Position-Aware Diagnosis and Reinforcement Learning for Detailed Image Captioning
2026-10-22 11:21:32Yuanze Hu, Zhichao Yang, Junwei Jing, Xin Yang, Wenxuan Liu, Tinghai Zhang, Jia Fu, Hongzhi Zhang, Xiangyu Wu, Ruiming Tang, Tingting Gao, Han LiAbstract
Detailed image captioning is usually framed as an amount problem: a model should mention more visual facts while avoiding hallucination. However, autoregressive VLMs generate captions as ordered sequences, and later generations may be produced under weaker or noisier visual grounding. We therefore argue that long-form captioning should be studied not only by what facts are mentioned, but also by where they appear. We introduce a position-aware rubric representation that decomposes dense human references into atomic visual claims, verifies generated captions against those claims, and induces a quartile-wise composition of supported, hallucinated, and generic content. This lens reveals two qualitatively different failures in long captions: models tend to produce unsupported specificity in the middle, while the final portion often escapes into vague, non-evidential language rather than simply accumulating more hallucination. A same-reference human control does not exhibit these patterns; additional human audits examine the visual validity of the induced labels. Motivated by this diagnosis, we propose PAGER (Position-Aware Grounded Evidence Reinforcement), a GRPO training objective with separate signals for supported fine-grained middle evidence, evidence-bearing tail continuation, claim coverage, and content-aware length control. The result is a closed loop: the same ordered claim representation identifies where long captions fail and supplies the position-specific reward needed to repair those failures. Experiments on detailed-caption benchmarks show that PAGER improves position-wise grounding and pairwise caption preference while preserving competitive global caption quality.
Abstract (translated)
URL
https://arxiv.org/abs/2610.31111
PDF
-
When Annotation Patterns Are Learned: Skin-Tone Dependence in Dermatology Concepts
2026-10-22 08:24:13Y. Li, M. Bigliardi-Qi, S. Bedros, P. Bigliardi, N. PapanikolopoulosAbstract
Concept bottleneck models (CBMs) let clinicians inspect and correct skin disease diagnoses through clinical concepts, but only if the concept annotations are reliable. We therefore audit the reliability of SkinCon, a widely used dermatology concept dataset, across skin tones. Within the same disease and with near duplicates removed, erythema is annotated far less often on dark than on light skin, with an odds ratio of 0.17. In contrast, hyperpigmentation and hypopigmentation are annotated more often, with odds ratios of 6.2 and 4.5, respectively. We refer to this pattern as tone-dependent concept annotation. An evidential concept model trained on these annotations learns the same pattern and reproduces it on unseen images. Such a model may agree with potential erythema omissions rather than contradict them. Consistent with this, three label error detectors underperform a disease-only baseline at finding erythema that dermatologist captions mention but SkinCon labels as absent. The three detectors are based on confident learning, training dynamics, and the model's own evidential uncertainty. Concept annotations should thus be audited within each subgroup before explanations built on them can be trusted.
Abstract (translated)
URL
https://arxiv.org/abs/2610.30901
PDF
-
Class-Grounded Batik Captioning under Scarce Text Annotation: Label Retention and Source-Specific Error Propagation
2026-10-22 03:22:08Novanto Yudistira, Muhammad Dzaka Aufa Fadhillah, Irawati Nurmala SariAbstract
Class labels offer compact supervision for describing cultural-heritage images when expert text annotations are scarce. We investigate a batik-captioning cascade that integrates a LoRA-adapted DINOv2 classifier with Qwen2.5-VL-7B, adapted using LoRA and loaded in 4-bit NF4. A $3\times2$ evaluation combines three captioner conditions with two inference prompts to examine adaptation and predicted-class injection. Across 150 caption-evaluation images, all three injected conditions achieve a ground-truth class-label mention rate of 0.9067. The classifier's separate test yields 96.10% pooled accuracy, with 98.33% on 300 public-source images and 75.76% on 33 local images. Supported-class macro-F1 and balanced accuracy on the local subset are 0.8493 and 0.8302, respectively. The classifier and captioner use separate partitions, and their training--evaluation overlap remains unquantified. Differences in label semantics and training support also affect interpretation of the source-specific accuracy gap. Reference-based caption metrics and image--text compatibility scores produce different model rankings. Qualitative examples show both retention of correctly predicted identities and propagation of classification errors into fluent descriptions. These findings characterize how class conditioning preserves label information and transfers upstream errors into batik captions under scarce text annotation.
Abstract (translated)
URL
https://arxiv.org/abs/2610.30661
PDF
-
Visual Grounding First, Multimodal In-context Learning Follows
2026-10-22 01:48:28Minhyuk Seo, Minjae Lee, Chaeeun Lee, Wei Lin, Muhammad Jehanzeb Mirza, Tinne Tuytelaars, Jonghyun ChoiAbstract
LLMs have shown strong in-context learning (ICL) capability, but extending it to Multimodal LLMs (MLLMs) remains challenging. Prior multimodal ICL methods often rely on ICL-specific datasets for additional training or task vector extraction, which can improve performance on the same ICL benchmark used for adaptation. However, they do not generalize well to other benchmarks and often induce forgetting in previously well-solved tasks. For this reason, we identify 'visual grounding' as the key bottleneck in multimodal ICL; MLLMs often fail to attend to task-relevant visual evidence in demonstrations, instead over-relying on textual cues and producing hallucinated outputs. To address this, we propose ProCoRe to enhance visual grounding through contrastive reinforcement learning on multimodal contrastive data for improving multimodal ICL. ProCoRe generates captions for contrastive image pairs, decomposes them into propositions, and optimizes contrastive rewards that encourage alignment with corresponding images while discouraging mismatched ones. Despite never seeing ICL-formatted training examples, ProCoRe improves average multimodal ICL classification accuracy by 5.6% and captioning ROUGE-L by 13.8% over the off-the-shelf Qwen-VL-3-8B model, outperforming ICL-data-dependent state of the arts across benchmarks. Finally, we introduce PICL, a personalized multimodal ICL captioning benchmark that tests whether MLLMs genuinely understand visual demonstrations rather than relying on pre-trained semantic priors.
Abstract (translated)
URL
https://arxiv.org/abs/2610.30592
PDF
-
HeyGen Voice: Scene-Aware Voice Cloning for Talking Videos
2026-10-21 21:42:44Pavel Andreev, Andrei Gusev, Adithya Narayan, Yi Ren, Ivan Somov, Zijian Li, Sergey Demyanov, Viktor Goriachko, Alan Tan, Nik Nolte, Sam Prokopchuk, Andy Pan, David Chou, Kirill Leonov, Penghan Wang, Jiewei Yuan, Zhibin Hong, Rong YanAbstract
Most voice-cloning systems are built for audio alone: they copy the reference recording's delivery and acoustics, so a voice cloned on a phone sounds like a phone recording whether the speaker is pictured in a studio, on a street, or in a cathedral. Speech for a talking video should sound like the pictured person speaking in the pictured place. We present HeyGen Voice, a scene-aware voice-cloning model that separates who is speaking from how and where: it keeps the speaker's voice from an everyday recording while the target image sets the delivery and acoustics. Prior image-conditioned TTS does half of this job: it either adds scene acoustics without keeping a specific speaker or infers a voice from a face. HeyGen Voice is a multimodal autoregressive transformer conditioned on the image, speaker and scene captions, and the voice recording, trained in five stages that move from multilingual TTS pretraining through simulated and real cross-scene adaptation to reinforcement and preference optimization. Two clone policies trade scene adaptation against fidelity to the original recording, and an audio anchor holds long scripts in one voice and one acoustic environment. Listeners prefer HeyGen Voice over leading commercial systems across a third-party arena, production votes, and talking-video studies: it ranks first (Elo 1201) on the public Artificial Analysis controlled-voice TTS leaderboard (October 2026); in a blind test, users cloning their own voice chose it over ElevenLabs v4 and Cartesia Sonic 3.6 in 62% and 54% of votes; and volunteers rated talking videos of their own digital twin about 10 to 30 points higher (out of 100) when voiced by HeyGen Voice than by leading commercial alternatives. Try HeyGen Voice and hear samples at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.30470
PDF
-
AT-Differ: Dataset for Audio-Text Difference Captioning
2026-10-21 14:59:17Akihiro Fujii, Yuki Okamoto, Yuki Saito, Hiroshi SaruwatariAbstract
Evaluating audio-text semantic alignment is important in various audio-language tasks. Existing methods mainly quantify audio-text semantic alignment using a single score and cannot explicitly describe how audio and text differ. In this paper, we propose audio-text difference captioning, a task that generates textual descriptions of differences between audio and text. We also construct a new dataset AT-DIFFER containing audio-text pairs with systematically simulated differences and corresponding difference captions. Using AT-DIFFER, we fine-tune an existing multimodal large language model to describe differences between audio and text. The experimental results show that the original model has limited ability to describe such differences, whereas fine-tuning on AT-DIFFER substantially improves this ability.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29836
PDF
-
MemoryTax: The Bitter Lesson of Agent Memory
Abstract
Agent memory systems such as Mem0 and Zep decide what to keep before any question arrives; coding agents keep the raw history and let the model search it with its own tools. We ask whether designed memory still helps once the model is strong enough to search for itself. On one LoCoMo conversation (419 turns, 152 questions), we run four memory methods (full context, BM25, Mem0, Zep) and three coding agents (Pi, Codex, Claude Code), each with GPT-4o-mini, GPT-6 Luna and Claude Opus 5.5, and record what every rollout costs. This is a short empirical study, with one conversation and one run per cell. Agent memory is a costly patch for a weak model: Mem0 beats full context only with GPT-4o-mini, and its build alone costs more than a whole full-context rollout. Agent memory also has a ceiling, set by the model that writes it: a stronger reader does not recover what a weaker writer left out. With Claude Opus 5.5, Pi tops the cost-accuracy frontier, scoring 99% for less than half the cost of full context; its one miss is a wrong answer key. Part of that lead comes from photo captions that only the agents could read: on the 88 questions whose evidence has no photo, Pi and full context differ by one answer.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29823
PDF
-
BiMoGen: Bidirectional Motion-Text Generation via Unified Masked Discrete Diffusion
2026-10-21 11:15:39Wanjiang Weng, Yongliang Wu, Xiaofeng Tan, Xingyu Zhu, Wenbo Zhu, Hongsong WangAbstract
Text-to-motion generation and motion-to-text captioning are two fundamental tasks in human motion modeling, both grounded in the same underlying motion-text correspondence. Existing unified approaches mostly rely on autoregressive modeling, which imposes a fixed generation order and is therefore poorly suited to the bidirectional dependencies between language and motion, allowing early prediction errors to persist as fixed context and degrade both temporal coherence and cross-modal consistency. Masked discrete diffusion, which models sequences through iterative bidirectional prediction, allows prediction from bidirectional context without a fixed generation order. We therefore propose BiMoGen (Bidirectional Motion-text Generation), a unified masked discrete diffusion framework for bidirectional motion-text modeling. To stabilize training, we design Decoupled Uni- and Cross-Modal Training, in which masked pretraining first establishes cross-modal correspondence on paired motion-text sequences, after which supervised fine-tuning specializes the model for bidirectional generation. Masked diffusion nonetheless faces a train-test mismatch, as the model is trained on clean ground-truth context yet encounters self-generated and potentially erroneous context at inference, with errors committed under heavily masked states propagating through subsequent steps. We further introduce Generation-Aware Self-Correction that exposes the model to its own predictions during training and uses correction passes at early sampling steps for further refinement of previously unmasked tokens. Extensive experiments on HumanML3D and KIT-ML demonstrate competitive performance on both tasks, validating the effectiveness of the proposed two-stage training and self-correction designs. The project page is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2609.35407
PDF
-
Generalized CycleGRPO for Fine-Grained Multimodal Understanding and Grounding
2026-10-21 10:39:56Xin Zhang, Haochen Wang, Beibei Lin, Zhaoxiang Zhang, Robby T. Tan, Jason Li, Dacheng TaoAbstract
Fine-grained multimodal intelligence requires models to both understand a designated region and localize it from language. Existing methods learn these directions with separate objectives and task-specific annotations. We present Generalized CycleGRPO, the first unified reinforcement learning framework coupling region understanding and localization across spatial, temporal, and spatio-temporal domains. Given a region, a multimodal large language model generates candidate descriptions and then localizes each description to reconstruct the input region. Reconstruction overlap provides a verifiable reward for captioning, while individual localization rewards improve the inverse mapping. A common region-space formulation extends this cycle from image masks and boxes to video intervals and timestamped spatial evidence, without reference captions or an external language-model judge during reinforcement learning. In videos, the cycle encourages descriptions that distinguish a target event from similar recurrences. We introduce DOVE (Distinctive Occurrence-level Video-caption Evaluation), a human-verified contrastive benchmark for this temporal distinctiveness. CycleGRPO improves SAMTok's overall gIoU from 57.5% to 68.2% on GroundingSuite and average gIoU from 71.2% to 78.7% on GRES. For temporal tasks, it achieves 51.9% tIoU on Charades and 33.3% Temporal Evidence Coverage on DOVE, compared with 45.6% and 14.8% for the Qwen3-VL initialization. These results establish region-text cycle consistency as a general self-evaluation principle for multimodal understanding and localization. Models, code, and DOVE are available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29552
PDF
-
Omnilingual-MsASR: Benchmarking Multimodal, Multilingual, Long-Form Multi-Speaker ASR of Omni Large Language Models
2026-10-21 09:11:30Yuan Ge, Pei Zhang, Xize Cheng, Lingzhou Mu, Mingyang Wang, Jin Xu, Baosong Yang, Shengxiang Gao, Jingbo Zhu, Zhengtao Yu, Tong XiaoAbstract
Omni-modal large language models (LLMs) have achieved remarkable progress in audio and video understanding. However, existing models struggle with long-form, multi-speaker videos, especially in understanding the entire video and recognizing who says what and when. To characterize the capability limit of current omni LLMs and guide future model development, we introduce Omnilingual-MsASR, a multilingual, multimodal benchmark for long-form Multi-speaker Automatic Speech Recognition (MsASR) covering six languages, recordings up to 30 minutes, and scenarios with up to 20 speakers. Omnilingual-MsASR provides parallel audio and video inputs to evaluate models' cross-modal consistency and their ability to exploit complementary cues across modalities. Extensive evaluations of leading omni-modal LLMs reveal substantial performance bottlenecks, while their tcpWERs frequently exceed 100%. These findings highlight fine-grained temporal localization, speaker disambiguation, and cross-modal synergy as key challenges for future omni-modal foundation models. Our work is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29455
PDF
-
OACap: Caption-Guided Active Observation for Detailed Video Captioning
2026-10-21 07:26:26Hao Shi, Hong Yan, Xiaoyang Zhang, Qing Cui, Chao HuangAbstract
Unlike question answering, where a query specifies a perceptual target, open-ended detailed captioning must determine what evidence to acquire and how to integrate it into a coherent global description. We propose Observation-Adaptive Video Captioner (OACap), a caption-centered agent that turns unsupported or missing information into targeted observation requests, uses spatiotemporal tools to acquire evidence, and iteratively verifies and revises the caption. OACap captions consistently improve performance across general video-understanding benchmarks using a fixed text-only reader. To better diagnose the information they preserve, we introduce Factual-Semantic Captioning Bench (FSCap-Bench). Its Core-Support Factuality Score (CSFS) combines claim precision with importance-aware fact coverage and responds as expected to four controlled perturbation types on 93.0-98.3% of valid caption pairs. Its Cross-Modal Semantic Round-Trip Score (CSRS) measures semantic preservation through a video-caption-video-caption round trip by matching atomic facts from source and recovered captions. Together, they connect downstream textual-memory utility with factual and semantic preservation.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29327
PDF
-
HORIA: 4D Human and Object Reconstruction from a Monocular Video with Interaction-Aware Re-Rendering
2026-10-21 07:08:30Hyeongjin Nam, Taeryung Lee, Daniel Sungho Jung, Kyoung Mu LeeAbstract
Joint 4D reconstruction of a human and object from a monocular video is essential for understanding how people interact with their environment. While recent methods have made noticeable progress, capturing such human-object interactions remains difficult due to two main challenges. First, a video typically contains composite semantic context spanning multiple interaction phases, making it difficult to track the temporal changes of the interaction in the reconstruction. Second, under human-object occlusion, a single-view observation contains incomplete visual information about the interaction, resulting in geometrically inconsistent reconstructions. To address these challenges, we introduce HORIA, built upon two core designs. First, we decompose the entire input video into temporally grounded and semantically distinct interaction phases with motion-aware captioning, obtaining a specific text description for each phase. Second, conditioned on these text descriptions of the interaction, we propose an interaction-aware re-rendering that compensates for the limited single-view observation. Specifically, we render the initially reconstructed 4D human and object from novel viewpoints and refine the renderings with a text-to-video diffusion prior, which provides auxiliary novel-view supervision beyond single-view cues. As a result, our framework produces temporally coherent and accurate 4D reconstructions under diverse interaction scenarios, achieving state-of-the-art performance.
Abstract (translated)
URL
https://arxiv.org/abs/2610.25960
PDF
-
Cache Locally, Recompute Selectively for Streaming Video Understanding
2026-10-21 05:39:35Junhyeok Lee, Sangyoon Lee, Jaeho LeeAbstract
Streaming video understanding (SVU) requires video-LLMs to process incoming video continuously while answering user queries with low latency. Recent methods reduce query-time cost by prefilling frames into KV caches during idle time and reusing the cached KVs at query time. To bound per-frame computation over unbounded video streams, these methods restrict the context available during prefill. This creates a fundamental context mismatch: the restricted context may omit information needed to properly contextualize video tokens, yet, still contain information irrelevant to the eventual query, which is unknown during prefill. We propose FrameBlend, a simple yet effective plug-and-play method to address this mismatch through selective KV recomputation, without requiring any additional training. At query time, FrameBlend re-prefills selected video tokens using query-relevant context and replaces their cached KVs with recomputed ones. Crucially, this recomputation enables shorter attention windows during idle-time prefill, substantially reducing prefill cost while improving accuracy. Experiments across three SVU frameworks and five streaming and offline benchmarks demonstrate accuracy gains of up to 3.9 pp. On RVS-Ego, FrameBlend reduces per-frame prefill FLOPs by over 20% and total processing time by 14-59% for retrieval-based frameworks, including the cost of query-time recomputation. These results demonstrate that a simple change in how contextualization is distributed between idle time and query time can improve the accuracy-efficiency trade-off for SVU.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29229
PDF
-
Agentic Video Understanding: A Survey
2026-10-21 02:49:27Xinyu Deng, Siwen Luo, Daochang LiuAbstract
As large language models (LLMs) become capable of processing increasingly diverse modalities and longer temporal contexts, an emerging line of work is moving beyond fixed video-language inference toward agentic systems that actively decide what information to inspect, retain, verify, and act upon. This survey reviews video understanding agents: systems that use video as the primary information source and solve understanding tasks through adaptive state construction and action selection. We first formalize an agent loop for video understanding, then address a central question: why do agents matter for video understanding? To answer this, we organize the literature through a challenge-to-design taxonomy, linking context bottlenecks to hierarchical evidence memory, evidence sparsity to active evidence acquisition, temporal causality to state and process tracking, and multimodal ambiguity to role-specialized coordination. We further review state space paradigms, learning paradigms, supervision signals, benchmarks, and evaluation protocols. Finally, we identify open directions toward agentic-native temporal modeling and video-native agents. Project page: this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2609.31713
PDF
-
JUICE: Unified Multimodal Modeling in Pretrained Representation Spaces
2026-10-20 18:25:09Jiahan Li, Huangyuan Su, Shengbang Tong, Shusheng Yang, Ruihan Guo, Suozhi Huang, Hexu Zhao, Pratyusha Sharma, Sherry Yang, Jiatao Gu, Rob Fergus, Kyunghyun ChoAbstract
Most unified multimodal models treat images and text asymmetrically: a pretrained language model generates text, an attached diffusion head generates images, and outputs that contain both are produced in a fixed order, so the first output cannot adapt to the second. We ask whether both modalities can instead be generated by one process that is trained on them jointly from scratch. Pretrained encoder spaces hold both modalities in one continuous form and carry knowledge learned at scale, and recent work finds them effective as generative targets for vision and language. We therefore model the joint distribution directly in these spaces. We introduce JUICE (Joint Unified multimodal modeling In Continuous Encoder spaces), a single flow matching Transformer over frozen image and text representations, in which each modality has its own noise level. Which modalities are clean and which are generated then decides the task: image generation, captioning, visual question answering, or joint generation. Controlled pretraining runs test each design choice. At equal guidance, generating text as continuous states leaves image generation intact (48.41 GenEval, against 47.86 for a model trained only for text-to-image generation), while captions lose some exact wording. Training on joint and conditional denoising together beats either alone, and semantic encoder spaces beat VAE latents and raw pixels as targets. At 2.1B parameters, with no language model initialization and at 224x224, JUICE reaches up to 0.88 on GenEval and 73.1 on VQAv2. By sampling both modalities from the joint distribution (coupled sampling), it can also draw a route on a map while writing out the matching moves. On mazes with several valid solutions, joint success rises from 54.6% with independent sampling to 93.7%, and no modality has to go first. On visual spatial planning, JUICE solves 99.8% of maps at its training sizes.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28801
PDF
-
Industrial ReMEmbR: Scaling Spatiotemporal Memory to Robot Fleets
2026-10-20 17:58:32Jad Yammine, Malek Wahidi, Abrar Anwar, Yan Chang, Mihir Acharya, Anthony RizkAbstract
Warehouse robot fleets continuously observe operational conditions, but preserving these observations as reusable semantic memory incurs substantial computation, communication, and storage costs. For example, explaining recurring aisle blockages can require reviewing video from multiple robots across several shifts. We present IndustrialRemembr, a shared spatiotemporal memory system that extends NVIDIA ReMEmbR to robot fleets and reduces server-side computation by generating scene descriptions only on demand. On each robot, an NVIDIA Jetson encodes video clips into retrieval embeddings with Cosmos-Embed1, vision tokens with Cosmos3-Edge, and reconstruction latents with the Wan2.2 VAE, without uploading raw camera frames. The encoding models are re-used to detect people in image patches, zeroing them out to support PII abstraction for GDPR-compliant deployments. We split Cosmos3-Edge at its vision-language boundary, placing the vision encoder on the robot and the language decoder on an on-premises server. A Nemotron agent retrieves memories by content, position, and time and queries the decoder using cached vision tokens. This design eliminates caption generation at write time, while cached-token decoding reproduces the unsplit model's output in all 450 tested decodes.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28624
PDF
