Transformer
-
DyPES-VLA: Learning Shared Dynamics Priors and Embodiment-Specific Control for Cross-Embodiment Manipulation
2026-08-06 17:59:20Junfeng Li, Junjie He, Zhide Zhong, Yangyang Zheng, Pingyue Sheng, Jiayu Dong, Ruixin Li, Haodong Yan, Jiaguan Zhu, Tianran Zhang, Runze Yu, Wen Chen, Liuqing Yang, Yuxiang Gao, Haoang LiAbstract
Vision-Language-Action (VLA) models have become a powerful paradigm for robot manipulation, but training a single generalist policy for heterogeneous robot embodiments remains an open problem. Existing methods have two main limitations. First, they underuse dynamics priors shared across diverse visual and interaction data, limiting cross-embodiment transfer. Second, they require extensive manual preprocessing to convert embodiment-specific actions into a common format. To overcome these limitations, we propose DyPES-VLA, a cross-embodiment VLA that learns shared Dynamics Priors and Embodiment-Specific control. First, we learn shared dynamics priors by training the vision-language model (VLM) with a future-prediction objective on cross-embodiment data, driving the shared query representation to capture object motion, contact, and interaction-induced scene changes. Second, an embodiment-specific Mixture-of-Experts (MoE) action head translates these shared dynamics priors into executable controls directly in each embodiment's native action space, without manually pre-aligning heterogeneous actions into a common format. This head shares attention layers to capture common temporal action structures, while its embodiment-specific feed-forward experts resolve the unique kinematic constraints and control semantics of distinct embodiments. As a generalist policy, our \ourmethod achieves state-of-the-art performance across simulation and real-world evaluations, reaching 98.0% success on LIBERO, 59.25% on RoboCasa-GR1, and 89.02% on RoboTwin~2.0.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06374
PDF
-
The Bitter Lesson of Tool Calling
2026-08-06 17:58:32Ishan Patel, Sahil Sen, Elias Lumer, Vamse Kumar SubbiahAbstract
Tool use transforms LLMs into agents that act beyond their training data, and for code-capable models, programmatic tool calling extends this further by replacing rigid JSON calls with scripts that chain and parallelize naturally. However, a systematic evaluation of tools as code on an established benchmark across current and prior model generations under real-world task conditions has not been conducted. In this work, we empirically compare programmatic tool calling (PTC) to native JSON tool calling across 14 language models on BFCL v4. In the programmatic tool calling paradigm, tools are exposed as typed Python stubs that the model invokes through code, with execution and results handled in a single agent turn. Programmatic tool calling matches or exceeds native JSON tool calling in 11 of 14 models on BFCL v4, with the GPT-5.6 family achieving a 10.6% improvement over the JSON tool calling baseline. Further, it matches or outperforms baseline in 13 of 14 models under parallel fan-out, and holds stable under context rot conditions where baseline degrades 2.3% on average. Our results demonstrate that programmatic tool calling is a viable and robust alternative to JSON tool calling, with performance tracking model capability across release generations.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06370
PDF
-
Bias Analysis of L2 Speaking Assessment Systems Using Concept Activation Vectors
2026-08-06 17:20:58Arya Labroo, Mengjie Qian, Kate KnillAbstract
Automatic speaking assessment systems are increasingly deployed in high-stakes settings to mark second language (L2) learners' speaking tests, making it critical to show that their scores depend on speaking proficiency rather than irrelevant speaker attributes such as first language (L1) or age. Transformer-based foundation models have improved the accuracy of these L2 speaking graders, but their black-box representations make fairness and interpretability analysis more difficult. Building on prior work that used Concept Activation Vectors (CAVs) to detect bias towards unwanted attributes (`concepts') in feature-based graders, we extend CAV-based analysis to two neural speaking assessment systems: a text-based BERT grader and a speech-and-text multimodal grader based on Whisper. CAVs represent human-interpretable concepts as directions in a model's activation space, allowing us to distinguish between whether a concept is encoded in a model's internal representations and whether it influences the predicted score, the latter quantified using a gradient-based sensitivity metric. Since CAVs rely on linear separability, which is less likely in complex neural embedding spaces, we also investigate whether sparse autoencoders (SAEs) provide cleaner concept directions by learning CAVs in a sparse latent space and mapping them back to activation space. Our analysis shows that concept recoverability depends strongly on the representation and architecture being probed, rather than on the concept alone. Sensitivity to concepts is also architecture-dependent. SAEs make concepts more linearly recoverable, but attenuate the original activation-space sensitivity, especially in low-dimensional layers. These findings highlight the need to distinguish concept recoverability from concept influence when auditing bias in speaking assessment systems.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06300
PDF
-
BaKron: Efficient Quantization with Kronecker-Factored Hessians
2026-08-06 17:15:37Johann Birnick, Rayan SaabAbstract
We accelerate a family of algorithms for neural network quantization whose geometry is informed by any Kronecker-factored approximation of the Hessian. GPTQ-style adaptive rounding typically uses one-sided information derived from input activations. Two-sided Kronecker-factored Hessian approximations can additionally capture correlations across output coordinates, but applying GPTQ directly in the vectorized weight domain is computationally expensive. Building on the two-sided adaptive-rounding formulation used by BoA and YAQA, we introduce BaKron, an efficient solver that combines anti-diagonal parallelism with a recursive divide-and-conquer construction. For an $m\times n$ weight matrix, BaKron uses $O(m+n)$ sequential steps while reducing the total work from $O(m^2n^2)$ to $O(mn(m+n))$. Thus, it matches the cubic scaling of GPTQ while exploiting richer curvature information. Moreover, BaKron is modular with respect to both the base quantizer and the Hessian estimator. We also provide practical benchmarks, consider a range of Hessians that BaKron can be called with, find an efficient technique to compute these Hessians, and evaluate the algorithm experimentally.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06291
PDF
-
EmoWorld: A Decoupled Affective Field for Controllable Emotional Video Generation
2026-08-06 16:20:23Bingyuan Wang, Baistan Zhyldyzbekov, Kunyu Feng, Zeyu WangAbstract
Emotion shapes how viewers interpret a scene, yet existing video generators entangle global atmosphere, affect-bearing semantic cues, and temporal progression within a single text condition. We present EmoWorld, a framework that decouples these factors within a frozen flow-matching video diffusion transformer (Video DiT). A one-time preparation stage extracts layer-specific affect directions and a reusable cue library from geometry-preserving neutral and emotion-edited panoramas. At inference, Visual Atmosphere Steering (VAS) injects atmosphere directions into hidden states, Semantic Affective Steering (SAS) isolates a separately scalable prompt residual for semantic cues, and Temporal Affective Steering (TAS) interpolates endpoint residual fields across denoising and video time. On Wan2.2, VAS improves target-emotion alignment by 19% while reducing a temporal-fluctuation proxy by 48%; SAS improves target-emotion alignment by 37% and increases detected affect-bearing cues by 36%; and TAS improves transition monotonicity by 15% over the strongest baseline. EmoWorld is evaluated across 27 emotion categories in text-to-video and image-to-video settings, demonstrates portability across multiple Video-DiT backbones, and supports camera-conditioned composition without updating generator parameters.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06231
PDF
-
TS-RAG: Retrieval Augmented Generation for Time Series Forecasting
2026-08-06 16:12:57Yixiong Xiao, Congxi Xiao, Jingbo ZhouAbstract
While deep learning models, particularly transformer-based architectures, have shown impressive performance in time series forecasting, the application of retrieval-augmented generation (RAG) in this domain remains limited. Since RAG has proven effective in enhancing the capabilities of large language models by incorporating relevant external information, retrieving similar time series sequences as references might also improve accuracy in time series forecasting tasks. However, most time series models are constrained by limited training data, smaller parameter scales, and a lack of the extensive generative capabilities found in large language models. Simply concatenating reference sequences into the prompt, as done in language models, may not yield the expected results. To address these challenges, we propose a novel approach, TS-RAG, which leverages RAG to enhance forecasting performance. The framework introduces specially designed reference tokens to effectively fuse information from the input sequence with that from retrieved similar sequences, enabling a more robust capture of complex temporal dynamics. Experimental results demonstrate that TS-RAG achieves consistent state-of-the-art performance across several real-world forecasting benchmarks.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06223
PDF
-
Continual Learning in Transition
2026-08-06 16:07:26Zhiyan Hou, Dan Zhang, Tao Feng, Liyuan Wang, Wei Li, Xiangzhao Hao, Hongyan An, Junfeng Fang, Haokai Ma, Zhaohui Xu, Haiyun Guo, Jinqiao Wang, Tat-Seng ChuaAbstract
Classical continual learning (CL) has primarily focused on enabling models to update and retain knowledge through parameter-centric mechanisms, e.g., training strategies, architectural designs, and weight adaptation. However, emerging paradigms are reshaping the scope of CL beyond this traditional model adaptation view. For instance, on-policy learning broadens the space of update mechanisms; test-time training extends CL from the training phase to inference; and external harness components such as memory, skill libraries, and interaction protocols extend the evolutionary boundaries of model capabilities far beyond the static parameter space. Collectively, these developments indicate a transition from parameter-centric learning toward system-level adaptation. To characterize this transition, we examine the evolution of continual learning through three dimensions: When, How, and Where learning occurs. The How dimension encompasses off-policy, on-policy, and beyond-gradient optimization mechanics. The When dimension captures evolution across pre-training, post-training, and inference-time stages. The Where dimension delineates updates occurring within internal parameters versus external structural constraints. Anchored by this tri-axial framework, we systematically survey representative methods, trace the ongoing transition of continual learning, and discuss the key challenges, broader implications, and future directions arising from this paradigm shift.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06216
PDF
-
What Current AI Benchmarks Leave Unmeasured: Modality, Search, Citations, and Implications
2026-08-06 15:58:32Ro Encarnaci\'on, Tina Behzad, Emma Lurie, Dana\'e MetaxaAbstract
Large language model (LLM) benchmark evaluations are routinely used to support claims about model safety, reliability, and deployment readiness. Yet most evaluations rely on a single access modality (model APIs), perform a single run per prompt, and report accuracy as the primary outcome metric, without accounting for conditions such as web search that may have effects on model behavior in deployment. We audit these assumptions for one of the most widely-used LLMs, comparing two modalities, ChatGPT's chat UI and OpenAI's API, with and without web search enabled. We use a stratified total sample of 401 prompts from two popular benchmarks, BBQ and SafetyBench, collecting 4,812 total responses across three repeated runs per prompt. Beyond standard performance measures, we evaluate model output dimensions including response consistency, response text similarity, citation grounding, and abstention behavior. For instance, chat UI responses were less accurate than API responses on both benchmarks with search disabled. Enabling web search reduced accuracy by up to 8 percentage points, and even reversed the direction of modality performance trends for one benchmark. Repeated runs of the same prompt produced inconsistent responses in up to 21\% of prompts. The two modalities also grounded answers in different citations, and abstention behavior was also inconsistent across both modalities. These results illustrate that, even within a model family, reporting only simple accuracy metrics can obscure important forms of model behavioral variation relevant to AI safety assessments. We argue that AI safety evaluations should systematically account for modality, multi-run consistency, search conditions, and response-level behaviors to better reflect how deployed AI systems behave in practice.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06202
PDF
-
HOPE: Hand-Object Pressure Estimation from Monocular Videos
2026-08-06 15:51:46Subin Jeon, Byungjun Kim, Hanbyul JooAbstract
Estimating physical pressure from vision is essential for understanding contact-rich hand-object interaction. However, prior vision-based pressure estimation methods are largely limited to planar surfaces and single image input, making them difficult to apply to dynamic hand-object interaction with diverse objects. We instead formulate pressure estimation as a hand-centric video prediction problem with monocular video as input. This formulation predicts temporally evolving per-vertex normal pressure and contact directly on the hand mesh, yielding a unified output space independent of object shape and sensor layout. Building on this formulation, we propose \textbf{HOPE}, a framework with two key components. First, we lift tactile-glove pressure, planar-sensor pressure, and distance-based hand-object contact annotations into a shared hand vertex space, allowing bare-hand contact data to regularize pressure learning where metric labels are unavailable. Second, we introduce a vertex-anchored video transformer that treats each vertex as a persistent token, aggregates visual features and hand pose over time, and uses a contact-gated pressure head to enforce that pressure vanishes without contact. Experiments on OpenTouch, PressureVisionDB, and hand-object contact benchmarks validate HOPE across object-pressure, surface-pressure, and contact-supervised HOI settings. Despite using metric pressure supervision primarily from gloved-hand videos, HOPE generalizes to bare-hand egocentric and in-the-wild videos, producing joint contact and pressure predictions beyond the scope of contact-only or planar-pressure baselines.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06192
PDF
-
Audio-to-Score Transcription using Pre-trained Features, Data Augmentation, and the New SheetSage-A2S Dataset
2026-08-06 15:33:18Eoin Cummins, Zhongyi Huang, Alexandre D'Hooge, Zhuoro Mo, Yaolong JuAbstract
Existing audio-to-score (A2S) systems primarily focus on classical music, and the application to popular music remains underexplored. This paper first presents the new SheetSage-A2S Dataset, which includes 61 hours of audio with \texttt{**kern} score encodings for 9,468 clips originating from 6,066 unique songs, the first of its kind to facilitate A2S research for popular music. Additionally, we improve on existing A2S approaches by using data augmentation and MuQ, a pretrained feature-extraction model for music audio, to enhance generalisation abilities and extract meaningful audio features. Results show that the proposed A2S model achieves 4.98\% symbol error rate (SER) on the Quartets collection for classical music, which significantly outperforms the 15.3\% SER from the existing state-of-the-art \cite{alfaro-contrerasTransformer2024}. Additionally, our model achieves 20.92\% SER on the SheetSage-A2S dataset for popular music, serving as a strong benchmark for future research. The dataset, model, and code are made publicly available at: this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06165
PDF
-
Visual Grounding in Zero-Shot Vision-Language Control
2026-08-06 15:21:42J. de Curt\`o, Dayani Plasencia, Diego S\'anchez, I. de Zarz\`aAbstract
Vision-language models (VLMs) are increasingly used as zero-shot controllers, but successful trajectories do not necessarily show that decisions are grounded in visual input: simulator dynamics and conservative action priors can produce favourable scores without meaningful perception. We investigate this with an input-ablation battery: blind-image controls, repeated identical inputs, lane-axis reflection, non-visual baselines, and pipeline-integrity checks. Across nine direct-action models, six structured local VLMs, and an exploratory VLM-MPC hierarchy, we analyse 32,874 scored calls over two embodiments and three simulators. The direct-control results are largely negative: a constant-SLOW policy outperforms a scripted geometric controller, several models are image-invariant or nearly constant, and models that recognize longitudinal hazards still fail to transform LEFT and RIGHT under reflection. No local VLM meets the joint longitudinal and lateral grounding criteria. However, an image-only deterministic positive control estimates the lead gap with 0.090 m MAE and exact mirror equivariance, confirming the stimuli carry sufficient visual information; the failures are modular, not universal. A post-hoc, leakage-controlled symmetry-consensus guardian selects two models from 16 calibration frames and freezes a 2-of-4 hazard vote across original and reflected views. On 272 held-out frames it reaches 0.954 balanced accuracy (episode-cluster bootstrap 95% CI [0.895,0.990]); nested leave-one-episode-out recovers the same pair and threshold in all 12 folds. Abstaining on ties raises committed balanced accuracy to 0.973 at 0.824 coverage. With deterministic perception retaining lateral authority, offline modular replay achieves 0.934 action agreement and exact mirror equivariance. These results support current VLMs as bounded, selective hazard assistants, not monolithic zero-shot controllers.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06154
PDF
-
Hardware Keystores for AI Agent Signing Workflows: A Zero-Trust MCP Enforcement Architecture
2026-08-06 15:04:26Leo Sambrook, Sampo SovioAbstract
AI agents performing cryptographic operations (signing Git commits, authenticating API calls, issuing certificates) currently store private keys in software-accessible locations: plaintext files, environment variables, or container memory. Any process with sufficient read privileges can extract the raw key material. A recent production incident demonstrated the practical severity: private keys were exfiltrated from a widely deployed framework via email injection in under five minutes. We aim to enforce both key confidentiality and content-aware authorisation for key use. To that end, we replace software-resident keys with hardware-confined keys accessible through a vendor-neutral PKCS#11 interface. A hardware keystore (HSM, TPM, smart card) executes cryptographic operations on-device; the host receives only the result via opaque handles. Hardware confinement is the primary contribution; it is enabled by a surrounding five-layer Zero-Trust enforcement stack comprising session identity (SAGA), scope bounds (Smax), semantic validation (RAV), taint tracking, and the hardware execution boundary. We evaluate against 12 injection scenarios derived from AgentDojo's ImportantInstructionsAttack template (Debenedetti et al., arXiv:2406.13352). We run four LLM models; three follow injections in baseline mode (gpt-oss-120b, Qwen2.5-72B, DeepSeek-V4-Flash, n=192 combined). Baseline Attack Success Rate (ASR): 19.3% [14.3%, 25.4%]; protected ASR: 0% (Wilson 95% CI upper bound 2.0%). Zero false positives across four benign task scenarios.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06130
PDF
-
Is Self-Pretraining really useful to improve diagnosis in medical Time Series?
2026-08-06 14:53:23Omar Coser, Antonio Orvieto, Paolo Soda, Loredana ZolloAbstract
Inspired by recent evidence that transformer architectures benefit from Self-PreTraining (SPT) on long-context benchmarks, we investigate whether similar gains extend to multimodal, multivariate, and even simple univariate medical time series. Our objective is to assess the impact of SPT on the performance and scalability of transformer-based models across diverse medical applications, particularly under limited data conditions. We evaluate transformer architectures on three representative medical time-series tasks: rehabilitation robotics (Camargo dataset), stress detection (Non-EEG Stress), and Parkinson's disease detection (Gait Parkinson's Disease). Models are trained either from scratch or through SPT using four masking-based objectives designed to promote temporal and cross-modal representation learning, and we systematically vary model depth to examine how capacity interacts with pre-training benefits. Across datasets and configurations, SPT consistently improves classification accuracy by 0-6 percentage points depending on masking strategy, dataset and architecture, with gains observed not only in multivariate settings but also when models are restricted to simple univariate inputs. The improvements increase for deeper models that can better exploit the enriched temporal representations learned during pre-training. These findings indicate that SPT is a simple and general strategy that enhances transformer performance on medical time-series tasks without requiring task-specific architectural changes, supporting its potential to improve robustness and accuracy in data-limited clinical settings.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06122
PDF
-
Confidence matters: Leveraging Multi-view Geometric Priors for GS-based Reconstruction
2026-08-06 14:50:29Hongyu Zhou, Zorah L\"ahnerAbstract
3D Gaussian splatting (3DGS) has emerged as a widely-used tool for novel view synthesis, offering real-time rendering in a sparse representation. However, the method's reliance on structure-from-motion initialization and photometric optimization can lead to suboptimal geometric reconstruction, particularly for objects with high specularity. In this work, we investigate the integration of geometric priors, in the form of predicted normal and depth maps, into the 3DGS framework to improve the reconstruction quality. We analyze the effect of incorporating these priors into GS-based methods and our evaluation reveals that multi-view predictions, as they are done by the recent visual geometry grounded transformer (VGGT), outperform single-view alternatives. A major factor is the existence of a confidence map for the estimations, which comes as a by-product of multi-view models and which can significantly improve the effectiveness of priors by weighting each prediction appropriately. Extensive experiments on standard benchmarks show consistent improvement in reconstruction quality and significant gains in complex scenes including specular objects.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06117
PDF
-
Beyond Sequence Order: Syntax-Informed Positional Embeddings for Transformers
2026-08-06 14:44:31Haris Riaz, Hyungji Kim, Mihai SurdeanuAbstract
Positional embeddings (PE) in Transformers encode token distance and order but are largely agnostic to \textit{syntactic structure}. We introduce \textbf{S}yntax-\textbf{i}nformed \textbf{P}ositional \textbf{E}mbeddings (\textbf{SiPE}), which learns a lightweight syntactic prior from dependency parses during pretraining and injects it across all three dominant PE families (absolute, relative, rotary), for both encoders and decoders, leaving self-attention and the rest of the architecture untouched. We isolate \emph{where} and \emph{how} the prior should enter the model, and find it depends on the architecture: for autoregressive decoders that use relative PE, the prior is strongest when coupled multiplicatively with the relative-position term of the attention score, outperforming injection into the input embeddings, into self-attention, or into the positional and attention terms jointly---while for encoders it is best added directly to the input embeddings, composing with each encoder's native positional mechanism. We find that models pre-trained with SiPE improve on the SyntaxGym benchmark by up to $10.3\%$ while simultaneously reducing perplexity by $9.0\%$ over a base model with no syntactic supervision---a metric nearly every existing syntax-injection method instead degrades. Crucially, these gains extend beyond syntactic generalization: SiPE also improves real-world language understanding, raising scores on the GLUE benchmark by up to $8.2\%$ over a model trained without it. Unlike existing syntactic language models that marginalize over many parses at inference or discard syntax at runtime, SiPE conditions on a single parse, establishing a new Pareto frontier between syntactic supervision and inference cost.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06111
PDF
-
ECHO: A Locally-Deployable Agentic Health Assistant with Temporal Memory, Safety Guardrails, and Speech Assessment
2026-08-06 14:44:12Abdulkadir K\"ul\c{c}e, Alihan Esen, Ca\u{g}la Fikir, Berke Kurt, Kuzey Arar, G\"okhan Ercan, Faik Boray TekAbstract
This paper presents ECHO (Enhanced Care \& Health Observer), a locally-deployable conversational health assistant for long-term chronic care management. ECHO integrates three complementary software modules developed under shared supervision as a unified system. The core module is an agentic chatbot built on a ReAct loop orchestrated via LangGraph, equipped with 17 clinical tools and a temporal knowledge graph for persistent cross-session memory; it achieves a 94.9\% tool-execution pass rate across a 59-scenario benchmark with GPT-5 Mini. A two-stage hybrid safety layer intercepts all incoming queries: a rule-based layer handles explicit crisis signals and jailbreak attempts in under 1ms, while a signed graph neural network (GNN) with APPNP-style propagation classifies boundary cases by clinical intent, achieving 88.8\% accuracy and 90.6\% unsafe recall on a 2,537-query annotated Turkish health dataset while outperforming zero-shot LLM baselines including Llama 3.3 70B. A multimodal speech assessment module combining Whisper acoustic encoding and BERT text encoding with cross-attention fusion estimates emotion, depression, and pain, reaching a mean macro F1 of 0.652. The full system is implemented as a web application that can run entirely on consumer hardware, with no patient data transmitted to external services, supporting compliance with GDPR and KVKK.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06110
PDF
-
Dense-Cast: A lightweight ensemble of deep learning architectures for precipitation nowcasting
2026-08-06 14:26:21Gourav Jyoti Kalita, Hidam Kumarjit SinghAbstract
Proper short-term forecasting of precipitation is crucial in disaster management and preparedness. Nonetheless, the variability and nonlinearity of precipitation make short-term forecasting challenging for meteorologists. Moreover, capturing temporal dependencies in spatiotemporal data is a challenge in precipitation nowcasting. In this article, we introduce a lightweight deep learning model for half-hourly precipitation nowcasting. This model has been designed by incorporating the DenseNet architecture, residual connections, and transformer encoders for effective precipitation nowcasting with reduced model parameters. The North-Eastern region of India has been selected as the area of interest for our study. The region receives the highest precipitation during the months of June-September due to the monsoon season. The proposed model takes the previous five time-steps of half-hourly precipitation as inputs and predicts the precipitation in the next two half-hours. The GPM IMERG precipitation dataset with a 30-minute cadence has been used in this study for training and testing the model. The proposed architecture achieves best MAE of 0.235 millimetres, RMSE of 0.735 millimetres, and KGE score of 0.816 at an interval of 30 minutes.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06082
PDF
-
Domain-Grounded Candidate Selection for Agentic Image Editing: A Shadow Removal Case
2026-08-06 14:19:22Shilin Hu, Jingyi Xu, Dimitris Samaras, Hieu LeAbstract
Commercial vision-language models are reshaping computer vision, with visual priors broad enough to rival task-specific systems. This raises a natural question: do they reduce the need for classic, physics-informed low-level vision? We study this through shadow removal, a problem shaped by scene geometry, illumination, materials, and occluders, where paired shadow and shadow-free data are hard to collect at scale. We find that a commercial generative editor, used directly, can produce clean shadow-free edits that preserve surface texture and local appearance. However, this comes with a new failure mode: the same editor can regenerate scene content, hallucinate objects, or misread a shadow as material or geometry, producing plausible but physically wrong edits. We address this with an agentic candidate-selection pipeline: the editor generates a guided probe, an evaluator screens for major failures, retries when needed, samples multiple candidates, filters them, and selects a final result balancing shadow removal against scene preservation. Grounding this process in shadow-formation physics makes it more reliable: prompting the generator and evaluator to treat shadows as illumination effects caused by light occlusion, not material or object structure, measurably improves quality and consistency. On the ShadowRemovalRefine benchmark, our physics-oriented pipeline achieves a CDD of 0.0075, reducing CDD by at least 47% over the strongest prior method. These results suggest that commercial vision-language models do not replace classic low-level vision priors; instead, such priors remain useful for constraining and steering physically underconstrained generation.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06075
PDF
-
The Next Screenshot Knows: Gated Hindsight Distillation for Mobile GUI Agents
2026-08-06 14:10:23Weiwei Li, Junzhuo Liu, Tong Chu, Hengfu Yu, Wen LiAbstract
GUI agents are commonly trained offline from successful interaction trajectories. Standard training decomposes each trajectory into prefix-action pairs: the agent predicts an action from the current screen and interaction history, while the subsequent observation is discarded. This removes the rationale of why an action is correct: the evidence often appears only on the subsequent screen. For example, to enable Soft Wrap, the agent should click Edit or View, but nothing reveals this until the menu opens. Without such evidence, standard imitation gives the model little chance of ever sampling and thus learning the correct reasoning. To address this issue, we propose Gated Hindsight Distillation (GHD), which uses the next screenshot as privileged information during training. A student predicts from the observable trajectory prefix, while a parameter-sharing teacher additionally observes the next screenshot and re-scores the student's on-policy responses. We apply distillation only when the student fails and the hindsight-conditioned teacher recovers the demonstrated action. GHD improves task success over GRPO on AndroidWorld and AndroidLab across two vision-language models. The code and checkpoints will be made available.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06065
PDF
-
Learning from Failures: Retrieval-Centric CoT via Hard Negatives for Unified Multimodal Retrieval
2026-08-06 14:04:56Zelong Sun, Jun Wang, Kaicheng Yang, Tiancheng Gu, Ziyong Feng, Zhiwu LuAbstract
Unified multimodal retrieval aims to identify candidates that satisfy complex user intent expressed through heterogeneous inputs. Although Large Vision-Language Model (LVLM)-based retrievers are efficient and scalable, directly encoding raw multimodal inputs often misses fine-grained discriminative cues, leading to confusion among semantically similar candidates. Recent methods mitigate this limitation by generating Chain-of-Thought (CoT) rationales to enrich the query representation. However, such reasoning is typically derived from the query alone: it explains what the query describes, but not what the retriever misunderstands. We argue that effective retrieval reasoning should instead be conditioned on retrieval feedback. Based on this insight, we introduce UniME-R1, an embedder-adviser framework that learns to reason over initially retrieved candidates and generate Retrieval-Centric Chain-of-Thought (RC-CoT). The adviser analyzes candidates individually to identify the discriminative cues confused by the embedder. If the target appears in the initial top-k set, UniME-R1 directly reranks the candidates; otherwise, it generates RC-CoT to refine the retrieval direction and performs full-corpus re-retrieval with a dual-mode embedder. To train the framework, we mine hard negatives to simulate realistic retrieval failures, jointly optimize direct retrieval and RC-CoT-augmented retrieval, and align the adviser with retrieval outcomes through supervised learning and retrieval-oriented reinforcement learning. Extensive experiments on MMEB-V2 and a diverse set of general multimodal retrieval benchmarks demonstrate that UniME-R1 consistently improves retrieval performance over strong baselines.
Abstract (translated)
URL
https://arxiv.org/abs/2608.06060
PDF
