PANDO: Efficient Multimodal AI Agents via Online Skill Distillation

Recent advances in multimodal web agents often rely on increased inference-time computation, including rollout search, verifier passes, offline skill discovery, and specialist model stacks. This raises a central question: can a web agent become more efficient as it accumulates experience, rather than more expensive? We first analyze trajectories from VisualWebArena and identify three recurring sources of inefficiency: repeat-action loops, hidden discovery costs, and low prompt-cache reuse. We then introduce PANDO, a single-rollout online skill-distillation framework that maintains a structured Skill Library and combines progress reflection, confidence-based skill demotion, hierarchical routing, visual compression, and cache-aware prompting. On the full set of 910 VisualWebArena tasks, PANDO achieves a 58.3% success rate, outperforming SGV (54.0%) and our WALT reproduction (45.2%), while using 58% fewer tokens than SGV and 61% fewer tokens than WALT, without any pre-evaluation discovery budget. A 300-task ablation further shows that rules and routines provide most of the success gains, while routing, compression, and cache-aware prompting convert the larger skill library into lower marginal token cost. Finally, we introduce three trajectory-level efficiency metrics -- Action Repetition Rate, Step Overhead Ratio, and Prompt Cache Utilization -- to make efficiency visible beyond terminal success.

MAP: A Map-then-Act Paradigm for Long-Horizon Interactive Agent Reasoning

Current interactive LLM agents rely on goal-conditioned stepwise planning, where environmental understanding is acquired reactively during execution rather than established beforehand. This temporal inversion leads to Delayed Environmental Perception: agents must infer environmental constraints through trial-and-error, resulting in an Epistemic Bottleneck that traps them in inefficient failure cycles. Inspired by human affordance perception and cognitive map theory, we propose the Map-then-Act Paradigm (MAP), a plug-and-play framework that shifts environment understanding before execution. MAP consists of three stages: (1) Global Exploration, acquiring environment-general priors; (2) Task-Specific Mapping, constructing a structured cognitive map; and (3) Knowledge-Augmented Execution, solving tasks grounded on the map. Experiments show consistent gains across benchmarks and LLMs. On ARC-AGI-3, MAP enables frontier models to surpass near-zero baseline performance in 22 of 25 game environments. We further introduce MAP-2K, a dataset of map-then-act trajectories, and show that training on it outperforms expert execution traces, suggesting that understanding environments is more fundamental than imitation.

NTIRE 2026 Challenge on Video Saliency Prediction: Methods and Results

This paper presents an overview of the NTIRE 2026 Challenge on Video Saliency Prediction. The goal of the challenge participants was to develop automatic saliency map prediction methods for the provided video sequences. The novel dataset of 2,000 diverse videos with an open license was prepared for this challenge. The fixations and corresponding saliency maps were collected using crowdsourced mouse tracking and contain viewing data from over 5,000 assessors. Evaluation was performed on a subset of 800 test videos using generally accepted quality metrics. The challenge attracted over 20 teams making submissions, and 7 teams passed the final phase with code review. All data used in this challenge is made publicly available - https://github.com/msu-video-group/NTIRE26_Saliency_Prediction.

Human Gaze-based Dual Teacher Guidance Learning for Semi-Supervised Medical Image Segmentation

In the field of medical image segmentation, the scarcity of labeled data poses a major challenge for existing models to accurately perceive target regions. Compared with manual annotation, gaze data is easier and cheaper to obtain. As a classical semi-supervised learning framework, mean-teacher can effectively use a large number of unlabeled medical images for stable training through self-teaching and collaborative optimization. Our study is based on the mean-teacher framework. By combining gaze data, it aims to address two crucial issues in semi-supervised medical image segmentation: 1) expand the scale and diversity of the dataset with limited labeled data; 2) enhance the network's perception ability. We propose the Human Gaze-based Dual Teacher Guidance Learning model (HG-DTGL). In this model, human gaze serves as an additional hidden `teacher' in the mean-teacher architecture. We introduce the GazeMix to generate reliable mixed data to expand the diversity and scale of the dataset, and the Multi-scale Gaze Perception (MGP) module is used to extract the multi-scale perception of the network. A Gaze Loss is designed to align the model's perception with human gaze. We have verified HG-DTGL on multiple datasets of different modalities and achieved superior performance on a total of ten different organs/tissues, with extensive experiments. This demonstrates that our method has strong generalization ability for medical images of different modalities, and shows the great application potential of gaze data in semi-supervised medical image segmentation.

Evidence Packing for Cross-Domain Image Deepfake Detection with LVLMs

Image Deepfake Detection (IDD) separates manipulated images from authentic ones by spotting artifacts of synthesis or tampering. Although large vision-language models (LVLMs) offer strong image understanding, adapting them to IDD often demands costly fine-tuning and generalizes poorly to diverse, evolving manipulations. We propose the Semantic Consistent Evidence Pack (SCEP), a training-free LVLM framework that replaces whole-image inference with evidence-driven reasoning. SCEP mines a compact set of suspicious patch tokens that best reveal manipulation cues. It uses the vision encoder's CLS token as a global reference, clusters patch features into coherent groups, and scores patches with a fused metric combining CLS-guided semantic mismatch with frequency-and noise-based anomalies. To cover dispersed traces and avoid redundancy, SCEP samples a few high-confidence patches per cluster and applies grid-based NMS, producing an evidence pack that conditions a frozen LVLM for prediction. Experiments on diverse benchmarks show SCEP outperforms strong baselines without LVLM fine-tuning.

Label Noise Cleaning for Supervised Classification via Bernoulli Random Sampling

Label noise - incorrect labels assigned to observations - can substantially degrade the performance of supervised classifiers. This paper proposes a label noise cleaning method based on Bernoulli random sampling. We show that the mean label noise levels of subsets generated by Bernoulli random sampling containing a given observation are identically distributed for all clean observations, and identically distributed, with a different distribution, for all noisy observations. Although the mean label noise levels are not independent across observations, by introducing an independent coupling we further prove that they converge to a mixture of two well-separated distributions corresponding to clean and noisy observations. By establishing a linear model between cross-validated classification errors and label noise levels, we are able to approximate this mixture distribution and thereby separate clean and noisy observations without any prior label information. The proposed method is classifier-agnostic, theoretically justified, and demonstrates strong performance on both simulated and real datasets.

Multimodal Protein Language Models for Enzyme Kinetic Parameters: From Substrate Recognition to Conformational Adaptation

Predicting enzyme kinetic parameters quantifies how efficiently an enzyme catalyzes a specific substrate under defined biochemical conditions. Canonical parameters such as the turnover number ($k_\text{cat}$), Michaelis constant ($K_\text{m}$), and inhibition constant ($K_\text{i}$) depend jointly on the enzyme sequence, the substrate chemistry, and the conformational adaptation of the active site during binding. Many learning pipelines simplify this process to a static compatibility problem between the enzyme and substrate, fusing their representations through shallow operations and regressing a single value. Such formulations overlook the staged nature of catalysis, which involves both substrate recognition and conformational adaptation. In this regard, we reformulate kinetic prediction as a staged multimodal conditional modeling problem and introduce the Enzyme-Reaction Bridging Adapter (ERBA), which injects cross-modal information via fine-tuning into Protein Language Models (PLMs) while preserving their biochemical priors. ERBA performs conditioning in two stages: Molecular Recognition Cross-Attention (MRCA) first injects substrate information into the enzyme representation to capture specificity; Geometry-aware Mixture-of-Experts (G-MoE) then integrates active-site structure and routes samples to pocket-specialized experts to reflect induced fit. To maintain semantic fidelity, Enzyme-Substrate Distribution Alignment (ESDA) enforces distributional consistency within the PLM manifold in a reproducing kernel Hilbert space. Experiments across three kinetic endpoints and multiple PLM backbones, ERBA delivers consistent gains and stronger out-of-distribution performance compared with sequence-only and shallow-fusion baselines, offering a biologically grounded route to scalable kinetic prediction and a foundation for adding cofactors, mutations, and time-resolved structural cues.

Gradient Flow Drifting: Generative Modeling via Wasserstein Gradient Flows of KDE-Approximated Divergences

We reveal a precise mathematical framework about a new family of generative models which we call Gradient Flow Drifting. With this framework, we prove an equivalence between the recently proposed Drifting Model and the Wasserstein gradient flow of the forward KL divergence under kernel density estimation (KDE) approximation. Specifically, we prove that the drifting field of drifting model (arXiv:2602.04770) equals, up to a bandwidth-squared scaling factor, the difference of KDE log-density gradients $\nabla \log p_{\mathrm{kde}} - \nabla \log q_{\mathrm{kde}}$, which is exactly the particle velocity field of the Wasserstein-2 gradient flow of $KL(q\|p)$ with KDE-approximated densities. Besides that, this broad family of generative models can also include MMD-based generators, which arises as special cases of Wasserstein gradient flows of different divergences under KDE approximation. We provide a concise identifiability proof, and a theoretically grounded mixed-divergence strategy. We combine reverse KL and $χ^2$ divergence gradient flows to simultaneously avoid mode collapse and mode blurring, and extend this method onto Riemannian manifold which loosens the constraints on the kernel function, and makes this method more suitable for the semantic space. Preliminary experiments on synthetic benchmarks validate the framework.

Enhancing Persona Following at Decoding Time via Dynamic Importance Estimation for Role-Playing Agents

The utility of Role-Playing Language Agents in sociological research is growing alongside the adoption of Large Language Models. For realism in social simulation, these agents must adhere to their personas defined by character profiles, yet existing strategies-static prompt engineering or costly fine-tuning-fail to adapt personas to dynamic scenarios. Psychological theories, such as the Cognitive-Affective Personality Systems, provide a crucial explanation for this failure: a persona's influence on behavior is not static but varies with the scenarios. This context-dependence highlights the critical need for adaptive persona management. To address this gap, we propose a novel, theory-driven method that dynamically estimates context-dependent persona importance and integrates it into weighted reward-guided decoding, enabling inference-time persona following. Specifically, we introduce the Persona Dynamic Decoding (PDD) framework, which consists of two key components: (1) Persona Importance Estimation (PIE) module, which dynamically quantifies the contextual importance of persona attributes without requiring ground-truth supervision; and (2) Persona-Guided Inference-Time Alignment (PIA) paradigm, which leverages these importance scores to construct weighted multi-objective rewards and modulate generation probabilities during inference. Extensive experiments show the effectiveness of our method in utterance consistency and behavioral fidelity.

TopoCurate:Modeling Interaction Topology for Tool-Use Agent Training

Training tool-use agents typically relies on outcome-based filtering: Supervised Fine-Tuning (SFT) on successful trajectories and Reinforcement Learning (RL) on pass-rate-selected tasks. However, this paradigm ignores interaction dynamics: successful trajectories may lack error recovery or exhibit redundancy, while pass rates fail to distinguish structurally informative tasks from trivial ones. We propose \textbf{TopoCurate}, an interaction-aware framework that projects multi-trial rollouts from the same task into a unified semantic quotient topology. By merging equivalent action-observation states, this projection transforms scattered linear trajectories into a structured manifold that explicitly captures how tool invocations and environmental responses drive the divergence between effective strategies and failure modes. Leveraging this representation, we introduce a dual-selection mechanism: for SFT, we prioritize trajectories demonstrating reflective recovery, semantic efficiency, and strategic diversity to mitigate covariate shift and mode collapse; for RL, we select tasks with high error branch ratios and strategic heterogeneity, maximizing gradient Signal-to-Noise Ratio to address vanishing signals in sparse-reward settings. Evaluations on BFCLv3 and Tau2 Bench show that TopoCurate achieves consistent gains of 4.2\% (SFT) and 6.9\% (RL) over state-of-the-art baselines. We will release the code and data soon for further investigations.