Many animals possess a remarkable capacity to rapidly construct flexible cognitive maps of their environments. These maps are crucial for ethologically relevant behaviors such as navigation, exploration, and planning. Existing computational models typically require long sequential trajectories to build accurate maps, but neuroscience evidence suggests maps can also arise from integrating disjoint experiences governed by consistent spatial rules. We introduce the Episodic Spatial World Model (ESWM), a novel framework that constructs spatial maps from sparse, disjoint episodic memories. Across environments of varying complexity, ESWM predicts unobserved transitions from minimal experience, and the geometry of its latent space aligns with that of the environment. Because it operates on episodic memories that can be independently stored and updated, ESWM is inherently adaptive, enabling rapid adjustment to environmental changes. Furthermore, we demonstrate that ESWM readily enables near-optimal strategies for exploring novel environments and navigating between arbitrary points, all without the need for additional training. Our work demonstrates how neuroscience-inspired principles of episodic memory can advance the development of more flexible and generalizable world models.

Vision-and-Language Navigation (VLN) tasks agents with locating specific objects in unseen environments using natural language instructions and visual cues. Many existing VLN approaches typically follow an 'observe-and-reason' schema, that is, agents observe the environment and decide on the next action to take based on the visual observations of their surroundings. They often face challenges in long-horizon scenarios due to limitations in immediate observation and vision-language modality gaps. To overcome this, we present VISTA, a novel framework that employs an 'imagine-and-align' navigation strategy. Specifically, we leverage the generative prior of pre-trained diffusion models for dynamic visual imagination conditioned on both local observations and high-level language instructions. A Perceptual Alignment Filter module then grounds these goal imaginations against current observations, guiding an interpretable and structured reasoning process for action selection. Experiments show that VISTA sets new state-of-the-art results on Room-to-Room (R2R) and RoboTHOR benchmarks, e.g.,+3.6% increase in Success Rate on R2R. Extensive ablation analysis underscores the value of integrating forward-looking imagination, perceptual alignment, and structured reasoning for robust navigation in long-horizon environments.

The substantial computational demands of modern large-scale deep learning present significant challenges for efficient training and deployment. Recent research has revealed a widespread phenomenon wherein deep networks inherently learn low-rank structures in their weights and representations during training. This tutorial paper provides a comprehensive review of advances in identifying and exploiting these low-rank structures, bridging mathematical foundations with practical applications. We present two complementary theoretical perspectives on the emergence of low-rankness: viewing it through the optimization dynamics of gradient descent throughout training, and understanding it as a result of implicit regularization effects at convergence. Practically, these theoretical perspectives provide a foundation for understanding the success of techniques such as Low-Rank Adaptation (LoRA) in fine-tuning, inspire new parameter-efficient low-rank training strategies, and explain the effectiveness of masked training approaches like dropout and masked self-supervised learning.

Colorectal cancer (CRC) is a major global cause of cancer-related deaths, with early polyp detection and removal during colonoscopy being crucial for prevention. While deep learning methods have shown promise in polyp segmentation, challenges such as high computational costs, difficulty in segmenting small or low-contrast polyps, and limited generalizability across datasets persist. To address these issues, we propose LGPS, a lightweight GAN-based framework for polyp segmentation. LGPS incorporates three key innovations: (1) a MobileNetV2 backbone enhanced with modified residual blocks and Squeeze-and-Excitation (ResE) modules for efficient feature extraction; (2) Convolutional Conditional Random Fields (ConvCRF) for precise boundary refinement; and (3) a hybrid loss function combining Binary Cross-Entropy, Weighted IoU Loss, and Dice Loss to address class imbalance and enhance segmentation accuracy. LGPS is validated on five benchmark datasets and compared with state-of-the-art(SOTA) methods. On the largest and challenging PolypGen test dataset, LGPS achieves a Dice of 0.7299 and an IoU of 0.7867, outperformed all SOTA works and demonstrating robust generalization. With only 1.07 million parameters, LGPS is 17 times smaller than the smallest existing model, making it highly suitable for real-time clinical applications. Its lightweight design and strong performance underscore its potential for improving early CRC diagnosis. Code is available at https://github.com/Falmi/LGPS/.

The strength of the human hand lies in its ability to manipulate small objects precisely and robustly. In contrast, simple robotic grippers have low dexterity and fail to handle small objects effectively. This is why many automation tasks remain unsolved by robots. This paper presents an optimization-based framework for in-hand manipulation with a robotic hand equipped with compact Magnetic Tactile Sensors (MTSs). The small form factor of the robotic hand from Shadow Robot introduces challenges in estimating the state of the object while satisfying contact constraints. To address this, we formulate a trajectory optimization problem using Nonlinear Programming (NLP) for finger movements while ensuring contact points to change along the geometry of the fingers. Using the optimized trajectory from the solver, we implement and test an open-loop controller for rolling motion. To further enhance robustness and accuracy, we introduce a force controller for the fingers and a state estimator for the object utilizing MTSs. The proposed framework is validated through comparative experiments, showing that incorporating the force control with compliance consideration improves the accuracy and robustness of the rolling motion. Rolling an object with the force controller is 30\% more likely to succeed than running an open-loop controller. The demonstration video is available at https://youtu.be/6J_muL_AyE8.

Machine learning models can inadvertently expose confidential properties of their training data, making them vulnerable to membership inference attacks (MIA). While numerous evaluation methods exist, many require computationally expensive processes, such as training multiple shadow models. This article presents two new complementary approaches for efficiently identifying vulnerable tree-based models: an ante-hoc analysis of hyperparameter choices and a post-hoc examination of trained model structure. While these new methods cannot certify whether a model is safe from MIA, they provide practitioners with a means to significantly reduce the number of models that need to undergo expensive MIA assessment through a hierarchical filtering approach. More specifically, it is shown that the rank order of disclosure risk for different hyperparameter combinations remains consistent across datasets, enabling the development of simple, human-interpretable rules for identifying relatively high-risk models before training. While this ante-hoc analysis cannot determine absolute safety since this also depends on the specific dataset, it allows the elimination of unnecessarily risky configurations during hyperparameter tuning. Additionally, computationally inexpensive structural metrics serve as indicators of MIA vulnerability, providing a second filtering stage to identify risky models after training but before conducting expensive attacks. Empirical results show that hyperparameter-based risk prediction rules can achieve high accuracy in predicting the most at risk combinations of hyperparameters across different tree-based model types, while requiring no model training. Moreover, target model accuracy is not seen to correlate with privacy risk, suggesting opportunities to optimise model configurations for both performance and privacy.

Large Vision-Language Models (VLMs) have achieved remarkable performance across a wide range of tasks. However, their deployment in safety-critical domains poses significant challenges. Existing safety fine-tuning methods, which focus on textual or multimodal content, fall short in addressing challenging cases or disrupt the balance between helpfulness and harmlessness. Our evaluation highlights a safety reasoning gap: these methods lack safety visual reasoning ability, leading to such bottlenecks. To address this limitation and enhance both visual perception and reasoning in safety-critical contexts, we propose a novel dataset that integrates multi-image inputs with safety Chain-of-Thought (CoT) labels as fine-grained reasoning logic to improve model performance. Specifically, we introduce the Multi-Image Safety (MIS) dataset, an instruction-following dataset tailored for multi-image safety scenarios, consisting of training and test splits. Our experiments demonstrate that fine-tuning InternVL2.5-8B with MIS significantly outperforms both powerful open-source models and API-based models in challenging multi-image tasks requiring safety-related visual reasoning. This approach not only delivers exceptional safety performance but also preserves general capabilities without any trade-offs. Specifically, fine-tuning with MIS increases average accuracy by 0.83% across five general benchmarks and reduces the Attack Success Rate (ASR) on multiple safety benchmarks by a large margin.

With the growing complexity of Cyber-Physical Systems (CPS) and the integration of Internet of Things (IoT), the use of sensors for online monitoring generates large volume of multivariate time series (MTS) data. Consequently, the need for robust anomaly diagnosis in MTS is paramount to maintaining system reliability and safety. While significant advancements have been made in anomaly detection, localization remains a largely underexplored area, though crucial for intelligent decision-making. This paper introduces a novel transformer-based model for unsupervised anomaly diagnosis in MTS, with a focus on improving localization performance, through an in-depth analysis of the self-attention mechanism's learning behavior under both normal and anomalous conditions. We formulate the anomaly localization problem as a three-stage process: time-step, window, and segment-based. This leads to the development of the Space-Time Anomaly Score (STAS), a new metric inspired by the connection between transformer latent representations and space-time statistical models. STAS is designed to capture individual anomaly behaviors and inter-series dependencies, delivering enhanced localization performance. Additionally, the Statistical Feature Anomaly Score (SFAS) complements STAS by analyzing statistical features around anomalies, with their combination helping to reduce false alarms. Experiments on real world and synthetic datasets illustrate the model's superiority over state-of-the-art methods in both detection and localization tasks.

Mixed-type time series (MTTS) is a bimodal data type that is common in many domains, such as healthcare, finance, environmental monitoring, and social media. It consists of regularly sampled continuous time series and irregularly sampled categorical event sequences. The integration of both modalities through multimodal fusion is a promising approach for processing MTTS. However, the question of how to effectively fuse both modalities remains open. In this paper, we present a comprehensive evaluation of several deep multimodal fusion approaches for MTTS forecasting. Our comparison includes three fusion types (early, intermediate, and late) and five fusion methods (concatenation, weighted mean, weighted mean with correlation, gating, and feature sharing). We evaluate these fusion approaches on three distinct datasets, one of which was generated using a novel framework. This framework allows for the control of key data properties, such as the strength and direction of intermodal interactions, modality imbalance, and the degree of randomness in each modality, providing a more controlled environment for testing fusion approaches. Our findings show that the performance of different fusion approaches can be substantially influenced by the direction and strength of intermodal interactions. The study reveals that early and intermediate fusion approaches excel at capturing fine-grained and coarse-grained cross-modal features, respectively. These findings underscore the crucial role of intermodal interactions in determining the most effective fusion strategy for MTTS forecasting.

Multi-agent credit assignment is a fundamental challenge for cooperative multi-agent reinforcement learning (MARL), where a team of agents learn from shared reward signals. The Individual-Global-Max (IGM) condition is a widely used principle for multi-agent credit assignment, requiring that the joint action determined by individual Q-functions maximizes the global Q-value. Meanwhile, the principle of maximum entropy has been leveraged to enhance exploration in MARL. However, we identify a critical limitation in existing maximum entropy MARL methods: a misalignment arises between local policies and the joint policy that maximizes the global Q-value, leading to violations of the IGM condition. To address this misalignment, we propose an order-preserving transformation. Building on it, we introduce ME-IGM, a novel maximum entropy MARL algorithm compatible with any credit assignment mechanism that satisfies the IGM condition while enjoying the benefits of maximum entropy exploration. We empirically evaluate two variants of ME-IGM: ME-QMIX and ME-QPLEX, in non-monotonic matrix games, and demonstrate their state-of-the-art performance across 17 scenarios in SMAC-v2 and Overcooked.

Panoptic maps enable robots to reason about both geometry and semantics. However, open-vocabulary models repeatedly produce closely related labels that split panoptic entities and degrade volumetric consistency. The proposed UPPM advances open-world scene understanding by leveraging foundation models to introduce a panoptic Dynamic Descriptor that reconciles open-vocabulary labels with unified category structure and geometric size priors. The fusion for such dynamic descriptors is performed within a multi-resolution multi-TSDF map using language-guided open-vocabulary panoptic segmentation and semantic retrieval, resulting in a persistent and promptable panoptic map without additional model training. Based on our evaluation experiments, UPPM shows the best overall performance in terms of the map reconstruction accuracy and the panoptic segmentation quality. The ablation study investigates the contribution for each component of UPPM (custom NMS, blurry-frame filtering, and unified semantics) to the overall system performance. Consequently, UPPM preserves open-vocabulary interpretability while delivering strong geometric and panoptic accuracy.

The computational efficiency of many neural operators, widely used for learning solutions of PDEs, relies on the fast Fourier transform (FFT) for performing spectral computations. As the FFT is limited to equispaced (rectangular) grids, this limits the efficiency of such neural operators when applied to problems where the input and output functions need to be processed on general non-equispaced point distributions. Leveraging the observation that a limited set of Fourier (Spectral) modes suffice to provide the required expressivity of a neural operator, we propose a simple method, based on the efficient direct evaluation of the underlying spectral transformation, to extend neural operators to arbitrary domains. An efficient implementation* of such direct spectral evaluations is coupled with existing neural operator models to allow the processing of data on arbitrary non-equispaced distributions of points. With extensive empirical evaluation, we demonstrate that the proposed method allows us to extend neural operators to arbitrary point distributions with significant gains in training speed over baselines while retaining or improving the accuracy of Fourier neural operators (FNOs) and related neural operators.

Robust disturbance rejection remains a longstanding challenge in humanoid locomotion, particularly on unstructured terrains where sensing is unreliable and model mismatch is pronounced. While perception information, such as height map, enhances terrain awareness, sensor noise and sim-to-real gaps can destabilize policies in practice. In this work, we provide theoretical analysis that bounds the return gap under observation noise, when the induced latent dynamics are contractive. Furthermore, we present Contractive Mapping for Robustness (CMR) framework that maps high-dimensional, disturbance-prone observations into a latent space, where local perturbations are attenuated over time. Specifically, this approach couples contrastive representation learning with Lipschitz regularization to preserve task-relevant geometry while explicitly controlling sensitivity. Notably, the formulation can be incorporated into modern deep reinforcement learning pipelines as an auxiliary loss term with minimal additional technical effort required. Further, our extensive humanoid experiments show that CMR potently outperforms other locomotion algorithms under increased noise.

Recent DiT-based text-to-image models increasingly adopt LLMs as text encoders, yet text conditioning remains largely static and often utilizes only a single LLM layer, despite pronounced semantic hierarchy across LLM layers and non-stationary denoising dynamics over both diffusion time and network depth. To better match the dynamic process of DiT generation and thereby enhance the diffusion model's generative capability, we introduce a unified normalized convex fusion framework equipped with lightweight gates to systematically organize multi-layer LLM hidden states via time-wise, depth-wise, and joint fusion. Experiments establish Depth-wise Semantic Routing as the superior conditioning strategy, consistently improving text-image alignment and compositional generation (e.g., +9.97 on the GenAI-Bench Counting task). Conversely, we find that purely time-wise fusion can paradoxically degrade visual generation fidelity. We attribute this to a train-inference trajectory mismatch: under classifier-free guidance, nominal timesteps fail to track the effective SNR, causing semantically mistimed feature injection during inference. Overall, our results position depth-wise routing as a strong and effective baseline and highlight the critical need for trajectory-aware signals to enable robust time-dependent conditioning.

Following their success across many domains, transformers have also proven effective for symbolic regression (SR); however, the internal mechanisms underlying their generation of mathematical operators remain largely unexplored. Although mechanistic interpretability has successfully identified circuits in language and vision models, it has not yet been applied to SR. In this article, we introduce PATCHES, an evolutionary circuit discovery algorithm that identifies compact and correct circuits for SR. Using PATCHES, we isolate 28 circuits, providing the first circuit-level characterisation of an SR transformer. We validate these findings through a robust causal evaluation framework based on key notions such as faithfulness, completeness, and minimality. Our analysis shows that mean patching with performance-based evaluation most reliably isolates functionally correct circuits. In contrast, we demonstrate that direct logit attribution and probing classifiers primarily capture correlational features rather than causal ones, limiting their utility for circuit discovery. Overall, these results establish SR as a high-potential application domain for mechanistic interpretability and propose a principled methodology for circuit discovery.

Reparameterization Policy Gradient (RPG) has emerged as a powerful paradigm for model-based reinforcement learning, enabling high sample efficiency by backpropagating gradients through differentiable dynamics. However, prior RPG approaches have been predominantly restricted to Gaussian policies, limiting their performance and failing to leverage recent advances in generative models. In this work, we identify that flow policies, which generate actions via differentiable ODE integration, naturally align with the RPG framework, a connection not established in prior work. However, naively exploiting this synergy proves ineffective, often suffering from training instability and a lack of exploration. We propose Reparameterization Flow Policy Optimization (RFO). RFO computes policy gradients by backpropagating jointly through the flow generation process and system dynamics, unlocking high sample efficiency without requiring intractable log-likelihood calculations. RFO includes two tailored regularization terms for stability and exploration. We also propose a variant of RFO with action chunking. Extensive experiments on diverse locomotion and manipulation tasks, involving both rigid and soft bodies with state or visual inputs, demonstrate the effectiveness of RFO. Notably, on a challenging locomotion task controlling a soft-body quadruped, RFO achieves almost $2\times$ the reward of the state-of-the-art baseline.

Diffusion Large Language Models (dLLMs) support arbitrary-order generation, yet their inference performance critically depends on the unmasking order. Existing strategies rely on heuristics that greedily optimize local confidence, offering limited guidance for identifying unmasking paths that are globally consistent and accurate. To bridge this gap, we introduce path log-likelihood (Path LL), a trajectory-conditioned objective that strongly correlates with downstream accuracy and enables principled selection of unmasking paths. To optimize Path LL at inference time, we propose POKE, an efficient value estimator that predicts the expected future Path LL of a partial decoding trajectory. We then integrate this lookahead signal into POKE-SMC, a Sequential Monte Carlo-based search framework for dynamically identifying optimal unmasking paths. Extensive experiments across 6 reasoning tasks show that POKE-SMC consistently improves accuracy, achieving 2%--3% average gains over strong decoding-time scaling baselines at comparable inference overhead on LLaDA models and advancing the accuracy--compute Pareto frontier.

Low-Rank Adaptation (LoRA) methods have emerged as crucial techniques for adapting large pre-trained models to downstream tasks under computational and memory constraints. However, they face a fundamental challenge in balancing task-specific performance gains against catastrophic forgetting of pre-trained knowledge, where existing methods provide inconsistent recommendations. This paper presents a comprehensive analysis of the performance-forgetting trade-offs inherent in low-rank adaptation using principal components as initialization. Our investigation reveals that fine-tuning intermediate components leads to better balance and show more robustness to high learning rates than first (PiSSA) and last (MiLoRA) components in existing work. Building on these findings, we provide a practical approach for initialization of LoRA that offers superior trade-offs. We demonstrate in a thorough empirical study on a variety of computer vision and NLP tasks that our approach improves accuracy and reduces forgetting, also in continual learning scenarios.

Integrating logical knowledge into deep neural network training is still a hard challenge, especially for sequential or temporally extended domains involving subsymbolic observations. To address this problem, we propose DeepDFA, a neurosymbolic framework that integrates high-level temporal logic - expressed as Deterministic Finite Automata (DFA) or Moore Machines - into neural architectures. DeepDFA models temporal rules as continuous, differentiable layers, enabling symbolic knowledge injection into subsymbolic domains. We demonstrate how DeepDFA can be used in two key settings: (i) static image sequence classification, and (ii) policy learning in interactive non-Markovian environments. Across extensive experiments, DeepDFA outperforms traditional deep learning models (e.g., LSTMs, GRUs, Transformers) and novel neuro-symbolic systems, achieving state-of-the-art results in temporal knowledge integration. These results highlight the potential of DeepDFA to bridge subsymbolic learning and symbolic reasoning in sequential tasks.

Large Reasoning Models (LRMs) achieve strong performance by generating long reasoning traces with reflection. Through a large-scale empirical analysis, we find that a substantial fraction of reflective steps consist of self-verification (recheck) that repeatedly confirm intermediate results. These rechecks occur frequently across models and benchmarks, yet the vast majority are confirmatory rather than corrective, rarely identifying errors and altering reasoning outcomes. This reveals a mismatch between how often self-verification is activated and how often it is actually useful. Motivated by this, we propose a novel, experience-driven test-time framework that reduces the overused verification. Our method detects the activation of recheck behavior, consults an offline experience pool of past verification outcomes, and estimates whether a recheck is likely unnecessary via efficient retrieval. When historical experience suggests unnecessary, a suppression signal redirects the model to proceed. Across multiple model and benchmarks, our approach reduces token usage up to 20.3% while maintaining the accuracy, and in some datasets even yields accuracy improvements.

In this study, we investigate how language models develop preferences for \textit{idiomatic} as compared to \textit{linguistically acceptable} Swedish, both during pretraining and when adapting a model from English to Swedish. To do so, we train models on Swedish from scratch and by fine-tuning English-pretrained models, probing their preferences at various checkpoints using minimal pairs that differ in linguistic acceptability or idiomaticity. For linguistic acceptability, we adapt existing benchmarks into a minimal-pair format. To assess idiomaticity, we introduce two novel datasets: one contrasting conventionalized idioms with plausible variants, and another contrasting idiomatic Swedish with Translationese. Our findings suggest that idiomatic competence emerges more slowly than other linguistic abilities, including grammatical and lexical correctness. While longer training yields diminishing returns for most tasks, idiom-related performance continues to improve, particularly in the largest model tested (8B). However, instruction tuning on data machine-translated from English -- the common approach for languages with little or no native instruction data -- causes models to rapidly lose their preference for idiomatic language.

LLM routing aims to achieve a favorable quality--cost trade-off by dynamically assigning easy queries to smaller models and harder queries to stronger ones. However, across both unimodal and multimodal settings, we uncover a pervasive yet underexplored failure mode in existing routers: as the user's cost budget increases, routers systematically default to the most capable and most expensive model even when cheaper models already suffice. As a result, current routers under-utilize small models, wasting computation and monetary cost and undermining the core promise of routing; we term this phenomenon routing collapse. We attribute routing collapse to an objective--decision mismatch: many routers are trained to predict scalar performance scores, whereas routing decisions ultimately depend on discrete comparisons among candidate models. Consequently, small prediction errors can flip relative orderings and trigger suboptimal selections. To bridge this gap, we propose EquiRouter, a decision-aware router that directly learns model rankings, restoring the role of smaller models and mitigating routing collapse. On RouterBench, EquiRouter reduces cost by about 17\% at GPT-4-level performance compared to the strongest prior router. Our code is available at https://github.com/AIGNLAI/EquiRouter.

Single-cell RNA-seq profiles are high-dimensional, sparse, and unordered, causing autoregressive generation to impose an artificial ordering bias and suffer from error accumulation. To address this, we propose scDiVa, a masked discrete diffusion foundation model that aligns generation with the dropout-like corruption process by defining a continuous-time forward masking mechanism in token space. ScDiVa features a bidirectional denoiser that jointly models discrete gene identities and continuous values, utilizing entropy-normalized serialization and a latent anchor token to maximize information efficiency and preserve global cell identity. The model is trained via depth-invariant time sampling and a dual denoising objective to simulate varying sparsity levels while ensuring precise recovery of both identity and magnitude. Pre-trained on 59 million cells, scDiVa achieves strong transfer performance across major benchmarks, including batch integration, cell type annotation, and perturbation response prediction. These results suggest that masked discrete diffusion serves as a biologically coherent and effective alternative to autoregression.

Object detection is pivotal in computer vision, yet its immense computational demands make deployment slow and power-hungry, motivating quantization. However, task-irrelevant morphologies such as background clutter and sensor noise induce redundant activations (or anomalies). These anomalies expand activation ranges and skew activation distributions toward task-irrelevant responses, complicating bit allocation and weakening the preservation of informative features. Without a clear criterion to distinguish anomalies, suppressing them can inadvertently discard useful information. To address this, we present InlierQ, an inlier-centric post-training quantization approach that separates anomalies from informative inliers. InlierQ computes gradient-aware volume saliency scores, classifies each volume as an inlier or anomaly, and fits a posterior distribution over these scores using the Expectation-Maximization (EM) algorithm. This design suppresses anomalies while preserving informative features. InlierQ is label-free, drop-in, and requires only 64 calibration samples. Experiments on the COCO and nuScenes benchmarks show consistent reductions in quantization error for camera-based (2D and 3D) and LiDAR-based (3D) object detection.

Deep Research (DR) agents extend Large Language Models (LLMs) beyond parametric knowledge by autonomously retrieving and synthesizing evidence from large web corpora into long-form reports, enabling a long-horizon agentic paradigm. However, unlike real-time conversational assistants, DR is computationally expensive and time-consuming, creating an autonomy-interaction dilemma: high autonomy on ambiguous user queries often leads to prolonged execution with unsatisfactory outcomes. To address this, we propose IntentRL, a framework that trains proactive agents to clarify latent user intents before starting long-horizon research. To overcome the scarcity of open-ended research data, we introduce a scalable pipeline that expands a few seed samples into high-quality dialogue turns via a shallow-to-deep intent refinement graph. We further adopt a two-stage reinforcement learning (RL) strategy: Stage I applies RL on offline dialogues to efficiently learn general user-interaction behavior, while Stage II uses the trained agent and a user simulator for online rollouts to strengthen adaptation to diverse user feedback. Extensive experiments show that IntentRL significantly improves both intent hit rate and downstream task performance, outperforming the built-in clarify modules of closed-source DR agents and proactive LLM baselines.

Integrating hard constraints into deep learning is essential for safety-critical systems. Yet existing constructive layers that project predictions onto constraint boundaries face a fundamental bottleneck: gradient saturation. By collapsing exterior points onto lower-dimensional surfaces, standard orthogonal projections induce rank-deficient Jacobians, which nullify gradients orthogonal to active constraints and hinder optimization. We introduce Soft-Radial Projection, a differentiable reparameterization layer that circumvents this issue through a radial mapping from Euclidean space into the interior of the feasible set. This construction guarantees strict feasibility while preserving a full-rank Jacobian almost everywhere, thereby preventing the optimization stalls typical of boundary-based methods. We theoretically prove that the architecture retains the universal approximation property and empirically show improved convergence behavior and solution quality over state-of-the-art optimization- and projection-based baselines.

Estimating treatment effects in networks is challenging, as each potential outcome depends on the treatments of all other nodes in the network. To overcome this difficulty, existing methods typically impose an exposure mapping that compresses the treatment assignments in the network into a low-dimensional summary. However, if this mapping is misspecified, standard estimators for direct and spillover effects can be severely biased. We propose a novel partial identification framework for causal inference on networks to assess the robustness of treatment effects under misspecifications of the exposure mapping. Specifically, we derive sharp upper and lower bounds on direct and spillover effects under such misspecifications. As such, our framework presents a novel application of causal sensitivity analysis to exposure mappings. We instantiate our framework for three canonical exposure settings widely used in practice: (i) weighted means of the neighborhood treatments, (ii) threshold-based exposure mappings, and (iii) truncated neighborhood interference in the presence of higher-order spillovers. Furthermore, we develop orthogonal estimators for these bounds and prove that the resulting bound estimates are valid, sharp, and efficient. Our experiments show the bounds remain informative and provide reliable conclusions under misspecification of exposure mappings.

Reinforcement learning with verifiable rewards (RLVR) is effective for training large language models on deterministic outcome reasoning tasks. Prior work shows RLVR works with few prompts, but prompt selection is often based only on training-accuracy variance, leading to unstable optimization directions and weaker transfer. We revisit prompt selection from a mechanism-level view and argue that an effective minibatch should provide both (i) a reliable positive anchor and (ii) explicit negative learning signals from rare failures. Based on this principle, we propose \emph{positive--negative pairing}: at each update, we sample a hard-but-solvable $q^{+}$ and an easy-but-brittle prompt $q^{-}$(high success rate but not perfect), characterized by low and high empirical success rates under multiple rollouts. We further introduce Weighted GRPO, which reweights binary outcomes at the pair level and uses group-normalized advantages to amplify rare successes on $q^{+}$ into sharp positive guidance while turning rare failures on $q^{-}$ into strong negative penalties. This bidirectional signal provides informative learning feedback for both successes and failures, improving sample efficiency without suppressing exploration. On Qwen2.5-Math-7B, a single paired minibatch per update consistently outperforms a GRPO baseline that selects two prompts via commonly used variance-based selection heuristics: AIME~2025 Pass@8 improves from 16.8 to 22.2, and AMC23 Pass@64 from 94.0 to 97.0, while remaining competitive with large-scale RLVR trained from a pool of 1209 training prompts. Similar gains are observed on Qwen2.5-Math-7B-Instruct.

Multi-subject image generation aims to synthesize images that faithfully preserve the identities of multiple reference subjects while following textual instructions. However, existing methods often suffer from identity inconsistency and limited compositional control, as they rely on diffusion models to implicitly associate text prompts with reference images. In this work, we propose Hierarchical Concept-to-Appearance Guidance (CAG), a framework that provides explicit, structured supervision from high-level concepts to fine-grained appearances. At the conceptual level, we introduce a VAE dropout training strategy that randomly omits reference VAE features, encouraging the model to rely more on robust semantic signals from a Visual Language Model (VLM) and thereby promoting consistent concept-level generation in the absence of complete appearance cues. At the appearance level, we integrate the VLM-derived correspondences into a correspondence-aware masked attention module within the Diffusion Transformer (DiT). This module restricts each text token to attend only to its matched reference regions, ensuring precise attribute binding and reliable multi-subject composition. Extensive experiments demonstrate that our method achieves state-of-the-art performance on the multi-subject image generation, substantially improving prompt following and subject consistency.

Lifelong learning is critical for embodied agents in open-world environments, where reinforcement learning fine-tuning has emerged as an important paradigm to enable Vision-Language-Action (VLA) models to master dexterous manipulation through environmental interaction. Thus, Continual Reinforcement Learning (CRL) is a promising pathway for deploying VLA models in lifelong robotic scenarios, yet balancing stability (retaining old skills) and plasticity (learning new ones) remains a formidable challenge for existing methods. We introduce CRL-VLA, a framework for continual post-training of VLA models with rigorous theoretical bounds. We derive a unified performance bound linking the stability-plasticity trade-off to goal-conditioned advantage magnitude, scaled by policy divergence. CRL-VLA resolves this dilemma via asymmetric regulation: constraining advantage magnitudes on prior tasks while enabling controlled growth on new tasks. This is realized through a simple but effective dual-critic architecture with novel Goal-Conditioned Value Formulation (GCVF), where a frozen critic anchors semantic consistency and a trainable estimator drives adaptation. Experiments on the LIBERO benchmark demonstrate that CRL-VLA effectively harmonizes these conflicting objectives, outperforming baselines in both anti-forgetting and forward adaptation.