Optimization
-
From Slices to Volumes: Comparing 2D, 2.5D, and 3D Residual Networks for Pulmonary Nodule Malignancy-Risk Classification
2026-10-25 10:03:36Bahodir Nazarov, Zineb Gharib, Fatima ZazoulAbstract
Purpose: To compare 2D, 2.5D, and 3D residual networks for pulmonary nodule malignancy-risk classification and evaluate momentum-contrast pretraining, mild augmentation, and exclusion of indeterminate ratings. Methods: A retrospectively prepared LIDC-IDRI cohort contained 2,654 nodules labeled benign, indeterminate, or malignant using median radiologist ratings. Patient-disjoint splits comprised 1,876 training, 382 validation, and 396 test nodules. Representations included one axial slice, five neighboring slices, and a local CT volume. Each track evaluated a randomly initialized baseline, momentum-contrast pretraining followed by fine-tuning, and supervised training with mild augmentation. Checkpoints were selected by validation Macro-F1. The 2D experiments used three seeds; 2.5D and 3D results were single-run estimates. A separate binary experiment excluded indeterminate nodules. Results: Augmented models achieved test Macro-F1 values of $0.6134\pm0.0075$ for 2D, $0.6537$ for 2.5D, and $0.5403$ for 3D; the 2D uncertainty denotes population standard deviation across seeds. Corresponding accuracies were $0.6128$, $0.6869$, and $0.5455$. Augmentation improved Macro-F1 within every track, whereas the tested pretraining recipes reduced it. Binary Macro-F1 values were $0.8227$, $0.8664$, and $0.7554$, respectively. Conclusion: Augmented 2.5D achieved the highest observed performance. Differences in preprocessing, optimization, capacity, and replication limit attribution to dimensionality. Labels reflect radiologist-assessed malignancy risk rather than pathology-confirmed cancer. Code and experimental results are available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.35960
PDF
-
Entropic Control for Homeostatic Optimization: Pre-empting Catastrophic Forgetting via Dynamic Feedback
2026-10-25 09:34:29Shangxi Wu, Xiaorong Yan, Xiaojun Meng, Yuxiang Zhang, Jiansheng WeiAbstract
Supervised Fine-Tuning (SFT) of LLMs suffers from catastrophic forgetting, and prevailing mitigation strategies struggle to reconcile the plasticity-stability dilemma. To address this, we propose Entropic Control for Homeostatic Optimization (ECHO), a closed-loop control framework. By decomposing the KL divergence, we identify output entropy as a dominant observable component of KL variation. This enables us to leverage entropy as a computationally efficient online proxy to monitor forgetting risks in real-time, bypassing expensive reference-model inference. Consequently, ECHO employs an elastic controller that dynamically modulates the gradient weight of replay data via a proportional control law, using deviations from the pre-trained entropic equilibrium as an error signal. Empirical evaluations across 0.6B to 9B parameter architectures demonstrate that ECHO lies on, and in several comparisons extends, the observed plasticity-stability frontier and maintains high operational robustness against varying data mixing ratios.
Abstract (translated)
URL
https://arxiv.org/abs/2610.34182
PDF
-
ASR-Protected Fusion and Reference-Anchored Policy Optimization for Joint ASR and Word-Level Speech Anti-Spoofing
2026-10-25 08:40:02Xueping Zhang, Xiaoxiao Miao, Tiantian Feng, Ming LiAbstract
Existing anti-spoofing systems report either an utterance-level score or a time-based spoof interval: the former cannot show which words are spoofed, while the latter does not directly match the transcript and must be post-aligned with words, so alignment errors can lead to wrong word-level labels. Standard ASR model outputs only text; spoof information in the audio is lost after recognition and is not passed with the transcript. Yet changing only a few words can change the meaning of a sentence, leaving downstream systems to use incorrect text without knowing which words are spoofed. We therefore treat partial anti-spoofing as tagged ASR: the model transcribes speech and marks spoofed words at the same time, directly linking spoof labels to output words. We adopt the audio LLM and use ASR-Protected Subspace Gated Fusion (APSGF) to fuse spoof cues with ASR features. Spoof cues are added only outside an ASR-protected subspace, so they do not disturb recognition features. After cross-entropy training, we fine-tune the model with Reference-Anchored Policy Optimization (RAPO). Since the SFT policy is already concentrated, multiple samples from the same input tend to receive similar rewards, resulting in weak group-wise advantages. RAPO instead generates a single sample with low sampling randomness and anchors its advantage to the reward of a reference model. This allows the model to reinforce responses that outperform the reference while suppressing those that underperform it. Experiments on five datasets demonstrate the effectiveness of our approach in jointly detecting spoofed words and preserving ASR accuracy. The code is publicly available.
Abstract (translated)
URL
https://arxiv.org/abs/2610.34156
PDF
-
Humans in Space: Can MLLMs Construct Human-Centric Spatial Understanding via Active Exocentric Exploration?
2026-10-25 08:35:20Tianchen Guo, Tengyue Zhang, Chen Liu, Ling Chen, Xin YuAbstract
While Multimodal Large Language Models (MLLMs) excel at zero-shot reasoning, their spatial capabilities are primarily evaluated through static, passive observation. In complex, human-centric scenes, occlusion often hides crucial geometric evidence, and passive benchmarks force models to guess rather than recognize their own visual uncertainty. To bridge this gap, we introduce SATE-Bench, a novel interactive benchmark comprising 316 real-world scenes and 2,525 multi-turn QA pairs to evaluate visual self-awareness and active exocentric view-exploration. It combines a Passive multi-view control with an Active exploration setting, where models start from a severely occluded view and must iteratively query adjacent viewpoints until they gather enough evidence to fully ``sate'' the question, scored by a step-penalized efficiency reward. Evaluating state-of-the-art MLLMs reveals that they terminate exploration prematurely while reporting high confidence, and struggle with fine-grained 3D geometric tasks; matched-view controls further attribute the Active--Passive gap primarily to insufficient evidence acquisition. Moreover, fine-tuning with Group Relative Policy Optimization (GRPO) shows that rewarding view-seeking increases active exploration but lowers overall accuracy, as models over-explore and fail to fuse expanding multi-view contexts across turns. These findings indicate that scaling passive multi-view perception does not intrinsically yield active spatial reasoning, highlighting the need for models that know where to look, how to integrate what they see, and when to stop. Code and data are available at this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2610.34149
PDF
-
Self-Evolution Reasoner: Continuous Optimization via Policy-Intrinsic Exploration and Exploitation
2026-10-25 08:18:39Wenhang Shi, Yiren Chen, Zhe Zhao, Jinhao Dong, Wei Lu, Xiaoyong DuAbstract
Reinforcement learning with verifiable rewards (RLVR) enables large language models (LLMs) to continuously self-improve their reasoning capabilities. Compared to RLVR's scalar rewards, recent self-supervised on-policy distillation methods leverage denser supervision signals. However, both paradigms suffer from exploration bottlenecks, struggling to escape local optima when the policy fails to sample valid solutions. Relying on external demonstrations can bypass this issue; however, the off-policy nature hinders genuine self-improvement and the mismatched-distribution reasoning can not be fully exploited. In this work, we propose Self-Evolution Policy Optimization (SEPO), a framework that drives continuous reasoning capability evolution via guided exploration and in-distribution exploitation. Specifically, SEPO utilizes labels to elicit correct reasoning from the policy, unlocking the exploration potential. By treating current policy conditioned on this in-distribution reasoning as a self-teacher, SEPO distills the newly discovered solution back into the policy. Through this iterative cycle of policy-intrinsic exploration and exploitation, SEPO achieves robust and continuous capability evolution. Across scientific, tool-use, and complex logical reasoning domains, SEPO consistently achieves the best performance under identical training time or steps, demonstrating superior efficiency and convergence ceilings. Notably, SEPO maintains stable improvements even in scenarios where baselines severely underperform or fail entirely, such as on complex tasks or with small-scale models. Our code is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.34137
PDF
-
T2D: Target-to-Draft Model Pruning & Adaptation for Efficient Speculative Decoding
2026-10-25 07:00:45Shahriar Kabir Nahin, Xiao Ye, Muhao Chen, Ben Zhou, Anshuman ChhabraAbstract
Speculative decoding significantly accelerates autoregressive text generation in Large Language Models (LLMs), but its efficacy depends largely on two key factors: the distributional alignment between the draft and the target model, and the draft model's latency during inference. However, existing approaches typically utilize independently pretrained small draft models or seek to prune the target model to a manually chosen architecture, thereby failing to explicitly balance the trade-off between draft model alignment and token generation efficiency. To address this gap, we formulate draft model design as an optimization problem and propose T2D, a framework that constructs draft models directly from the target model using latency-aware structured pruning and knowledge distillation. In this manner, T2D obtains draft models that are jointly optimized for both measured inference latency and cross-entropy loss on the target model's responses, resulting in superlative speculative decoding performance. Through extensive experiments on multi-domain speculative decoding benchmarks and across several target model families of differing scale, we demonstrate how T2D prunes up to 92\% of the target model to obtain highly performant draft models compared to competitive baselines. Moreover, draft models obtained via T2D can also be utilized by other speculative decoding token-alignment and knowledge distillation methods in a plug-and-play fashion, leading to further gains. Our findings underscore the importance of jointly navigating the draft model architecture-alignment frontier to unlock the full potential of speculative decoding for LLMs.
Abstract (translated)
URL
https://arxiv.org/abs/2610.34103
PDF
-
Hierarchical Cooperative GFlowNets
2026-10-25 00:46:45Pedro Dall'Antonia, Juan Belieni, Ana Carolina Erthal, Saullo H. G. de Oliveira, Eliezer de Souza da Silva, Diego MesquitaAbstract
Generative Flow Networks (GFlowNets) are trained to sample objects from compositional spaces with probability proportional to a reward function, but ineffective exploration can cause learned samplers to concentrate on a few high-reward regions. An intuitive remedy is to train multiple GFlowNets independently in parallel and combine their samples. However, this does not prevent multiple models from collapsing onto the same high-reward areas, leaving large regions of the state space unexplored. With this in mind, we propose Hierarchical Cooperative GFlowNets (HCG), an approach that dynamically partitions the search space to coordinate discovery without relying on hand-crafted boundaries. HCG imposes an ordering in which the highest-ranked model may explore the entire space, while lower-ranked models are progressively pushed toward regions not covered by higher-ranked ones. We provide a theoretical characterization of how specific learning objectives can shift probability mass away from designated areas, instantiating this mechanism within our architecture. A shared classifier coordinates the hierarchy by partitioning the search space, keeping the GFlowNets independently parameterized and enabling asynchronous training. Furthermore, we derive a normalizer-weighted aggregation rule for the ensemble and prove that it recovers the target reward-proportional distribution under the stated assumptions. Finally, experiments across multimodal distributions, sequence generation, combinatorial optimization, and biological sequence design show that HCG improves mode coverage and target approximation, translating increases in the number of coordinated samplers into broader, complementary exploration.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33968
PDF
-
Curriculum-Driven Deep Reinforcement Learning for Autonomous Far-Range Rocket Landing
2026-10-24 19:57:44Atin Chowdhury, Piyush JoshiAbstract
Autonomous vertical landing of reusable launch vehicles is a critical yet formidable challenge in aerospace engineering, primarily due to highly nonlinear dynamics, stringent precision require- ments, and inherent environmental uncertainties. While reinforcement learning (RL) has emerged as a promising alternative to traditional model-based control, existing approaches frequently rely on flat training paradigms and highly simplified initial conditions, drastically limiting their real-world applicability. This paper proposes a comprehensive curriculum-driven deep reinforcement learning framework designed to master both far-range navigation and precision landing. By employing a three-stage curriculum augmented with a critic warm-up procedure and a rehearsal mechanism to prevent catastrophic forgetting, alongside a high-dimensional observation vector utilizing a full 3 X 3 rotation-matrix, the agent is trained to navigate from an initial horizontal distance of 1,000 to 1,500 units vastly farther than prior benchmarks before executing a soft touchdown. Crucially, we intro- duce a redesigned reward function that reduces unnecessary computational overhead and addresses limitations of earlier formulations, including poorly calibrated reward components and potentially conflicting optimization objectives. The redesigned function bounds the per-step reward magnitude by (O(1)), while its progress term accumulates the total distance closed over the trajectory. The available 100,000-episode results report a 99.98% success rate for the proposed policy and 68.31% for the baseline under the evaluation conditions. In the evaluation, the proposed policy retained 69.1% of its initial fuel capacity and achieved a mean final distance from the pad centre of 7.07 units, compared with 13.10 units for the baseline. Landing visualizations are available in Video 1 and Video 2.
Abstract (translated)
URL
https://arxiv.org/abs/2610.35954
PDF
-
Stable Dimensionality Reduction via Isometric Centroid-Encoder
2026-10-24 19:26:33Tomojit Ghosh, Siva Rama Krishna Prasad Changala, Prathyusha Kanakamalla, Harigovind HarikumarAbstract
Many supervised dimensionality reduction methods promote intra-class compactness and inter-class separation but do not explicitly constrain class geometry, leading to variability in the relative arrangement of classes across independent runs. Such variability limits the reproducibility and interpretability of low-dimensional embeddings. We propose the Isometric Centroid-Encoder (ICE), a supervised dimensionality reduction framework in which class centroids are learned directly in the latent space while preserving a prescribed pairwise class-centroid geometry. Unlike existing centroid- or prototype-based methods that primarily encourage intra-class compactness, ICE explicitly regularizes the relative arrangement of latent class centroids. Unlike multidimensional scaling and many metric-learning approaches that preserve sample-level relationships, ICE operates on class centroids, preserving inter-class geometry without requiring large numbers of sample-pair or triplet constraints. We further show that the intrinsic dimensionality required to represent the centroid-distance configuration of $M$ classes is at most $M-1$, explaining the observed saturation of distance-preservation measures as the latent dimensionality increases. To quantify embedding reproducibility, we introduce the Centroid Neighborhood Stability (CNS) metric, which measures the consistency of class-level neighborhood topology across independent runs and complements existing embedding quality measures. Experiments on image and biological datasets demonstrate that ICE produces stable embeddings with improved preservation of class-centroid geometry while remaining computationally efficient. Additional analyses of optimization dynamics, sensitivity, and preprocessing further provide practical insights into latent dimensionality and explicit geometry modeling for supervised dimensionality reduction.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33872
PDF
-
Adapting DNNs to Embrace Voltage Scaling Induced Errors via Accumulation-Aware Quantization
2026-10-24 17:42:49Alexis Maras, Panagiotis Chaidos, Georgios Alexandris, Dimitrios Soudris, Sotirios XydisAbstract
The pervasive deployment of Deep Neural Networks (DNNs) across the computing spectrum has brought growing attention to energy efficiency, alongside raw performance. While voltage scaling is a potent knob for power reduction, its application in next-generation nodes is hampered by severe throughput degradation and susceptibility to process variation. This work aims to address these energy-reliability trade-offs by introducing a hardware-software co-design methodology, based on the observation that timing-robust bit ranges within the accumulator of a systolic array vary significantly across different voltage-frequency operating points. At the software level, we propose an Accumulation-Aware Training approach that incorporates this information as a first-order optimization constraint during training. By integrating mixed-precision quantization with unstructured pruning, we effectively constrain the model's weight distribution to avoid activating the hardware's sensitized high-rank carry chains. At the architectural level, we complement this with selective MSB-masking and fine-grained clock-gating to minimize unnecessary dynamic switching activity. Furthermore, we employ a layer-wise Dynamic Frequency Scaling strategy to further adapt to the accumulator's reduced dynamic range, based on the numerical sensitivity of each layer. We experimentally evaluate the proposed solution on a deep scaled 2nm technology node and demonstrate that, when operating at a low-voltage frontier, our co-optimized post-layout architecture delivers substantial energy and energy-delay product gains of 39% and 23% respectively, relative to the nominal voltage implementation, while maintaining accuracy degradation within 1%.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33824
PDF
-
Computer Vision Navigation for an Autonomous Vehicle Model Using Reinforcement Learning
2026-10-24 15:52:59Dinis Fernandes, Tiago Ribeiro, Ant\'onio Ribeiro, Carolina Lopes, Gil Lopes, A. Fernando RibeiroAbstract
This work aims to assess the capabilities of the reinforcement learning algorithm, Proximal Policy Optimization (PPO), in developing a policy capable of controlling the speed and steering angle of a vehicle model to navigate within simulated road lines. The control system's input is a camera, and its outputs are the rotational speed of the rear wheels and the steering angle of the front wheels, controlling the vehicle's speed and direction, respectively. The algorithm is trained and tested in a simulated environment, where it receives image frames from the vehicle's camera, processes them through a feature extraction algorithm, and feeds them into the policy. For each frame, the policy outputs speed and direction commands for the vehicle. The policy is iteratively updated to enhance its performance. The reinforcement learning algorithm successfully generates a policy capable of navigating the vehicle in a simplified environment. Potential improvements are suggested to enhance the system's robustness in more complex scenarios. Creating a simulation environment suitable for testing an advanced reinforcement learning algorithm is a complex task. This project offers a framework for other developers to test different approaches and algorithms. The tested algorithm, PPO, uses proximity intervals - an innovative approach in reinforcement learning. In some scenarios, the algorithm achieved better performance than predecessor algorithms, which are typically harder to implement and require tuning more hyperparameters.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33755
PDF
-
GE-Act 2.0: Pretraining and Scaling a World-Action Model for Robotic Manipulation
2026-10-24 15:41:00Renhang Liu, Wenzhi Zhao, Zhuo Yang, Liliang Chen, Pengfei Zhou, Shengcong Chen, Guanghui Ren, Youlun Peng, Rongjun Jin, Nan Wang, Sukai Wang, Xindong He, Jinyuan Feng, Ziyu Xiong, Linqing Zhong, Yifei Wei, Feng Han, Long Zhang, Da Huang, Nanshu Zhao, Chenghao Yin, Mo Wu, Zhaodong Yan, Kongtao Hu, Yuxiang Yan, Aogelijiang Niyazi, Yu Fang, Jia Zeng, Lizhu Meng, Daizhen Lv, Haoyu Cao, Zhiwen Hou, Lianjin Ye, Yuehan Niu, Zhikai Cai, Xuan Hu, Hui Min, Xiongfeng Cai, Yue Liao, Jing Wu, Soujanya Poria, Ye Li, Sanping Zhou, Maoqing Yao, AgiBot Research TeamAbstract
World-action models (WAM) predict future states to guide robot actions, enabling learning from both action-free video and action-labeled interaction. Most inherit pretrained video generators, leaving WAM pretraining and scaling underexplored. We introduce Genie Envisioner Act 2.0 (GE-Act 2.0), a world-action model whose trainable generative and action components are all initialized from scratch on manipulation data. It combines a control-oriented autoencoder (CoAE), a single-step visual planner (SVP), and an inverse dynamics model (IDM). CoAE retains action- and instruction-relevant information under aggressive compression, while SVP produces a complete future state in one differentiable pass, so visual planning and inverse dynamics can be pretrained separately on complementary data. The components are then jointly trained with knowledge-aligned selective optimization (KASO), which reduces mismatched supervision by selecting only predicted futures judged behaviorally compatible with the recorded action. We evaluate pretrained checkpoints directly, without per-task fine-tuning, on 100 tasks across 20 manipulation skill groups with held-out scenes, backgrounds, lighting, and object instances. Scaling co-training data from 300 to 30,000 hours raises success from 17.1% to 44.1% on G1-OP and from 13.4% to 31.1% on G2-90D; despite comprising less than 2% of the co-training data, G2-90D improves by 17.7 points, suggesting cross-embodiment transfer. Gains span 19/20 and 18/20 skill groups, and skill-specific coverage strongly correlates with zero-shot out-of-distribution (OOD) success (Pearson r=0.80; Spearman rho=0.85). Under the same protocol, the model grounds object, color, shape, and position references in at least 90% of trials and follows explicit instructions even when they conflict with an already-committed behavior or a conventional scene association.
Abstract (translated)
URL
https://arxiv.org/abs/2609.05588
PDF
-
E3tereo: Improving Stereo Matching at Depth Boundaries with Learned Geometric Edges
2026-10-24 14:35:22Qi Xiong, Qinghua Zhao, Yongqiang XiAbstract
Depth boundaries remain a major source of error in stereo matching. For IGEV on Scene Flow, geometric edge pixels account for only 6% of evaluated pixels but about half of the total absolute disparity error. Photometric edges can respond to texture changes without corresponding depth discontinuities, making them unreliable cues for boundary refinement. We propose E3Stereo, which learns geometric edges from disparity supervision and combines these boundary cues with the backbone disparity and image details to predict full-resolution disparity corrections. The same edge map also emphasizes boundary errors during training, while joint optimization allows the edge predictor to adapt to disparity refinement. On the Scene Flow dataset, E3Stereo reduces end-point error (EPE) at depth boundaries by 9.5%-14.3% across RAFT-Stereo, IGEV, and MonSter, while also reducing overall EPE. After fine-tuning, our IGEV-based model improves upon IGEV on all eleven reported metrics across ETH3D, KITTI 2012, and KITTI 2015. The source code will be made publicly available upon acceptance.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33704
PDF
-
PACE: Research Proving Agent for Adaptive Conjecture Exploration
2026-10-24 14:13:06Shu Miao, Feiming Wang, Wentao Long, Chenyi Li, Kun Yuan, Zaiwen WenAbstract
Many mathematical research agents focus on solving problems with fixed statements. However, conjecture research, especially in applied mathematics, may require revising candidate questions or conclusions as investigation proceeds, coupling problem formulation with proof development. We present PACE, a research-proof agent that supports this process through a structured research loop that selects mathematical objectives and develops or challenges candidate mechanisms. The whole prover constructs proofs from these mechanisms, and an independent verifier checks them against the original target. Evidence-tracking memory stores assumptions and dependencies arising during research and marks evidence for revalidation when a candidate or condition changes. This memory supports task-specific context composition that keeps checked premises, exploratory material and unresolved obligations distinct. To evaluate proof construction and conjecture research, we introduce ResearchProofBench, an applied-mathematics benchmark focused on optimization, comprising 51 known-proof tasks with 331 hidden checkpoints under historical abstract access and six open-problem cases. The known-proof tasks assess reconstruction of published results, while the open-problem cases assess mathematical investigation and proof development. On the 51 known-proof tasks, PACE achieves an aggregate Proof Score of 75.16 and completes 20 proofs. In conjecture research, PACE develops conditional results, with its submissions receiving higher review scores than the compared baselines.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33690
PDF
-
Meta-CLIP: Metadata-Conditioned Contrastive Learning for Zero-Shot Disease Inference
2026-10-24 13:45:22Haozhe Luo, Nirosh Sivanesan, Shelley Zixin Shu, Francesco Nitti, Mauricio ReyesAbstract
Medical vision--language pretraining learns transferable representations from paired images and radiology reports, yet differences in scanners and acquisition protocols can impair transfer. Existing approaches primarily enrich clinical semantics, leaving structured acquisition information in DICOM headers underused. We introduce Meta-CLIP, a metadata-conditioned contrastive framework that uses acquisition similarity to reweight cross-modal negatives during pretraining. A kernel over numerical and categorical DICOM fields defines a bounded pairwise margin. This margin increases the contribution of acquisition-similar negatives, emphasizing distinctions among studies that share imaging conditions. The kernel remains fixed during image--text optimization, and inference follows the standard dual-encoder architecture without requiring metadata. We evaluate Meta-CLIP on chest radiography and volumetric chest CT through zero-shot classification, appearance perturbations, grounding, and representation analyses. On CT-RATE, Meta-CLIP improves macro-AUROC over CT-CLIP by 1.41 points, with a 95\% paired bootstrap confidence interval of 0.96--1.85 points. It also achieves the highest macro-AUROC on four of five CXR benchmarks. An exploratory BrgSA evaluation extends the analysis to another CT backbone. These findings support acquisition metadata as a practical training-time prior for medical vision--language learning. Code will be made publicly available upon publication.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33674
PDF
-
TAD: Distilling Vision-Language-Action Policies from Teacher Trajectories Alone
2026-10-24 13:05:18Yun Zhu, Da WangAbstract
Online distillation of vision-language-action (VLA) policies commonly relies on teacher action scores that an inference-only interface may not expose. Trajectory-only Adversarial Distillation (\gad{}) instead turns a fixed bank of teacher joint-position features into adaptive feedback: a discriminator compares feature marginals from teacher and student rollouts and supplies rewards for on-policy learning. The student interacts with the environment without online teacher queries or environment task rewards. On ManiSkill \texttt{PutOnPlateInScene25Main}, \gad{} improves OpenVLA-OFT (7B) from $26.2\%$ to $76.6\%$ \successonce{} ($+50.4$ percentage points) and from $20.3\%$ to $58.6\%$ \successend{}. It trails white-box on-policy distillation (OPD) at $81.6\%$ / $59.4\%$ by $5.0$ / $0.8$ percentage points ($n{=}256$ evaluation trajectories per checkpoint). The tested four-epoch continuation achieves the highest \gad{} result without additional teacher data. This best run uses $5\times$ OPD's student interaction budget and $12.5\times$ its complete-rollout batch-passes. The comparison demonstrates online VLA improvement without teacher scoring while quantifying unequal interaction and optimization budgets.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33650
PDF
-
$xi$-DPO: Direct Preference Optimization via Ratio Reward Margin
2026-10-24 12:03:18Zhengyuan Fan, Zhonghua Wu, Yuxuan Du, Qun ChenAbstract
Reference-free preference optimization has emerged as an efficient alternative to reinforcement learning from human feedback, with Simple Preference Optimization(SimPO) demonstrating strong performance by eliminating the explicit reference model through a simple objective. However, the joint tuning of the hyperparameters $\beta$ and $\gamma$ in SimPO remains a central challenge. We argue that this difficulty arises because the margin formulation in SimPO is not easily interpretable across datasets with different reward gap structures. To better understand this issue, we conduct a comprehensive analysis of SimPO and find that $\beta$, rather than merely serving as a reward scaling factor, primarily functions to filter noisy samples, while the reward margin $\gamma$ defined by SimPO is not semantically well aligned with explicitly enlarging the reward gap between chosen and rejected responses. Motivated by these observations, we propose $\xi$-DPO: Direct preference optimization via ratio reward margin. We first reformulate the preference objective through an equivalent transformation, changing the optimization target from maximizing the likelihood of reward gaps to minimizing the distance between reward gaps and optimal margins. Then, we redefine the reward in a ratio form between the chosen and rejected, which effectively cancels the effect of $\beta$ and yields a bounded and interpretable margin. This margin is called the ratio reward margin and is denoted by $\xi$. Unlike the $\beta$ and $\gamma$ in SimPO, $\xi$ can be determined from the initial reward gap distribution, thus much easier to tune. Our extensive experimental results show that $\xi$-DPO substantially outperforms existing preference optimization methods across open benchmarks on multiple evaluation metrics.
Abstract (translated)
URL
https://arxiv.org/abs/2605.10981
PDF
-
Graph-dependent pooling: adaptive granularity in hierarchical graph neural networks
2026-10-24 11:54:16Bohan Yoon, Hyejin JangAbstract
Graph Neural Networks are widely used to learn structural representations of relational data. Hierarchical graph pooling learns multi-level representations by progressively coarsening graphs; however, prior methods rely on predefined cluster numbers or fixed pooling ratios, limiting adaptability to individual graphs. We propose Graph-Dependent Pooling (GDPool), which dynamically determines pooling granularity for each graph. We analyze optimization conflicts among layer-wise objectives through gradient alignment, showing that independently supervised pooling levels can send conflicting gradients to shared parameters. GDPool uses self-attention to select graph-specific anchors representing structurally important nodes and constructs coarse graphs through soft node assignment. Experiments on metal-organic framework and molecular graph benchmarks show that GDPool consistently achieves the best predictive performance across all benchmarks. Evaluations using modularity and the effective number of clusters show that GDPool achieves consistently higher modularity than existing methods while the number of active clusters decreases monotonically with depth. Sensitivity analyses confirm stable behavior across hyperparameter and architectural choices.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33623
PDF
-
HA-DiPO: Diffusion Policy Optimization for Heterogeneous Cooperative Transport
2026-10-24 10:24:37Xuebin Li, Jinpeng Xiao, Zong Chen, Pierluigi Arpenti, Bruno Siciliano, Jiebao Sun, Yiqun LiAbstract
Cooperative transport with heterogeneous legged robots requires coordinating different motion capabilities while supporting an unfixed shared payload. We present HA-DiPO (Heterogeneous-Agent Diffusion Policy Optimization), a hierarchical multi-agent reinforcement learning framework under centralized training with decentralized execution (CTDE). The framework decouples team-level coordination from embodiment-specific motion control: at the high level, a centralized critic uses team-value estimates to jointly train independent coordination policies, aligning robot-specific action responses with a shared transport objective; at the low level, embodiment-specific robust motion policies track the coordination commands. The high-level diffusion policies offer more flexible action distributions than single-Gaussian parameterizations, allowing potentially multimodal coordination responses to be represented. The coordination policies learn from scratch through reward-driven simulator interaction without demonstrations. At execution, each robot generates coordination commands using only local robot and payload observations, without partner-state inputs or inter-robot messages. Simulation comparisons show that HA-DiPO achieves higher final mean returns and velocity-tracking success rates, as well as lower velocity errors, than the evaluated baselines. Real-world experiments further demonstrate quadrupedal-wheeled-legged and human-robot cooperative transport.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33584
PDF
-
Distilling Dense Trajectory Teachers For Large-Motion Video World Models
2026-10-24 09:35:13Shuofeng Sun, Linqing Zhao, Haibin YanAbstract
Video world models are expected to generate videos that accurately follow specified camera trajectories while maintaining consistent 3D structures. However, existing models are predominantly trained on densely sampled videos with small inter-frame camera motions. When applied to sparse, large step trajectories, they often produce frozen frames, inaccurate camera motion, and severe geometric drift. We attribute this degradation to a mismatch between the gradual viewpoint transitions seen during training and the large viewpoint changes required at inference time. To address this problem, we propose Dense Trajectory Teacher Distillation (DTTD), a post-training framework that transfers a model's generation capability from dense trajectories to large camera motions. Given a sparse trajectory, DTTD interpolates additional camera poses to construct a dense trajectory covering the same path. The frozen base model then generates reliable teacher videos along this familiar dense trajectory, and pose aligned teacher observations are used to evaluate candidate videos generated under the original sparse trajectory. In this way, the model serves as its own generation teacher, requiring neither additional video data nor a stronger external video generator. We further combine dense teacher supervision with geometry aware rewards that assess cross frame consistency and camera trajectory following, and use the resulting preferences for offline preference optimization. To stably expand the model's motion range, we introduce a cumulative curriculum that progressively incorporates larger motion steps while retaining simpler trajectories to reduce capability forgetting.
Abstract (translated)
URL
https://arxiv.org/abs/2610.33566
PDF
