arXiv subjects

Ningxin Su

Publications and source records attributed to Ningxin Su.

12 recordsLinked to original sources

arXiv2026

Beyond Appearance Shifts: Task-Semantic Action Calibration for VLA Models

Vision-language-action (VLA) models have achieved strong performance in embodied manipulation, but still struggle to distinguish between changes that should preserve behavior and those that require different actions. We identify two complementary failure modes. Under task-preserving changes, where task semantics remain unchanged but scene appearance varies (e.g., style, illumination, clutter, or paraphrasing), policies often exhibit unnecessary action drift. Conversely, under semantic-breaking changes, where key task semantics such as the target object or constraint are altered, policies frequently continue following the original trajectory instead of adapting to the new task. To address these failures, we propose BAS-VLA, a task-semantic action calibration framework built on top of a frozen base VLA. BAS-VLA adopts a breaking-centered calibration core as its default path, complemented by a selective evidence-gated preserving auxiliary that corrects appearance-induced drift when task-preserving evidence is strong. Across four target-object swap tasks, BAS-VLA improves changed-task success from 2.3% to 68.5% while maintaining strong clean and semantics-preserving performance and suppressing outdated behavior. On validated style-preserving shifts, it further improves success from 42% to 70% without degrading clean performance. These results demonstrate the importance of calibrating VLA policies to remain stable when task semantics are preserved and adapt when task semantics change.

cs.RO↗

arXiv2026

VLA-Fabric: Learning Interaction for Composable Multi-Arm VLA Agents

Vision-language-action (VLA) models are rapidly expanding what individual robots can accomplish, yet many real-world manipulation settings require multiple arms to operate as a coordinated system. Existing approaches organize multi-arm coordination in different ways, including centralized joint policies, high-level task or role assignment, and locally executed policies that react to teammates. However, how complete action-producing VLAs should explicitly interact with one another during action generation remains largely unexplored. We introduce VLA-Fabric, which organizes one complete VLA per arm while preserving local perception and action ownership, and enables coordination through learned interaction across policy boundaries. We first systematically study where such interaction should occur, how peer information should be aggregated, and how the interaction should be learned, revealing complementary coordination functions spanning shared context, peer-specific latent information, and action-stage interaction. We then re-instantiate these functions in a structurally different VLA architecture, showing that the coordination principle is not tied to a particular tensor layout or backbone implementation. Finally, we compose teams of three and four arms and evaluate them in simulation and physical experiments. On all four physical tasks, Full interaction improves success over independently trained local policies by 16--40 percentage points. Together, these results suggest that explicit interaction among complete VLA policies offers a scalable path toward networked embodied intelligence.

cs.RO↗

arXiv2026

PortBench: A Correlation-Aware, Full-Pipeline Benchmark for LLM-Driven Portfolio Management

Large language models (LLMs) have shown strong performance across diverse financial tasks, yet portfolio management (PM) remains poorly evaluated. Existing benchmarks exhibit two gaps: they are often equity-only and ignore cross-asset correlations; they provide limited insight into where errors arise and how they affect subsequent decisions. We introduce PortBench, a benchmark with market data spanning six asset classes from 2015 to 2025. PortBench comprises a static QA dataset of 6,269 questions across seven task templates and a dynamic five-stage portfolio evaluation. Dynamic evaluation uses a class-level correlation score for allocation and CEPS to summarize reference-based stage scores. We further evaluate under three stress windows and three risk profiles, and support real-time evaluation on newly arriving market data. Evaluating ten LLMs reveals a disconnect: high financial QA scores do not ensure consistent portfolio outperformance. LLMs concentrate exposure in correlated assets, and their portfolios achieve higher Sharpe than equal weighting in 54.2\% of 120 evaluations across four market periods. Our source code is available at \href{https://github.com/AgenticFinLab/portbench}{this https URL}.

cs.AI↗

arXiv2026

MimicX: Policy-in-the-Loop Supervision Refinement for Video-Driven Humanoid Motion Tracking

Human videos provide rich motion targets for humanoid learning, yet visually plausible references can still produce persistent failures under physics-based execution. These failures reveal where training supervision should change. We present MimicX, a policy-in-the-loop framework that uses execution feedback to refine video-driven humanoid motion tracking. Starting from reconstructed and retargeted motion, MimicX localizes difficult transitions and affected body regions, then jointly adapts tracking objectives and the reset curriculum for policy continuation. Repeated rollout verification selects execution-priority improvements subject to tracking guards. Across four core video tasks, MimicX consistently improves tracking accuracy and Robust Execution Horizon relative to the Fixed Reference baseline. Task-averaged results show a 25.7% reduction in body-tracking error and a 255.6% increase in execution horizon. Additional video, motion-reference, and collision-scene studies evaluate the method beyond the core tasks, while MimicX-HLoop accelerates feedback through heterogeneous execution. Overall, MimicX turns policy failure into actionable supervision for deciding what to refine and which refinement to retain.

cs.RO↗

arXiv2026

EnvSimBench: A Benchmark for Evaluating and Improving LLM-Based Environment Simulation

Scalable AI agents training relies on interactive environments that faithfully simulate the consequences of agent actions. Manually crafted environments are expensive to build, brittle to extend, and fundamentally limited in diversity. A promising direction is to replace manually crafted environments with LLM-simulated counterparts. However, this paradigm hinges on an unexamined core assumption: LLMs can accurately simulate environmental feedback. In practice, LLM-simulated environments suffer from hallucinations, logical inconsistencies, and silent state drift failures that corrupt agent reward signals and compound the construction costs that the paradigm was designed to eliminate. To address this gap, we propose EnvSimBench with four contributions: 1) We provide the first formal definition and operationalization of Environment Simulation Ability (EnvSim Ability) as a quantifiable research objective. 2) We construct EnvSimBench, a rigorous benchmark covering 400 samples across 167 diverse environments, equipped with verifiable labels and fine-grained difficulty stratification along three axes. 3) Systematic evaluations reveal that all state-of-the-art language models suffer from a universal state change cliff: they achieve near-perfect accuracy on tasks when the environment state remains invariant, yet fail catastrophically when multiple states need simultaneous updates. This finding exposes EnvSim Ability as a critical yet largely unaddressed capability gap. 4) We design a constraint-driven simulation pipeline that substantially reduces hallucination, boosts environment synthesis yield by 6.8%, and cuts costs by over 90%. Overall, EnvSimBench serves as both a diagnostic framework and a practical optimization path for reliable LLM-based environment simulation, establishing a foundation for scalable agent training. Code and data are available at https://github.com/cookieApril/EnvSimBench

cs.AI↗

arXiv2026

D-JEPA: A Decision-Aligned Latent World Model

Latent world models predict the consequences of actions, but accurate prediction does not guarantee that latent distance reflects which candidate will execute successfully. We identify a decision-local prediction gap: among the few futures competing for execution, a candidate predicted closer to the goal can produce a worse realized outcome than an available alternative. We introduce D-JEPA, a decision-aligned latent world model that learns decision-relevant relations among candidate futures from executed outcomes. A bounded, permutation-equivariant operator jointly reasons over goal-relative predictive features and ordinal evidence, refining pretrained predictive geometry where action choices are most consequential. Restricted predictor adaptation and a shared ordinal interface extend this alignment across complementary predictive geometries. D-JEPA further realizes the learned decision structure in JEPA-compatible future representations, enabling deployment through native latent-distance planning. Evaluations across latent control, manipulation, pretrained action-producing models, physical robots and autonomous driving demonstrate improved action selection, including 87.89% success on PushT, a 15.04-point average gain on RoboTwin, and a 17-point gain on physical robot tasks. These results establish decision-relevant relational structure as a direct bridge between predictive world modeling and effective control.

cs.RO↗

arXiv2026

How Wrong Can a Good Predictor Be? Diverging Updates with Vanishing Predictive KL

Accurate posterior prediction need not require accurate approximation of Bayesian updates. We prove that an unbounded gap between the update maps can coexist with vanishing predictive KL for every fixed finite $K\ge2$ in a stationary symmetric Gaussian HMM. Exact Bayesian mixing and an explicit deterministic radial filter act on the same $K-1$ belief coordinates. As $q\to0^+$, their separation in centered logits in the worst case grows at least linearly in the natural confidence scale $L_K(q)$, while their categorical $D_{\mathrm{KL}}(\mathrm{exact}\|\mathrm{radial})$ vanishes at the same explicit witness. Along stationary HMM trajectories, the expected terminal KL between filtered posteriors also converges to zero at $H(q)=\lceil-\log(q)/c\rceil+1$. Typical blocks without switches drive both filters into a common confidence cone, where softmax curvature suppresses their disagreement; a single Gaussian maximal event controls adaptive noise. A sweep with equally spaced Gaussians over $K\in\{2,4,8\}$ illustrates the opposing trends, and binary controls at long horizons compare saturating and nonsaturating recurrences. The result isolates two missing links between internal update gaps and predictive cost: the contribution of separating states to expected loss and decoder sensitivity. Thus even an unbounded internal update gap does not by itself certify predictive failure. The construction is fixed in $K$ and does not provide a universal criterion for when compression is harmless or characterize when internal gaps must incur task loss.

cs.LG↗

arXiv2026

CoWAM: Coordination Contracts for Selective Policy Intervention with WAMs

World Action Models (WAMs) augment robot policies with action-conditioned predicted futures, but a plausible future alone does not justify changing the action that a bimanual policy would execute. We present CoWAM, a selective intervention layer that expresses synchronization, role compatibility, and collision convergence as coordination contracts. Each contract combines typed admissibility checks with event-conditioned verification and calibrated intervention gates. CoWAM preserves the nominal action unless an alternative satisfies every active obligation and provides a clear, low-risk improvement; when the nominal action is also inadmissible, it invokes a predefined abstention fallback. To separate selector quality from proposal quality, all methods operate on identical candidate pools and commit their decisions before shared oracle labeling. Across eight simulated bimanual tasks, CoWAM improves coordination-valid selection by 16.7 percentage points over the contract-only variant and raises closed-loop success by 9.6 percentage points over the strongest selective baseline, while keeping harmful interventions below 1%. Together, these results establish coordination contracts as an effective interface for conservative policy intervention with predicted world-action evidence across coordination-rich bimanual tasks.

cs.RO↗

arXiv2026

When Replanning Becomes the Bottleneck: Budgeted Replanning for Embodied Agents

Embodied agents replan frequently to recover from execution drift, partial observability, and coordination hazards, but each LLM-based replanning call can consume an accumulated textual context that grows over time and across agents. Once this context becomes large, replanning latency develops heavy tails and can miss real-time deadlines even when task success remains high, a failure mode that is hard to detect from average latency or success alone. We present BRACE, a controller that formulates replanning as a budgeted control loop by deciding whether to replan, selecting a replanning mode, and allocating an explicit token budget and latency service-level objective (SLO) while accounting for optional efficiency modules. As a reusable component, we introduce E-RECAP, a cost-aware progressive token pruning method that predicts token utility and prunes replanning contexts across transformer layers while preserving critical head and tail tokens. Across Meta Habitat, RoboFactory, and AirSim, BRACE with E-RECAP reduces replanning-call token counts by 62-92% and SLO violation rates from 85.5-100.0% to 4.7-50.0% in settings where task success is already saturated. In a harder RoboFactory setting where open-loop, frozen-plan, and No BRACE all fail, BRACE + E-RECAP reaches 80.0% success with 4.6% SLO violations, demonstrating that tail-aware per-call budgeting is effective across embodied platforms.

cs.RO↗

arXiv2026

When Does Multi-Agent Collaboration Help? An Entropy Perspective

Multi-agent systems (MAS) have emerged as a prominent paradigm for leveraging large language models (LLMs) to tackle complex tasks. However, the mechanisms governing the effectiveness of MAS built upon publicly available LLMs, specifically the underlying rationales for their success or failure, remain largely unexplored. In this paper, we revisit MAS through the perspective of \textit{entropy}, considering both intra- and inter-agent dynamics by investigating entropy transitions during problem-solving across various topologies, six reasoning benchmarks, and two agentic tasks. By analyzing 245 features spanning token-, agent-, and round-level entropy, we counterintuitively find that a single agent outperforms MAS in approximately 43.3\% of cases, and that entropy dynamics are largely determined during the first round of interaction. Furthermore, we provide three key observations: 1) \textit{Certainty Preference}: peak entropy directly harms and stable entropy directly benefits MAS correctness; 2) \textit{Base Entropy}: base models with lower entropy during problem-solving causally drive MAS performance; and 3) \textit{Task Awareness}: entropy dynamics of MAS play varying roles across different tasks. Building on these insights, we introduce a simple yet effective algorithm, the \textit{Entropy Judger}, to select solutions from MAS's pass@$k$ results, leading to consistent accuracy improvements across all MAS configurations and tasks. Our source code is available at \href{https://github.com/AgenticFinLab/multiagent-entropy}{this https URL}.

cs.MA↗

arXiv2025

GradES: Significantly Faster Training in Transformers with Gradient-Based Early Stopping

Early stopping monitors global validation loss and halts all parameter updates simultaneously, which is computationally costly for large transformers due to the extended time required for validation inference. We propose \textit{GradES}, a novel gradient-based early stopping approach that operates within transformer components (attention projections and Feed-Forward layer matrices). We found that different components converge at varying rates during fine-tuning for both language and vision-language models. \textit{GradES} tracks the magnitude of gradient changes in backpropagation for these matrices during training. When a projection matrix's magnitude of gradient changes fall below a convergence threshold $τ$, we exclude that projection matrix from further updates individually, eliminating costly validation passes while allowing slow converging matrices to continue learning. \textit{GradES} speeds up training time by 1.57--7.22$\times$ while simultaneously enhancing generalization through early prevention of overfitting, resulting in 1.2\% higher average accuracy in language tasks and 3.88\% on multimodal benchmarks.

cs.LG↗

arXiv2024

Calibre: Towards Fair and Accurate Personalized Federated Learning with Self-Supervised Learning

In the context of personalized federated learning, existing approaches train a global model to extract transferable representations, based on which any client could train personalized models with a limited number of data samples. Self-supervised learning is considered a promising direction as the global model it produces is generic and facilitates personalization for all clients fairly. However, when data is heterogeneous across clients, the global model trained using SSL is unable to learn high-quality personalized models. In this paper, we show that when the global model is trained with SSL without modifications, its produced representations have fuzzy class boundaries. As a result, personalized learning within each client produces models with low accuracy. In order to improve SSL towards better accuracy without sacrificing its advantage in fairness, we propose Calibre, a new personalized federated learning framework designed to calibrate SSL representations by maintaining a suitable balance between more generic and more client-specific representations. Calibre is designed based on theoretically-sound properties, and introduces (1) a client-specific prototype loss as an auxiliary training objective; and (2) an aggregation algorithm guided by such prototypes across clients. Our experimental results in an extensive array of non-i.i.d.~settings show that Calibre achieves state-of-the-art performance in terms of both mean accuracy and fairness across clients. Code repo: https://github.com/TL-System/plato/tree/main/examples/ssl/calibre.

cs.LG↗