Searcharxiv⌕ Search

arXiv subjects

Xiao-Ming Wu

Publications and source records attributed to Xiao-Ming Wu.

At least 19 recordsLinked to original sources
arXiv2026

Where Do Vision Language Models Fail at 3D Reasoning?

Vision-language models can get individual judgments right while struggling to recover and consistently use the exact properties of a 3D object. To characterize this gap, we introduce BedBench, a controlled benchmark of thirteen procedurally generated tasks spanning individual judgments and composite answers. Across twelve open-weight and four frontier models, per-color counting exposes a pronounced weakness: humans achieve 93.3% accuracy, compared with 7.2% for the best model. Targeted fine-tuning raises one model's accuracy from 3.0% to 68.6%, but reveals a further difficulty: its answers about the most common colors remain inconsistent with its own per-color counts. Supplying those counts as text improves its judgments, showing that recovering a property does not mean using it consistently. Reasoning, scaling, and added visual cues each improve individual judgments, but none makes exact answers reliable. To locate the difficulty, we probe Qwen3.5-27B: exact per-color counts remain recoverable at the language model's input, yet the model's answers rarely recover them (65.0% for the probe vs. 3.0% for the model). Image-removal controls further show that some apparent successes require no visual evidence. These findings motivate evaluating 3D reasoning by exact correctness, consistent use, and image dependence, rather than accuracy alone. Code and dataset are publicly available: https://github.com/leduckhai/BedBench

cs.CV↗
arXiv2026

ThinkLocoManip: A Whole-Body Control Agentic Harness for Legged Loco-Manipulation

In this paper, we propose ThinkLocoManip, an agentic harness for legged loco-manipulation designed to tackle complex, long-horizon tasks in large-scale environments. Existing approaches typically follow either direct action generation or a plan-then-act paradigm, both of which can struggle with the generalization and error-recovery capabilities required for extended task execution. In contrast, ThinkLocoManip adopts an interleaved reasoning-and-acting paradigm, allowing the agent to continuously reason about its current state, execute actions, observe their outcomes, and adapt its strategy accordingly. Building on this paradigm, our agentic harness orchestrates a suite of specialized tools for perception, navigation, manipulation, and locomotion, enabling flexible composition of heterogeneous capabilities. Unlike existing robotic agentic harnesses, ThinkLocoManip explicitly incorporates locomotion as a tool domain for agent reasoning and execution. We further design a set of skills that explicitly account for both tool usage within each functional domain and coordination across domains, supporting effective execution of complex behaviors over long horizons. To tackle task failure during long-horizon interactions, we develop keyframe thinking mechanism to preserve critical execution context and support efficient error recovery. Extensive real-world experiments demonstrate that ThinkLocoManip substantially outperforms representative methods based on both direct action generation and plan-then-act paradigms across challenging long-horizon legged loco-manipulation tasks. Notably, ThinkLocoManip achieves a success rate of 77%, more than twice that of the strongest plan-then-act baseline (37.1%). Our experiments further validate the effectiveness of interleaved reasoning and acting, tool orchestration, and coordinated skill execution.

cs.RO↗
arXiv2026

MUSE: Benchmarking Assembly Structure and Interface Geometry in Multi-Component Text-to-CAD Generation

Large language models (LLMs) have recently advanced text-driven 3D generation, yet generating CAD models that satisfy practical design requirements remains challenging. Existing benchmarks focus primarily on single-part CAD models and geometric similarity, providing limited assessment of assembly structure and interface geometry. To address this gap, we introduce MUSE, a Text-to-CAD benchmark for generating multi-component, editable boundary representation (B-Rep) assemblies. MUSE pairs practical design instances with structured \textit{Design Specifications} and evaluates generated outputs across four dimensions: code executability, geometric validity, structural and interface fidelity, and functional consistency. We further introduce Structure and Interface Evaluation (SIE), which combines reference Physical Assembly Graphs with direct B-Rep analysis to assess component correspondence, connectivity, interface types, and interface clearances. Experiments on proprietary and open-weight LLMs show that high execution success does not necessarily translate into correct assembly structure or interface geometry. Together, MUSE provides a benchmark and evaluation framework for studying multi-component Text-to-CAD generation beyond shape similarity. Our project website, including the leaderboard, dataset, and code, is available at \url{https://dong7313.github.io/muse-benchmark/}.

cs.AI↗
arXiv2026

Gödel Forest: Balancing Search Depth and Breadth for Data-Centric Recursive Self-Improvement

Recursive self-improvement (RSI) aims to achieve compounding gains by having models improve themselves. While most existing RSI systems optimize external agent harnesses or prompts around a frozen base model, data-centric RSI directly updates the model's own parameters by training on agent-generated data. However, because validating data strategies requires expensive model training, existing methods face a fundamental dilemma: a single agent gets trapped in narrow directions and lacks exploration breadth, while naive parallel search or heavy trace sharing sacrifices long-horizon search depth. To address this challenge, we introduce G"odel Forest, a multi-agent framework that organizes recursive self-improvement as an ensemble of co-evolving search trees. In G"odel Forest, each agent autonomously grows a persistent tree, deepening, branching, or pruning data strategies based on model feedback to secure depth, while parallel trees explore distinct regions of the data space to expand breadth. Crucially, rather than leaving trees isolated or flooding them with heavy execution logs, a dynamically co-evolving memory connects the forest: agents continuously distill their successes and failures into compact procedural lessons anchored to a global leaderboard. Through this forest ecosystem, a dead-end in one tree instantly warns the whole forest against unpromising paths, while an empirical breakthrough quickly seeds new exploration branches in neighboring trees. Evaluated on RSIBench-Data across six diverse domains, G"odel Forest outperforms the single-agent baseline by an average of 10.70% while reducing wall-clock time on five tasks. Ablations confirm that co-evolving shared memory yields a +7.00% gain over independent parallel search, demonstrating that collective distillation is key to scalable self-improvement. The code is available at https://github.com/evolvent-ai/Godel-Forest.

cs.CL↗
arXiv2026

VLANeXt Family: A Systematic Study of VLA Models from Core Recipes to Emerging Paradigms

Following the rise of large foundation models, Vision-Language-Action models (VLAs) emerged, leveraging strong visual and language understanding from Vision-Language Models for general-purpose policy learning. Yet, the current VLA landscape remains fragmented and exploratory. Although many groups have proposed their own VLA models, inconsistencies in training protocols and evaluation settings make it difficult to identify which design choices truly matter. To bring structure to this evolving space, we reexamine the VLA design space under a unified framework and evaluation setup. Starting from a simple VLA baseline similar to RT-2, which is the origin of VLA, we systematically dissect design choices along three dimensions: foundational components, perception essentials, and action modeling perspectives. From this study, we distill 12 key findings that together form a practical recipe for building strong VLA models. The outcome of this exploration is a simple yet effective model, VLANeXt. It outperforms the state-of-the-art methods on the LIBERO and LIBERO-plus benchmarks and demonstrates strong performance in real-world experiments. Beyond identifying the core recipe, we further ask how far these design principles extend to the emerging paradigms in VLAs. We thus expand VLANeXt along several emerging directions, including model scaling, latent-action pretraining, latent predictive representation learning, and world action modeling. These studies give rise to the VLANeXt family, spanning compact and scaled VLA variants, latent-action models, JEPA-style predictive models, and World Action Models. Our results show that the core recipe provides a strong foundation across different model scales and emerging paradigms.

cs.CV↗
arXiv2026

Beyond Imitation: Reflective On-Policy Self-Distillation for LLM Reasoning

On-policy self-distillation (OPSD) improves the reasoning capabilities of large language models (LLMs) by providing dense token-level supervision for on-policy rollouts. However, existing OPSD methods often yield limited gains on complex reasoning tasks and suffer from severe training instability. We identify two key causes: conditioning the self-teacher on a complete verified solution encourages imitation of complete reference trajectories rather than extraction of transferable reasoning insights, while indiscriminate full-response distillation imposes superfluous supervision on already-valid reasoning prefixes. Together, these issues suppress reasoning diversity and contribute to late-stage mode collapse. We propose Reflective On-policy Self-Distillation (ROSD), which distills transferable reasoning insights rather than complete reference trajectories. For each erroneous rollout, a self-reflector contrasts it with a correct rollout from the same group to derive a corrective idea and identify the sentence containing the first reasoning error. The corrective idea provides the self-teacher with targeted guidance, while the diagnosed error boundary allows ROSD to mask out the distillation loss over the valid prefix and apply token-level distillation only from the first erroneous sentence onward. Experiments across multiple reasoning benchmarks and model backbones show that ROSD consistently outperforms standard OPSD and reinforcement learning baselines, better preserves reasoning diversity, stabilizes training, and mitigates late-stage mode collapse. Code is available at https://github.com/ZiqiZhao1/ROSD.

cs.CL↗
arXiv2026

Puffin-World: Scaling a Unified Multimodal Model with Native 3D World States

We propose Puffin-World, a unified multimodal architecture that integrates physical understanding, spatial simulation, and 3D world generation and reconstruction without relying on external offline modules. To reliably construct and interact with 3D worlds, our framework jointly models three native world states: physics (gravity field and latitude), geometry (depth), and appearance (image), together with a unified Omni-Camera representation that supports diverse tasks and flexible motions. Beyond modeling these states, we introduce a strategy for propagating physical dynamics across future frames. By grounding absolute camera properties in the real world, Puffin-World enables physically consistent and visually stable world generation. We further couple appearance and geometry within a single generative process, jointly synthesizing each future view and reconstructing its underlying geometry. This unified paradigm enables interleaved closed-loop applications requiring synergy across multiple tasks, including mimic and self-calibrated world exploration. To scale Puffin-World to complex scenarios, we construct Puffin-16M, comprising 15 million vision-language-camera triplets and 1 million trajectories featuring various and challenging motions. To foster further research in this area, we released the code, models, and datasets.

cs.CV↗
arXiv2026

X2C: A Large-Scale Benchmark for Nuanced Humanoid Facial Expression Imitation

Fine-grained facial expression transfer from humans to humanoid agents presents a unique pattern recognition challenge due to the significant domain gap between biological facial dynamics and mechanical control spaces. While visual synthesis of talking heads has advanced rapidly, mapping high-dimensional visual cues to precise, physically constrained actuation signals remains an open problem, primarily due to the lack of large-scale paired data. To bridge this gap, we introduce X2C, a comprehensive benchmark dataset comprising 100,000 (image, control value) pairs. Unlike existing resources, X2C features nuanced, physically grounded expressions annotated with 30 continuous control parameters, establishing a high-fidelity standard for this task. Building on this resource, we propose X2CNet, a two-stage deep learning framework that explicitly decouples visual motion features from mechanical control regression to model the correspondence between human perceptual cues and humanoid actuation. Extensive experiments, including quantitative benchmarking and real-world physical validation, demonstrate that our approach achieves superior cross-domain consistency and enables robust, in-the-wild expression imitation. Code and Data: https://lipzh5.github.io/X2CNet/

cs.RO↗
arXiv2026

REAL: Reading Out Transformer Activations for Precise Localization in Language Model Steering

Inference-time steering aims to alter a large language model's (LLM's) responses without changing its parameters, but a central challenge is identifying the internal modules that most strongly govern the target behavior. Existing approaches often rely on simplistic cues or ad hoc heuristics, leading to suboptimal or unintended effects. We introduce REAL, a framework for identifying behavior-relevant modules (attention heads or layers) in Transformer models. For each module, REAL trains a vector-quantized autoencoder (VQ-AE) on its hidden activations and uses a shared, learnable codebook to partition the latent space into behavior-relevant and behavior-irrelevant subspaces. REAL quantifies a module's behavioral relevance by how well its VQ-AE encodings discriminate behavior-aligned from behavior-violating responses via a binary classification metric; this score guides both module selection and steering strength. We evaluate REAL across eight LLMs from the Llama and Qwen families and nine datasets spanning truthfulness enhancement, open-domain QA under knowledge conflicts, and general alignment tasks. REAL enables more effective inference-time interventions, achieving an average relative improvement of 20% (up to 81.5%) over the ITI method on truthfulness steering. In addition, the modules selected by REAL exhibit strong zero-shot generalization in cross-domain truthfulness-steering scenarios.

cs.CL↗
arXiv2026

SCI-PRM: A Tool Aware Process Reward Model for Scientific Reasoning Verification

While Process Reward Models (PRMs) have achieved remarkable success in mathematical reasoning, their application in complex scientific domains-such as biology, chemistry, and physics remains largely unexplored. Scientific problems demand not only logical rigor but also factual consistency and the precise usage of domain-specific tools, areas where current models often suffer from hallucinations and lack of verification. In this paper, we first construct SCIPRM70K, a large-scale dataset featuring Chain-of-Tool trajectories that explicitly interleave reasoning with the execution of scientific tools. Building upon this, we train an efficient reward model called Sci-PRM to provide fine-grained supervision on tool selection, execution accuracy, and result interpretation at each step in one inference. Experiments demonstrate that Sci-PRM significantly enhances foundation models in two key aspects: (1) it enables effective test-time scaling via Best-of-N selection; and (2) when integrated into Reinforcement Learning, it serves as a dense reward signal that mitigates the critical issue of advantage disappearance, allowing the model to break through existing performance ceilings.

cs.AI↗
arXiv2026

HATS: A Human-Agent Teleoperation System for Multi-Arm Data Collection

Many real-world manipulation scenarios, such as handling complex collaborative tasks and dealing with large workspaces, require coordination of more than two robotic arms. Consequently, an effective multi-arm teleoperation system is required to collect demonstrations for training coordinated multi-arm manipulation policies. However, existing teleoperation frameworks mainly focus on single-operator or multi-operator setups, facing a practical trade-off between the cognitive load placed on a single operator and the coordination cost incurred by multiple operators. To address this problem, we introduce HATS, a human-agent teleoperation system that enables a single human operator, assisted by an MLLM-based agent, to collect data for multi-arm manipulation tasks. Our system decouples the control space: two primary arms are directly teleoperated by the human, while two assistive arms are controlled by a training-free agent that handles sub-tasks. In addition, the human operator can use voice commands to prevent collisions and correct assistive arm behaviors during execution. Extensive evaluations demonstrate that HATS achieves data collection efficiency and success rates comparable to expert dual-human teams. Moreover, downstream policy evaluations demonstrate the efficacy and quality of the data collected through HATS.

cs.RO↗
arXiv2026

Earth Science Foundation Models: From Perception to Reasoning and Discovery

Large foundation models (FMs) are transforming Earth science by integrating heterogeneous multimodal data, such as multi-platform imagery, gridded reanalysis data, diverse geophysical and geochemical observations, and domain-specific text, to support tasks ranging from basic perception to advanced scientific discovery. This paper provides a unified review of Earth science foundation models (Earth FMs) through two complementary dimensions: depth, which traces the evolution of model capabilities from perception to multimodal reasoning and agentic scientific workflows, and breadth, which summarizes their expanding applications across the atmosphere, hydrosphere, lithosphere, biosphere, anthroposphere, and cryosphere, as well as coupled Earth system processes. Using this framework, we review representative multimodal Earth foundation models and compile more than 200 datasets and benchmarks spanning diverse Earth science tasks and modalities. We further discuss key challenges in multimodal data heterogeneity, scientific reliability and continual updating, scalability and sustainability, and the transition from foundation models to agentic and embodied Earth intelligence, and outline future directions toward more integrated, trustworthy, and actionable AI Earth scientists. Overall, this paper offers a structured roadmap for understanding the development of Earth foundation models from both capability depth and application breadth.

astro-ph.IM↗
arXiv2026

Task Editing for Generalizable 3D Visuomotor Policy Learning

3D visuomotor policies offer a promising direction for complex robotic manipulation, as depth maps and point clouds provide rich geometric information for spatial reasoning. However, their success often depends on large-scale real-world demonstrations, which are costly and time-consuming to collect. To this end, existing methods commonly use demonstration generation strategies to improve data efficiency by applying object-centric transformations to human-collected demonstrations, such as varying object poses or scales. While effective for local variation, these transformations largely preserve the original scene structure and skill sequence, limiting their ability to synthesize diverse scene-skill-object combinations for complex tasks. In this paper, we propose Task-Edit, a novel demonstration generation framework that generates diverse trajectories from a task-centric editing perspective. The key insight of Task-Edit is to decompose a task into scene, skill and object components, and flexibly recombine them. In this way, Task-Edit enables scalable demonstration generation and significantly improves generalization for long-horizon manipulation tasks. We evaluate Task-Edit through extensive real-world experiments and demonstrate three advantages: (1) Effectiveness: Task-Edit significantly improves 3D visuomotor policies across various real-world tasks and robot embodiments. (2) Generalizability: Task-Edit improves model generalization across different scenario setups. (3) Applicability: Task-Edit enables models to handle scenarios that are difficult to collect in the real world, including disturbance resistance, obstacle avoidance and unseen cluttered scenes.

cs.RO↗
arXiv2026

MSEarth: A Multimodal Benchmark for Earth Science Phenomenon Discovery with MLLMs

The rapid advancement of multimodal large language models (MLLMs) offers new opportunities for complex scientific challenges, yet their application in earth science-especially at the graduate level-remains underexplored due to a lack of benchmarks reflecting the depth and complexity of geoscientific reasoning. Existing datasets often rely on synthetic data or simple figure-caption pairs, failing to capture the nuanced reasoning required for real-world applications. To address this, we introduce MSEarth, a multimodal scientific dataset and benchmark curated from high-quality, open-access publications. Covering the five major spheres of Earth science-atmosphere, cryosphere, hydrosphere, lithosphere, and biosphere-MSEarth features over 289K figures with refined captions enriched by contextual discussions and reasoning from the original papers. The benchmark supports tasks such as scientific figure captioning, multiple choice questions, and open-ended reasoning, providing a scalable, high-fidelity resource for developing and evaluating MLLMs in scientific reasoning.

cs.AI↗
arXiv2026

FOREVER: Forgetting Curve-Inspired Memory Replay for Language Model Continual Learning

Continual learning (CL) for large language models (LLMs) aims to enable sequential knowledge acquisition without catastrophic forgetting. Memory replay methods are widely used for their practicality and effectiveness, but most rely on fixed, step-based heuristics that often misalign with the model's actual learning progress, since identical training steps can result in varying degrees of parameter change. Motivated by recent findings that LLM forgetting mirrors the Ebbinghaus human forgetting curve, we propose FOREVER (FORgEtting curVe-inspired mEmory Replay), a novel CL framework that aligns replay schedules with a model-centric notion of time. FOREVER defines model time using the magnitude of optimizer updates, allowing forgetting curve-inspired replay intervals to align with the model's internal evolution rather than raw training steps. Building on this approach, FOREVER incorporates a forgetting curve-based replay scheduler to determine when to replay and an intensity-aware regularization mechanism to adaptively control how to replay. Extensive experiments on three CL benchmarks and models ranging from 0.6B to 13B parameters demonstrate that FOREVER consistently mitigates catastrophic forgetting.

cs.LG↗
arXiv2026

SignDPO: Multi-level Direct Preference Optimisation for Skeleton-based Gloss-free Sign Language Translation

We present SignDPO, a novel multi-level Direct Preference Optimisation (DPO) framework designed to enhance the alignment of skeleton-based Sign Language Translation. While current skeleton-based models have made significant progress using Maximum Likelihood Estimation, they are primarily constrained by an imitation-based paradigm that lacks discriminative sensitivity to the fine-grained spatio-temporal nuances of sign language, often leading to semantic drift. To address this, SignDPO shifts the optimisation goal from simple sequence mimicry to structured preference alignment across spatial, temporal, and linguistic dimensions. Our framework involves three key designs. First, we introduce a hierarchical perturbation strategy to construct spatial and temporal non-preferred samples at both global and local granularities automatically. Second, we propose a self-guiding mechanism that leverages decoder cross-attention scores to identify and perturb semantically salient skeletal regions, forcing the model to distinguish genuine sign signals from structural distortions. Third, we establish an automated language-level preference generator by fine-tuning a dedicated perturbation model, capturing complex output-level failure modes without manual annotation. Extensive experiments on three widely adopted benchmarks, CSL-Daily, How2Sign, and OpenASL, demonstrate that SignDPO consistently outperforms state-of-the-art gloss-free methods and even rivals established gloss-based ones. Our results suggest that multi-level preference alignment is a powerful paradigm for bridging the gap between high-entropy skeletal trajectories and discrete linguistic semantics.

cs.CL↗
arXiv2026

CodeSpecBench: Benchmarking LLMs for Executable Behavioral Specification Generation

Large language models (LLMs) can generate code from natural language, but the extent to which they capture intended program behavior remains unclear. Executable behavioral specifications, defined via preconditions and postconditions, provide a concrete means to assess such understanding. However, existing work on specification generation is constrained in evaluation methodology, task settings, and specification expressiveness. We introduce CodeSpecBench, a benchmark for executable behavioral specification generation under an execution-based evaluation protocol. CodeSpecBench supports both function-level and repository-level tasks and encodes specifications as executable Python functions. Constructed from diverse real-world codebases, it enables a realistic assessment of both correctness (accepting valid behaviors) and completeness (rejecting invalid behaviors). Evaluating 15 state-of-the-art LLMs on CodeSpecBench, we observe a sharp performance degradation on repository-level tasks, where the best model attains only a 20.2% pass rate. We further find that specification generation is substantially more challenging than code generation, indicating that strong coding performance does not necessarily reflect deep understanding of intended program semantics. Our data and code are available at https://github.com/SparksofAGI/CodeSpecBench.

cs.SE↗
arXiv2026

Micro-Macro Retrieval: Reducing Long-Form Hallucination in Large Language Models

Large Language Models (LLMs) achieve impressive performance across many tasks but remain prone to hallucination, especially in long-form generation where redundant retrieved contexts and lengthy reasoning chains amplify factual errors. Recent studies highlight a critical phenomenon: the closer key information appears to the model outputs, the higher the factual accuracy. However, existing retrieval-augmented language models (RALMs) lack effective mechanisms to ensure this proximity - external evidence is injected into reasoning via multi-turn retrieval, but this cannot ensure key information stays close to the outputs. We propose Micro-Macro Retrieval (M2R), a novel retrieve-while-generate framework to fill this gap. At the macro level, M2R retrieves coarse-grained evidence from external sources; at the micro level, it extracts essential results from a key information repository built during reasoning and reuses them while generating answers. This design directly addresses the key-information-to-output proximity bottleneck, effectively reducing hallucination in long-form tasks. M2R is trained with a curriculum learning-based reinforcement learning strategy using customized rule-based rewards, enabling stable acquisition of retrieval and grounding skills. Extensive experiments across different benchmarks demonstrate the effectiveness of M2R, especially in lengthy-context settings.

cs.CL↗