arXiv subjects

Fei Ding

Publications and source records attributed to Fei Ding.

At least 37 records · Page 2Linked to original sources

arXiv2026

State commitment learning: training language models to distinguish computation from memory

Reasoning language models do not distinguish tokens used for computation from tokens that constitute persistent state: once generated, all hidden thoughts remain in context and influence future predictions. As a result, downstream reasoning may depend on failed attempts, dead ends, and private scratch work that should not be safely relied on later. We recast this phenomenon as a new training objective, state commitment learning: training models to explicitly distinguish information that should be committed as persistent state from temporary computation that can be discarded. We define a counterfactual criterion, persistent-state sufficiency, which makes it trainable and measurable whether an answer remains usable after hidden thoughts are erased. We then propose Counterfactual Erasure RL (CERL), which evaluates, under the same prefix, both a path that keeps hidden thoughts and a path that erases them, and gives reward only when the erasure path remains correct. We also introduce the Erasure Dependence Protocol and show across mathematics, long-chain logic, scientific QA, and multi-turn tool-use evaluation that CERL substantially reduces answer dependence on hidden thoughts without sacrificing accuracy, consistently outperforming correctness-only RL and long-answer SFT baselines.

cs.LG↗

arXiv2026

Scaffold-Mediated Post-Training: Co-Evolving Model Parameters and Procedural Scaffold Graphs

Post-training of large language models optimizes only parameters, while inference-time procedural scaffolds are typically designed independently of parameter training. This disconnect makes it difficult to automatically acquire and internalize complex strategies. We propose scaffold-mediated post-training: procedural scaffolds are organized into an evolvable graph structure that co-evolves with model parameters through discovery, distillation, and dynamic recompilation. We instantiate this paradigm as Skill Training. On FeatureBench, automatically discovered skills improve the passed rate by 8.1pp, and after progressive distillation the model still achieves a 27.7% passed rate without any external scaffold (distillation retention rate 85.2%, defined as post-distillation / with-skill passed rate), significantly outperforming standard SFT on the same data.

cs.CL↗

arXiv2026

Lance: Unified Multimodal Modeling by Multi-Task Synergy

We present Lance, a lightweight native unified model supporting multimodal understanding, generation, and editing for both images and videos. Rather than relying on model capacity scaling or text-image-dominant designs, Lance explores a practical paradigm for unified multimodal modeling via collaborative multi-task training. It is grounded in two core principles: unified context modeling and decoupled capability pathways. Specifically, Lance is trained from scratch and employs a dual-stream mixture-of-experts architecture on shared interleaved multimodal sequences, enabling joint context learning while decoupling the pathways for understanding and generation. We further introduce modality-aware rotary positional encoding to mitigate interference among heterogeneous visual tokens and boost cross-task alignment. During training, Lance adopts a staged multi-task training paradigm with capability-oriented objectives and adaptive data scheduling to strengthen both semantic comprehension and visual generation performance. Experimental results demonstrate that Lance substantially outperforms existing open-source unified models in image and video generation, while retaining strong multimodal understanding capabilities. The homepage is available at https://lance-project.github.io.

cs.CV↗

arXiv2026

SilverTorch: A Unified Model-based System to Democratize Large-Scale Recommendation on GPUs

Serving deep learning based recommendation models (DLRM) at scale is challenging. Existing approaches rely on dedicated ANN indexing and filtering services on CPUs, suffering from non-negligible costs and missing co-design opportunities. Such inefficiency makes them difficult to support complex model architectures, such as learned similarities and multi-task retrieval. In this paper, we present SilverTorch, a model-based serving system that brings all components into one unified model. It unifies model serving by replacing standalone indexing and filtering services with model layers. We propose a model-based GPU Bloom index for feature filtering and a fused Int8 ANN kernel for nearest neighbor search. Through co-design of the ANN search and feature filtering, we reduce GPU memory usage and eliminate computation. Benefiting from this design, we scale up retrieval by introducing an OverArch scoring layer and a multi-task retrieval with a Value Model to aggregate scores. These advancements improve the retrieval accuracy and enable future studies for serving more complex models. Our evaluation on industry-scale datasets show that SilverTorch achieves up to 23.7\times higher throughput compared to the state-of-the-art approaches. We also demonstrate that SilverTorch solution is 13.35\times more cost-efficient than CPU-based solution while improving accuracy via serving more complex models. SilverTorch is deployed at scale, serving hundreds of models online and supporting recommendation for diverse applications.

cs.IR↗

arXiv2026

DexSim2Real: Foundation Model-Guided Sim-to-Real Transfer for Generalizable Dexterous Manipulation

Sim-to-real transfer remains a critical bottleneck for deploying dexterous manipulation policies learned in simulation to real-world robots. Existing approaches rely on manually designed domain randomization or task-specific adaptation, limiting their generalizability across diverse manipulation scenarios. We present DexSim2Real, an integrated framework that leverages vision-language foundation models to bridge the sim-to-real gap for dexterous manipulation. Our system combines three components: (1) Foundation Model-Guided Domain Randomization (FM-DR), which uses a vision-language model as a visual realism critic to optimize simulation parameters via closed-loop CMA-ES, complementing text-based approaches like DrEureka with direct visual feedback; (2) a Tactile-Visual Cross-Attention Policy (TVCAP) that adapts cross-attention visuo-tactile fusion to zero-shot sim-to-real RL; and (3) a Progressive Skill Curriculum (PSC) that builds on LLM-based task decomposition with a difficulty scheduler tailored to contact-rich dexterous tasks. Extensive experiments on six challenging manipulation tasks with blinded evaluation demonstrate that DexSim2Real achieves a 78.2% average real-world success rate, outperforming DrEureka and DeXtreme while reducing the sim-to-real performance gap to only 8.3%.

cs.RO↗

arXiv2026

Physics-Aware LLM-Based Probabilistic Wind Power Scenario Generation under Extreme Icing Conditions

Accurately characterizing wind power uncertainty under icing and post-disaster conditions remains a critical challenge for resilient power system operation. To address this issue, this paper proposes a physics-aware large language model (LLM) framework for probabilistic wind power scenario generation under extreme icing conditions. The proposed framework integrates supervisory control and data acquisition (SCADA)-based physical modeling, multimodal tokenization, and a causal Transformer architecture trained in an autoregressive manner. A physics-aware decoding scheme effectively enforces rated power limits and ramping constraints on the generated trajectories while preserving stochastic diversity. Case studies using real wind turbine data show that the proposed method reproduces icing-induced power degradation and temporal variability observed during extreme weather. The resulting scenarios are physically consistent and high-fidelity, thereby significantly enhancing resilience assessment and recovery planning in renewable-integrated power systems.

eess.SY↗

arXiv2026

Spectral shadows of a single GaAs quantum dot

Semiconductor quantum dots are a promising platform for generating single and entangled photons.Still, their use is limited even in the most advanced structures by changes in the charge state of the quantum dot and its environment. Here, we present detailed time-resolved resonance fluorescence measurements on a single charge-tunable GaAs quantum dot, shedding new light on the spectral shadows invoked by the complex impurity environment. Detuning-dependent measurements reveal the existence of multiple Stark-shifted resonances, which are associated with rare spectral jumps smaller than the homogeneous linewidth and, therefore, typically concealed in the measurement noise. We observe similar environmentally induced Stark shifts for both the neutral exciton and negatively charged trion transitions, while the positively and doubly negatively charged trions exhibit significant differences. Our investigation quantifies the underlying impurity charge dynamics over a range from well below milliseconds to seconds, revealing that the hole occupation of the positively charged trion transition is constrained by rapid hole loss and slow hole recapture dynamics. Utilizing a second non-resonant laser, we increase the hole occupancy by over an order of magnitude and identify both a prolonged hole residence time and an enhanced hole tunneling rate into the quantum dot. These findings are supported by complementary spin noise spectroscopy measurements, which offer a significantly higher bandwidth compared to the time-resolved resonance fluorescence measurements.

cond-mat.mes-hall↗

arXiv2026

HELM: Harness-Enhanced Long-horizon Memory for Vision-Language-Action Manipulation

Vision-Language-Action (VLA) models fail systematically on long-horizon manipulation tasks despite strong short-horizon performance. We show that this failure is not resolved by extending context length alone in the current reactive execution setting; instead, it stems from three recurring execution-loop deficiencies: the memory gap, the verification gap, and the recovery gap. We present HELM, a model-agnostic framework that addresses these deficiencies with three components: an Episodic Memory Module (EMM) that retrieves key task history via CLIP-indexed keyframes, a learned State Verifier (SV) that predicts action failure before execution from observation, action, subgoal, and memory-conditioned context, and a Harness Controller (HC) that performs rollback and replanning. The SV is the core learning contribution: it consistently outperforms rule-based feasibility checks and ensemble uncertainty baselines, and its effectiveness depends critically on access to episodic memory. On LIBERO-LONG, HELM improves task success rate by 23.1 percentage points over OpenVLA (58.4% to 81.5%), while extending the context window to H=32 yields only a 5.4-point gain and same-budget LoRA adaptation remains 12.2 points below HELM. HELM also improves long-horizon performance on CALVIN and substantially boosts recovery success under controlled perturbations. Ablations and mechanism analyses isolate the contribution of each component, and we release LIBERO-Recovery as a perturbation-injection protocol for evaluating failure recovery in long-horizon manipulation.

cs.LG↗

arXiv2026

Seedance 2.0: Advancing Video Generation for World Complexity

Seedance 2.0 is a new native multi-modal audio-video generation model, officially released in China in early February 2026. Compared with its predecessors, Seedance 1.0 and 1.5 Pro, Seedance 2.0 adopts a unified, highly efficient, and large-scale architecture for multi-modal audio-video joint generation. This allows it to support four input modalities: text, image, audio, and video, by integrating one of the most comprehensive suites of multi-modal content reference and editing capabilities available in the industry to date. It delivers substantial, well-rounded improvements across all key sub-dimensions of video and audio generation. In both expert evaluations and public user tests, the model has demonstrated performance on par with the leading levels in the field. Seedance 2.0 supports direct generation of audio-video content with durations ranging from 4 to 15 seconds, with native output resolutions of 480p and 720p. For multi-modal inputs as reference, its current open platform supports up to 3 video clips, 9 images, and 3 audio clips. In addition, we provide Seedance 2.0 Fast version, an accelerated variant of Seedance 2.0 designed to boost generation speed for low-latency scenarios. Seedance 2.0 has delivered significant improvements to its foundational generation capabilities and multi-modal generation performance, bringing an enhanced creative experience for end users.

cs.CV↗

arXiv2026

Order Is Not Layout: Order-to-Space Bias in Image Generation

We study a systematic bias in modern image generation models: the mention order of entities in text spuriously determines spatial layout and entity--role binding. We term this phenomenon Order-to-Space Bias (OTS) and show that it arises in both text-to-image and image-to-image generation, often overriding grounded cues and causing incorrect layouts or swapped assignments. To quantify OTS, we introduce OTS-Bench, which isolates order effects with paired prompts differing only in entity order and evaluates models along two dimensions: homogenization and correctness. Experiments show that Order-to-Space Bias (OTS) is widespread in modern image generation models, and provide evidence that it is primarily data-driven and manifests during the early stages of layout formation. Motivated by this insight, we show that both targeted fine-tuning and early-stage intervention strategies can substantially reduce OTS, while preserving generation quality.

cs.CL↗

arXiv2026

Time-bin encoded quantum key distribution over 120 km with a telecom quantum dot source

Quantum key distribution (QKD) with deterministic single photon sources has been demonstrated over intercity fiber and free-space channels. The previous implementations relied mainly on polarization encoding schemes, which are susceptible to birefringence, polarization-mode dispersion and polarization-dependent loss in practical fiber networks. In contrast, time-bin encoding offers inherent robustness and has been widely adopted in mature QKD systems using weak coherent laser pulses. However, its feasibility in conjunction with a deterministic single-photon source has not yet been experimentally demonstrated. In this work, we construct a time-bin encoded QKD system employing a high-brightness quantum dot (QD) single-photon source operating at telecom wavelength. Our proof-of-concept experiment successfully demonstrates the possibility of secure key distribution over fiber link of 120 km, while maintaining extraordinary long-term stability over 6 hours of continuous operation. This work provides the first experimental validation of integrating a quantum dot single-photon source with time-bin encoding in a telecom-band QKD system. In addition, it demonstrates the highest secure key rate among the time-bin QKDs based on single-photon sources. This development signifies a substantial advancement in the establishment of a robust and scalable QKD network based on solid-state single-photon technology

quant-ph↗

arXiv2026

Quantum Dots as Solid-State Sources of Entangled Photon Pairs

Semiconductor quantum dots (QDs) have emerged as a premier solid-state platform for the deterministic generation of nonclassical light, offering a compelling pathway toward scalable quantum photonic systems. While single-photon emission from QDs has reached a high level of maturity, the realization of high-fidelity entangled photon-pair sources remains an active and rapidly evolving frontier. In this review, we survey the recent progress in QD-based entangled photon sources, highlighting the conceptual evolution from the established biexciton-exciton cascade to the emerging paradigm of spontaneous two-photon emission. We further examine how advances in nanophotonic architectures and coherent control strategies are redefining fundamental performance limits, enabling concurrent improvements in source brightness, coherence, and entanglement fidelity. Finally, we discuss the key physical and technological challenges that must be addressed to bridge the gap between laboratory demonstrations and large-scale deployment. We conclude by outlining the future opportunities for integrating QD-based entangled photon sources into practical quantum communication, computation, and sensing platforms.

physics.optics↗

arXiv2026

PoseStreamer: A Multi-modal Framework for 3D Tracking of Unseen Moving Objects

Six degree of freedom (6DoF) pose estimation for novel objects is a critical task in computer vision, yet it faces significant challenges in high-speed and low-light scenarios where standard RGB cameras suffer from motion blur. While event cameras offer a promising solution due to their high temporal resolution, current 6DoF pose estimation methods typically yield suboptimal performance in high-speed object moving scenarios. To address this gap, we propose PoseStreamer, a robust multi-modal 6DoF pose estimation framework designed specifically on high-speed moving scenarios. Our approach integrates three core components: an Adaptive Pose Memory Queue that utilizes historical orientation cues for temporal consistency, an Object-centric 2D Tracker that provides strong 2D priors to boost 3D center recall, and a Ray Pose Filter for geometric refinement along camera rays. Furthermore, we introduce MoCapCube6D, a novel multi-modal dataset constructed to benchmark performance under rapid motion. Extensive experiments demonstrate that PoseStreamer not only achieves superior accuracy in high-speed moving scenarios, but also exhibits strong generalizability as a template-free framework for unseen moving objects.

cs.CV↗

arXiv2025

Multi-GRPO: Multi-Group Advantage Estimation for Text-to-Image Generation with Tree-Based Trajectories and Multiple Rewards

Recently, Group Relative Policy Optimization (GRPO) has shown promising potential for aligning text-to-image (T2I) models, yet existing GRPO-based methods suffer from two critical limitations. (1) \textit{Shared credit assignment}: trajectory-level advantages derived from group-normalized sparse terminal rewards are uniformly applied across timesteps, failing to accurately estimate the potential of early denoising steps with vast exploration spaces. (2) \textit{Reward-mixing}: predefined weights for combining multi-objective rewards (e.g., text accuracy, visual quality, text color)--which have mismatched scales and variances--lead to unstable gradients and conflicting updates. To address these issues, we propose \textbf{Multi-GRPO}, a multi-group advantage estimation framework with two orthogonal grouping mechanisms. For better credit assignment, we introduce tree-based trajectories inspired by Monte Carlo Tree Search: branching trajectories at selected early denoising steps naturally forms \emph{temporal groups}, enabling accurate advantage estimation for early steps via descendant leaves while amortizing computation through shared prefixes. For multi-objective optimization, we introduce \emph{reward-based grouping} to compute advantages for each reward function \textit{independently} before aggregation, disentangling conflicting signals. To facilitate evaluation of multiple objective alignment, we curate \textit{OCR-Color-10}, a visual text rendering dataset with explicit color constraints. Across the single-reward \textit{PickScore-25k} and multi-objective \textit{OCR-Color-10} benchmarks, Multi-GRPO achieves superior stability and alignment performance, effectively balancing conflicting objectives. Code will be publicly available at \href{https://github.com/fikry102/Multi-GRPO}{https://github.com/fikry102/Multi-GRPO}.

cs.CV↗

arXiv2025

DreamO: A Unified Framework for Image Customization

Recently, extensive research on image customization (e.g., identity, subject, style, background, etc.) demonstrates strong customization capabilities in large-scale generative models. However, most approaches are designed for specific tasks, restricting their generalizability to combine different types of condition. Developing a unified framework for image customization remains an open challenge. In this paper, we present DreamO, an image customization framework designed to support a wide range of tasks while facilitating seamless integration of multiple conditions. Specifically, DreamO utilizes a diffusion transformer (DiT) framework to uniformly process input of different types. During training, we construct a large-scale training dataset that includes various customization tasks, and we introduce a feature routing constraint to facilitate the precise querying of relevant information from reference images. Additionally, we design a placeholder strategy that associates specific placeholders with conditions at particular positions, enabling control over the placement of conditions in the generated results. Moreover, we employ a progressive training strategy consisting of three stages: an initial stage focused on simple tasks with limited data to establish baseline consistency, a full-scale training stage to comprehensively enhance the customization capabilities, and a final quality alignment stage to correct quality biases introduced by low-quality data. Extensive experiments demonstrate that the proposed DreamO can effectively perform various image customization tasks with high quality and flexibly integrate different types of control conditions.

cs.CV↗

arXiv2025

RealCustom++: Representing Images as Real Textual Word for Real-Time Customization

Given a text and an image of a specific subject, text-to-image customization aims to generate new images that align with both the text and the subject's appearance. Existing works follow the pseudo-word paradigm, which represents the subject as a non-existent pseudo word and combines it with other text to generate images. However, the pseudo word causes semantic conflict from its different learning objective and entanglement from overlapping influence scopes with other texts, resulting in a dual-optimum paradox where subject similarity and text controllability cannot be optimal simultaneously. To address this, we propose RealCustom++, a novel real-word paradigm that represents the subject with a non-conflicting real word to firstly generate a coherent guidance image and corresponding subject mask, thereby disentangling the influence scopes of the text and subject for simultaneous optimization. Specifically, RealCustom++ introduces a train-inference decoupled framework: (1) during training, it learns a general alignment between visual conditions and all real words in the text; and (2) during inference, a dual-branch architecture is employed, where the Guidance Branch produces the subject guidance mask and the Generation Branch utilizes this mask to customize the generation of the specific real word exclusively within subject-relevant regions. In contrast to previous methods that excel in either controllability or similarity, RealCustom++ achieves superior performance in both, with improvements of 7.48% in controllability, 3.04% in similarity, and 76.43% in generation quality. For multi-subject customization, RealCustom++ further achieves improvements of 4.6% in controllability and 6.34% in multi-subject similarity. Our work has been applied in JiMeng of ByteDance, and codes are released at https://github.com/bytedance/RealCustom.

cs.CV↗

arXiv2025

Dynamic and Continuous Control of Second-Harmonic Chirality through Lithium Niobate Nonlocal Metasurface

Nonlinear chiral light sources are crucial for emerging applications in chiroptics, including ultrafast spin dynamics and quantum state manipulation. However, achieving precise and dynamic control over nonlinear optical chirality with natural materials and metasurfaces, particularly those based on non-centrosymmetric materials such as lithium niobate (LN), is hindered by the complex tensorial nature of the second-order nonlinear susceptibility. Here, we demonstrate a nonlinear nonlocal metasurface, comprising plasmonic nanoantennas atop an x-cut LN thin film, that enables dynamic and continuous control of second-harmonic (SH) chirality. By leveraging two spectrally detuned resonances arising from the excitation of orthogonally propagating guided modes, enabled by lattice anisotropy and LN birefringence, we achieve full-range tuning of SH chirality, from right- to left-handed circular polarization, simply by rotating the polarization of a linearly polarized pump. The SH chirality, quantified by the Stokes parameter S3, is thereby continuously tuned from 0.991 to -0.993 in simulations and from 0.920 to -0.815 in experiments, while maintaining consistently enhanced SH intensity across the entire tuning range. Our approach opens new avenues for developing compact and tunable chiral sources, with potential applications in integrated nonlinear photonics and adaptable quantum technologies.

physics.optics↗

arXiv2025

The enhanced X-ray Timing and Polarimetry mission -- eXTP for launch in 2030

In this paper we present the current status of the enhanced X-ray Timing and Polarimetry mission, which has been fully approved for launch in 2030. eXTP is a space science mission designed to study fundamental physics under extreme conditions of matter density, gravity, and magnetism. The mission aims at determining the equation of state of matter at supra-nuclear density, measuring the effects of quantum electro-dynamics, and understanding the dynamics of matter in strong-field gravity. In addition to investigating fundamental physics, the eXTP mission is poised to become a leading observatory for time-domain and multi-messenger astronomy in the 2030's, as well as providing observations of unprecedented quality on a variety of galactic and extragalactic objects. After briefly introducing the history and a summary of the scientific objectives of the eXTP mission, this paper presents a comprehensive overview of: 1) the cutting-edge technology, technical specifications, and anticipated performance of the mission's scientific instruments; 2) the full mission profile, encompassing spacecraft design, operational capabilities, and ground segment infrastructure.

astro-ph.HE↗