arXiv subjects

Haoyang Zhang

Publications and source records attributed to Haoyang Zhang.

At least 19 recordsLinked to original sources

arXiv2026

GeoGSplat: Geometry-Grounded Non-Pixel-Aligned Gaussian Splatting

Feed-forward Gaussian Splatting enables efficient novel view synthesis, but pixel-aligned prediction makes the primitive count grow with image resolution and view count, leading to substantial redundancy and increased computational cost. Non-pixel-aligned methods address this limitation by predicting a fixed number of Gaussians. Without explicit ray and depth constraints, these methods yield misplaced Gaussian centers, causing rendering artifacts under large viewpoint changes. To address this issue, we introduce GeoGSplat, a geometry-grounded feed-forward model that explicitly incorporates geometric priors into the non-pixel-aligned framework. A dual-branch encoder constructs Anchors and Triplane to capture coarse 3D structure and fine-grained scene features, respectively. Subsequently, a hierarchical decoder progressively decodes dense 3D points and Gaussian primitives, enabling 3D reconstruction and novel view synthesis. GeoGSplat surpasses both state-of-the-art pixel-aligned and non-pixel-aligned Gaussian Splatting methods on DTU and NRGBD in terms of novel view synthesis and reconstruction. The advantage of our geometry-grounded paradigm further increases as target viewpoints move farther from the input cameras, demonstrating stronger generalization to large viewpoint changes. Project Page: https://comedy0913.github.io/projects/geogsplat

cs.CV↗

arXiv2026

AlignDrive: Aligned Lateral-Longitudinal Planning for End-to-End Autonomous Driving

Practical autonomous driving requires models that generalize by reasoning through spatial-temporal possibilities to exclude unsafe outcomes. While state-of-the-art (SOTA) methods use parallel planning architectures, they fail to explicitly couple speed decisions with agent behavior along the driving path, leading to suboptimal coordination. To address this, we propose a cascaded framework that transforms longitudinal planning from an independent prediction task into a path-conditioned reasoning process. On the model side, we introduce an anchor-based regression design that conditions longitudinal prediction on the lateral drive path, and reformulate longitudinal planning as 1D displacement prediction along the path. This reduces geometric uncertainty and sharpens the model's focus on interaction-driven dynamics. On the data side, we introduce a planning-oriented data augmentation strategy that simulates rare safety-critical events by programmatically inserting agents and relabeling longitudinal targets to enforce collision avoidance. Evaluated on the challenging Bench2Drive benchmark, our method achieves SOTA performance with a driving score of 89.07 and a success rate of 73.18%, demonstrating significantly improved coordination and safety. Further evaluation on Fail2Drive confirms strong generalization to rare edge cases where parallel formulations typically fail. Project page:https://yanhaowu.github.io/AlignDrive/.

cs.RO↗

arXiv2026

Fish Audio Drama 3 Technical Report

We introduce Fish Audio Drama 3, an open-source speech generation model that turns text into directed performances, built on a 35B-parameter mixture-of-experts backbone that activates 3B parameters per token. It clones voices from multiple references, follows natural-language direction, revises a take when the next turn gives a note, voices multi-character scenes, edits existing speech by infilling, and synthesizes long-form and streaming audio. Multi-turn revision, scene dubbing, editing by infilling, and streaming text input are new in Drama 3, all in one model. Drama 3 pairs the hybrid linear-attention backbone with a new codebook multi-token-prediction audio decoder that predicts the acoustic codebooks of each frame in dedicated layers that attend over the entire sequence. The model is trained with a three-stage pre-training recipe, instruction fine-tuning, and large-batch reinforcement learning whose only reward is a single learned model of human preference. Its inference engine decodes a single stream at 271 frames per second on one H200, against 177 for Fish Audio S2, although Drama 3 has more than seven times as many parameters. The model weights are available on GitHub (https://github.com/fishaudio/fish-speech) and Hugging Face (https://huggingface.co/fishaudio/drama-3), and inference is available through SGLang-Omni (https://github.com/sgl-project/sglang-omni). We highly encourage readers to visit https://fish.audio to direct their own performances.

cs.SD↗

arXiv2026

Artificial Foveated Perception for Mitigating Shortcut Learning in Robotic Foundation Models

Robotic foundation models still need task-specific fine-tuning before deployment, and the fine-tuned policies often break under modest changes in scene layout, lighting, or nearby distractors. We trace this brittleness to \textit{shortcut learning}: fine-tuning supervises actions but not the visual evidence the policy uses, so the policy can settle on scene-level correlations that predict the demonstrations without causing success. We propose Artificial Foveated Perception (AFP), a lightweight, policy-agnostic module that takes the same vision and language inputs as existing Vision-Language-Action and World Action Model pipelines and predicts task-conditioned masks over the relevant objects, the robot, and other action-critical regions. During fine-tuning the masks serve as an auxiliary grounding signal that aligns the policy's visual attention with task-relevant regions; the policy architecture is unchanged, and at inference the policy runs on the original observation stream with no AFP call in the control loop. In simulation with four robotic foundation models and on a real robot with $π_{0.5}$, AFP improves generalization under environmental perturbations, reduces overfitting, and shortens fine-tuning. Ablations over mask quality and grounding-loss design show that these gains come from directing policy learning toward task-relevant visual evidence. Code, data, and videos are available at https://apollo-lab-yale.github.io/26-CoRL-AFP-website/.

cs.RO↗

arXiv2026

StepAudio 3 Realtime Technical Report

Realtime spoken interaction demands deep reasoning, prompt responses, and fluid turn-taking. We present StepAudio 3 Realtime, an audio-language foundation model organized around a continuous listen-converse-think-act loop. Deep Perception captures rich acoustic cues to interpret user intent, while Seamless Duplex models synchronized audio streams to handle pauses, backchannels, and interruptions naturally. Crucially, we resolve the tension between deep deliberation and latency via Think-While-Speaking, executing private reasoning in parallel with spoken delivery. In reasoning mode, StepAudio 3 reaches a 73.0 macro average on StepAudioChat. With Think-While-Speaking, it achieves dialogue and reasoning performance comparable to dedicated reasoning models while speaking in real time. Furthermore, an integrated Voice Agent handles asynchronous tool execution without disrupting the dialogue flow. StepAudio 3 Realtime achieves top-tier performance across key dimensions: an exceptional 90.6 on the MMSU benchmark, 98.9 Overall on the Artificial Analysis Full-Duplex Bench, and a 56.0% macro task-success rate on $τ$-Voice.

cs.SD↗

arXiv2026

Is This Evidence Decision-Critical? Learning to Verify Rule-Governed Decisions

Rule-based reasoning, as in eligibility checks and contract reviews, requires language models to assess evidence against individual conditions and combine their judgments under explicit rules. Errors in evidence assessment can leave a decision unchanged, but misinterpreting or overlooking decision-critical evidence can reverse it. Identifying such evidence allows more capable models to focus on checking the corresponding condition judgments, supporting accurate and safe decisions. Recognizing the evidence's criticality requires understanding how evidence affects a condition judgment and how that judgment affects the decision. To achieve the goal, we propose a INTERvention-based imPACT learning framework (InterPact), which enables counterfactual verification of evidence criticality in rule-governed decisions. Specifically, its evidence intervention constructor generates training pairs for a propagation verifier by editing case facts with a frozen language model while holding rules and non-target conditions fixed. Human-reviewed labels record the resulting condition and decision changes, while complete state-to-decision mappings supervise consequences beyond the observed edit. During training, the verifier weights learned conditional decision predictions by evidence-based condition probabilities through a fixed composition operation, propagating decision-change supervision into the base model. At inference, the trained base model directly judges criticality from the original case and target evidence, without human or stronger-model supervision. On single-case evidence criticality verification over adapted rule-governed decision cases, InterPact achieves 68.28% accuracy, outperforming all six baselines. These results support learned decision sensitivity as a basis for prioritizing evidence checks.

cs.CL↗

arXiv2026

Beyond Textual Chain-of-Thought: JEPA-Conditioned Latent Reasoning for Large Audio Language Models

Explicit textual Chain-of-Thought (CoT) has improved the reasoning ability of large audio language models (LALMs). However, textual CoTs are often constructed from text captions of audio and provide limited access to the acoustic evidence, which can introduce problems like hallucination. To address this modality-gap issue, we introduce JELAR, a Joint-Embedding Predictive Architecture (JEPA)-based latent reasoning framework that conditions latent reasoning supervision on acoustic representations learned from raw waveforms. During training, a frozen WavJEPA model provides representations learned directly from raw waveforms. A non-causal expert first constructs answer-aware queries, which cross-attend to WavJEPA embeddings to produce latent reasoning targets. The LALM is trained to predict these targets before generating its response. Experimental results show that JELAR improves the Audio-Reasoner baseline by 2.70 and 9.10 absolute percentage points on MMAU-mini and MMAR, respectively, demonstrating the effectiveness of JEPA-conditioned latent reasoning as an alternative to explicit textual CoT supervision.

eess.AS↗

arXiv2026

RGDT-Bench: Benchmarking LLM Reasoning for Rule-Governed Decisions and Their Justifications

We study reasoning in Rule-Governed Decision Tasks (RGDTs), where models apply external rules to case facts and justify decisions, as required in policy, contract, and compliance settings. Beyond the deductive capability emphasized by standard mathematical and logical reasoning tasks, RGDTs require interpreting rules and their applicability, assessing conditions from evidence, combining judgments under rules and exceptions, and providing checkable justifications. These demands motivate a benchmark assessing both decisions and their stated grounds. We introduce RGDT-Bench, providing 202.1K condition-level supervision slots across four task tracks and eight supported task-probe combinations that vary access to supporting information. Label-blind extraction and deterministic checks produce labels for warrant completeness: source-referenced coverage and consistency of stated decision grounds. The benchmark attributes failures to four process layers: rule use, condition, evidence, and aggregation, and checks the final outcome. Among evaluable correct responses, warrant incompleteness averages 40.2% across six evaluated LLMs and supported task-probe combinations. Such warrant incompleteness poses potential safety risks and remains difficult to detect: the best of seventeen existing evaluators reaches only 57.69% (random: 50%) task-averaged area under the receiver operating characteristic curve (AUROC). To address this difficulty, we train a simple reward model with warrant supervision. It achieves 69.24% task-averaged AUROC among correct answers, exceeding the matched outcome-supervised baseline by 10.37 pp (percentage points) and the best existing evaluator by 11.55 pp. Beyond completeness assessment, the model outperforms both outcome-supervised baselines across nearly all response-selection comparisons, supporting RGDT-Bench's warrant supervision for RGDT reasoning.

cs.CL↗

arXiv2026

Detection of a puzzling dual-superorbital hard X-ray modulation in the X-ray binary GX 301-2

The superorbital modulations (SMs) observed in wind-fed X-ray binaries remain a puzzling phenomenon in astrophysics. To investigate this behavior observationally, we analyzed the long-term hard X-ray light curve from the Swift/BAT 157-Month Hard X-ray Survey in X-ray binary GX 301-2. Using three timing analysis methods--the Lomb-Scargle periodogram, the weighted wavelet Ztransform, and Gaussian processes--we identify a rare dual-SM behavior in this source: the 115-day modulation exceeds the 5$σ$ global significance level, whereas the 65-day signal only marginally reaches the 4$σ$ level. Because the 115-day period is more consistent with the previously reported linear relation between orbital and superorbital periods, we interpret 115 days as the actual superorbital period, while the weaker and less stable 65-day period is its beat modulation with the orbital period.By assessing the applicability of different physical scenarios to our results, we suggest that this dual-SM behavior is most plausibly associated with corotating interaction regions (CIRs) in the stellar wind. This framework can also account for the observed linear orbital-superorbital relation, despite the unclear physical mechanism that sets the apparent ratio between the CIR and orbital periods across sources. Further long-term monitoring of this system, together with continued theoretical development of the CIR scenario, will be essential for clarifying the origin of wind-fed SMs.

astro-ph.HE↗

arXiv2026

Square-Root Higher-Order Exceptional Points with Symmetry-Induced Multiple Spectral Responses

We generalize square-root procedure to non-Hermitian systems with finite lattices, providing a spectral operation scheme applicable to arbitrary tight-binding models. Via this generalized square-root approach, we construct novel chiral-symmetric higher-order exceptional points (EPs) with multiple spectral responses. By taking square-root of a parent Hamiltonian hosting an $n$th-order EP (EP$_n$), an EP$_{2n+1}$ chiral-symmetric square-root system is obtained, whose lattice sites are inherited from both the parent system and an auxiliary residual system. The chiral-symmetry-induced structure of the generalized eigenspace enables onsite and coupling perturbations to selectively generate spectral responses of different orders. The proposed scheme is universal, applicable to any existing tight-binding EP system and iterable to generate EPs of arbitrarily high order. With enriched ultrasensitive spectral responses, the chiral-symmetric higher-order EP system provides a promising platform for signal amplification, detection and non-Hermitian control.

physics.optics↗

arXiv2026

From Procedural Skills to Strategy Genes: Towards Experience-Driven Test-Time Evolution

This beta technical report asks how reusable experience should be represented so that it can function as effective test-time control and as a substrate for iterative evolution. We study this question in 4.590 controlled trials across 45 scientific code-solving scenarios. We find that documentation-oriented Skill packages provide unstable control: their useful signal is sparse, and expanding a compact experience object into a fuller documentation package often fails to help and can degrade the overall average. We further show that representation itself is a first-order factor. A compact Gene representation yields the strongest overall average, remains competitive under substantial structural perturbations, and outperforms matched-budget Skill fragments, while reattaching documentation-oriented material usually weakens rather than improves it. Beyond one-shot control, we show that Gene is also a better carrier for iterative experience accumulation: attached failure history is more effective in Gene than in Skill or freeform text, editable structure matters beyond content alone, and failure information is most useful when distilled into compact warnings rather than naively appended. On CritPt, gene-evolved systems improve over their paired base models from 9.1% to 18.57% and from 17.7% to 27.14%. These results suggest that the core problem in experience reuse is not how to supply more experience, but how to encode experience as a compact, control-oriented, evolution-ready object.

cs.SE↗

arXiv2026

AutoResearch: Insight In, Hallucination Out

Autonomous research systems are increasingly capable of executing long research workflows, yet automation alone does not ensure that the resulting process remains scientifically grounded. We introduce AutoResearch, a two-stage system that connects Idea Generation with Idea Execution to address both how research ideas are formed and how they are reliably established through experimentation. In Idea Generation, AutoResearch continuously integrates emerging research signals with accumulated domain knowledge, identifies transferable mechanistic insights, and uses multi-model generation and cross-review to produce grounded, testable research plans. In Idea Execution, coordinated agents decompose these plans into experiments, iteratively implement and diagnose them, and employ independent evidence-based review before accepting research conclusions. Across representative settings in cross-modal retrieval, systems optimization, and benchmark-driven machine learning, AutoResearch turns generated ideas into measurable progress, detects and corrects unreliable experimental results, and makes evidence-conditioned decisions to continue, revise, or terminate research directions. For example, on RSICD benchmark, an AutoResearch-generated idea improves mean Recall from 32.84 to 34.69, while recording only 5 audit-confirmed issue events compared with 11-27 for other autonomous research systems. These results demonstrate a research process in which meaningful insight is grounded before experimentation and conclusions are grounded before acceptance: Insight In, Hallucination Out.

cs.AI↗

arXiv2026

LongWoF-Bench: Evaluating EvoMap Genes for Verifiable Long-Workflow Tasks

Large language models are increasingly expected to execute complex workflows whose success depends on maintaining interdependent constraints and producing artifacts that satisfy strict end-to-end verification. Yet successful execution experience is typically lost after a single run, forcing subsequent models to rediscover strategies and failure modes from scratch. We study whether such experience can instead be externalized and reused through EvoMap, where verifier-confirmed execution trajectories are consolidated into structured Gene. To evaluate this setting, we introduce the Long-Workflow Benchmark (LongWoF-Bench), comprising 778 machine-verifiable tasks across code generation, agent-environment synthesis, mathematical reasoning, and rule following. On the 252 tasks with verifier-confirmed Opus trajectories, evolved EvoMap Gene outperform Skill across all seven evaluated models by 8.7-15.5 percentage points, with the gains extending to consumer models from different model families. In contrast, reference-distilled Gene do not exhibit the same advantage, indicating that compact representation alone is insufficient and that Gene utility is closely associated with verified experience provenance. For Claude Opus, Gene reuse also completes 39 more tasks than Skill while reducing solve-time token consumption by 9.9%. Together, these results show that verified execution experience can be retained and shared as a reusable external resource, enabling models to improve long-workflow completion without repeatedly paying the full cost of experience discovery.

cs.CL↗

arXiv2026

Decoupling Vision, Language, and Action for Efficient Multi-Task Robot Policies

Vision-Language-Action (VLA) policies commonly run Vision-Language Model (VLM) backbones with billions of parameters at every policy inference, which costs latency and energy. We revisit a decoupled alternative for multi-task manipulation: separate vision and language encoders whose representations condition a compact action head. We run a standardized comparison that varies the vision encoder, the language encoder, and the action head while holding the demonstrations, the training-step budget, the tasks, the evaluation protocol, and the measurement platform fixed, against seven VLA baselines. The resulting Decoupled Embodiment Model (DEM) combines a fine-tuned DINOv3 vision encoder, a frozen NeoBERT language encoder, and a MeanFlow head that generates an action chunk in one forward pass. On 18 RoboCasa tasks evaluated with held-out instruction paraphrases and randomized scenes, DEM reaches 55.6\% mean success against 56.9\% for GR00T N1.7 and 54.6\% for $π_{0.5}$, and on three real-robot tasks it reaches 66.0\% against 68.0\% for GR00T N1.7. On the same workstation, DEM needs 6.1\,ms per policy forward pass, a maximum throughput of 162.7 policy calls per second, and draws an estimated 2.07\,J of GPU energy per call, eight to seventeen times the throughput and six to fifteen times less energy than these VLM-backbone policies. Within this trained-task regime, DEM sits on the observed success--latency--energy frontier and provides a strong, efficient baseline for language-conditioned robot skills.

cs.RO↗

arXiv2026

An encoded asymmetric ligand for metal-selective topological assembly of two-dimensional metal-organic frameworks

Two-dimensional metal-organic frameworks (2D MOFs), with diverse topological architectures, provide a powerful platform for exploring unconventional electronic and lattice-dynamical responses. Yet their structural diversity remains fundamentally constrained by the fixed geometry of high-symmetry ligands. Here, we introduce an encoded asymmetric ligand, benzo[b]triphenylene-2,3,6,7,11,12-hexaol (BTH), for metal-selective topological assembly. By integrating multi-site coordination fields with sterically differentiated environments, BTH exhibits distinct topological programmability: different divalent metal ions direct divergent framework architectures. Specifically, coordination of BTH with divalent Cu(II) and Zn(II) ions assembles Cu-BTH-MOF with a dual-mode hexagonal pore topology and Zn-BTH-MOF with uniform hexagonal channels, respectively, as supported by PXRD Pawley refinement, structural simulations, and pore-size distribution analysis. Furthermore, this topological divergence is accompanied by a significant divergence in charge-transport properties, with Cu-BTH-MOF reaching an electrical conductivity of 1.186 x 10-3 S cm-1, more than six orders of magnitude higher than that of Zn-BTH-MOF (3.38 x 10-10 S cm-1). This work establishes ligand desymmetrization as a programmable strategy for metal-selective topological diversification in 2D MOFs.

cond-mat.mtrl-sci↗

arXiv2026

Efficient Waveform Capture without Absorption with Synthesis of Complex Frequencies

A optical waveform can be synthesized by complex-frequency waves as well as by real-frequency harmonic waves. While single complex-frequency wave with exponentially rising waveform can be perfectly absorbed in lossless structures. Here, we propose that diverse optical waveforms can be captured without any absorption through the synthesis of complex frequencies in a lossless system. The scattering matrix zeros of the system correspond to a set of complex frequencies with exponentially rising waveforms, each of which can be virtually and perfectly absorbed. Thus, an input waveform, decomposed into these complex frequencies automatically, can be captured without any absorption. Then, in a well-designed coupled cavity system, various waveforms such as exponentially decaying, Gaussian, rectangular, triangular, and randomly generated piecewise-linear profiles, are captured with high efficiency. The proposed mechanism has potential applications in enhancing light-matter interactions and optical energy storage.

physics.optics↗

arXiv2026

First Observational Evidence for QPO-like Coevolution between Characteristic Damping Timescales and X-ray Time Lags among AGNs

Quasi-periodic oscillations (QPOs) and stochastic variability provide complementary probes of the inner accretion flow around supermassive black holes in active galactic nuclei (AGNs). Previous multi-epoch studies of narrow-line Seyfert 1 galaxy RE~J1034+396 revealed a coevolution between the QPO frequency and the X-ray time lag, but whether the stochastic variability participates in the same structural evolution has remained unclear. We analyze the multi-epoch XMM-Newton observations of RE~J1034+396 and model the soft (0.3--1~keV) and hard (1--4~keV) light curves separately using a damped random walk process. We obtain reliable characteristic damping timescales (CDTs) for 17 observations, with the soft-band CDT consistently longer than its hard-band counterpart. When combined with the X-ray time lag, the hard-band CDT traces a counterclockwise closed loop that closely resembles the previously reported QPO-frequency--time-lag loop, whereas the soft-band CDT exhibits more complex trajectory. Under plausible dynamical, thermal, and viscous interpretations, the hard-band CDT is associated with characteristic scales in the inner hot accretion flow/corona. The observed loops may represent different projections of a common cyclic evolution of the inner accretion flow/corona, as the three timing observables may respond to its dynamical, stochastic, and radiative properties on different timescales. These results provide the first observational support among AGNs for the physical picture in which changes in the spatial extent of the hot inner flow/corona simultaneously affect QPO and stochastic variability.

astro-ph.HE↗

arXiv2026

Theory of Deterministic Photon-Loss Subspaces for Quantum Interferences

Quantum coherence, quantum decoherence, and photon number reduction are coexistent in linear lossy optical systems. However, how these three elements combine together to determine the evolution of the quantum light remains unclear. Here, based on singular value decomposition (SVD), we propose the theory of deterministic photon-loss subspace (DPLS) for quantum interferences in lossy systems. By performing an SVD of scattering matrices with singular values either 0 or 1, a series of completely lossy and lossless input modes are first defined. According to n_1,...,n_i photons in the first,..., i-th lossy modes, the Hilbert space of the input states can be decomposed into a set of orthogonal subspaces H_((n_1,...,n_i ))^in, i.e., deterministic photon-loss subspaces (DPLSs). When the concept of DPLS is established, the input state can be projected onto these DPLSs. In each DPLS, the photons in lossy modes will be completely dissipated, while those in lossless modes experience a unitary evolution. The output state is a statistical mixture of the evolved outcomes of all projections, since decoherence is a concurrent process. Then, based on the DPLS theory, we not only revisited Anti-HOM interference and the distillation of quantum states, but also demonstrate a robust W-state generation for various input states in a three-port lossy system with one-dimensional DPLSs. Through investigating the loss-induced subspace structure of the system, our general theory for analyzing quantum state evolution in lossy systems explicitly reveals the interplay among quantum coherence, quantum decoherence, and photon number reduction. By engineering the loss, the constructed DPLSs can be used to precisely control quantum interferences in dissipative systems, with potential applications in quantum state preparation, quantum logic operations, and other quantum information processes.

quant-ph↗