arXiv subjects

Cheng Peng

Publications and source records attributed to Cheng Peng.

At least 19 recordsLinked to original sources

arXiv2026

TRACE: Memory-Efficient Time Series Class-Incremental Learning via Temporal-Spectral Reconstruction

Web-connected sensing applications continuously generate multivariate time series with evolving behavioral categories, making Time Series Class-Incremental Learning (TSCIL) essential for adapting to new classes without forgetting previously acquired knowledge. However, preserving diverse and evolving class distributions under strict privacy and memory constraints remains a fundamental challenge. Existing approaches often face a trade-off between retaining sufficient historical knowledge for accurate recognition and minimizing the memory required for continual adaptation. To address these limitations, we propose TRACE, a memory-efficient, replay-free framework that preserves class-specific knowledge through generative modeling rather than sample rehearsal. TRACE employs a compact, frozen temporal-spectral representation and lightweight class-specific autoencoders to capture complex intra-class variations. By independently preserving each class model, TRACE avoids recursive generation and catastrophic forgetting, while enabling direct classification through reconstruction-based scoring without retaining historical samples or large statistical matrices. Extensive experiments across five TSCIL benchmarks demonstrate that TRACE achieves competitive classification accuracy while reducing mean retained memory by approximately 98.26% compared with TS-ACL, the strongest evaluated analytical baseline, and achieving an approximately 33.43% relative improvement in accuracy on GrabMyo.

cs.LG↗

arXiv2026

OmniRay: Camera-Agnostic Monocular 3D Reconstruction with Composite Ray Fields

Monocular models recover 3D geometry from a single image in one pass, but most assume a pinhole camera and produce distorted geometry on the wide-angle imagery produced by fisheye and panoramic lenses. Camera-aware models remove this assumption by predicting a per-pixel ray field. However, they represent different camera geometries using a single fixed basis to capture both the overall camera projection and its finer variations. We argue that camera geometry can be viewed as having a hierarchy of complexity for representing rays. We present OmniRay, a feed-forward model whose Ray Module is composed of three components: a monotone radial base that captures radially symmetric structure, a spherical-harmonic residual for smooth deviations from it, and a free-form residual for remaining local variation. These components are composed as successive rotations on the sphere, allowing the representation to accommodate diverse camera geometries. Given a single image, OmniRay predicts a ray direction and an up-to-scale radial distance for every pixel, from which a 3D point cloud is recovered without needing camera information at inference time. Extensive experiments across different camera configurations demonstrate that OmniRay establishes a new state-of-the-art in camera-agnostic monocular 3D reconstruction.

cs.CV↗

arXiv2026

Memorizing Who, Knowing How: Benchmarking Personal and Procedural Memory for Persistent Assistants

Memory systems for persistent personal agents must move beyond isolated memories, maintaining both personal memory about whom they serve and procedural memory about how to act. Crucially, in harness-based personal assistance, procedures or skills are inherently user-conditioned: personal context determines which procedures apply, how they are instantiated, and when they should change. Existing benchmarks typically emphasize either personal or procedural memory, leaving their interaction underexplored. We introduce PraxisMemBench, a harness-based benchmark for persistent personal assistance across daily-life and work trajectories, where personal and procedural memory evolve together and later tasks depend on both. Experiments show that maintaining two memory types separately is insufficient, with completion dropping from single-type to interdependent dependencies (84.6% to 77.9%). More broadly, across existing methods, relation-related failures increase from early to late sessions as interaction history grows (59.8% to 72.9%), revealing a continual-maintenance challenge: writing correct memories is not enough if their evolving content and organization cannot be preserved for later retrieval. To address this, we propose PraxisMem, which uses privileged joint-state supervision to internalize the ability to maintain memory content and organization into the memory manager, with location-aware GRPO and dynamic residual-routed OPSD. PraxisMem reaches 40.78% overall Session Success Rate, compared with 20.60% for Claude Code's native memory. Our code is publicly available at https://github.com/QMMMS/PraxisMemBench.

cs.AI↗

arXiv2026

Exploring Recurrent-Depth Language Models for Multitask Clinical Text Processing

Clinical natural language processing includes diverse tasks such as information extraction, classification, and question answering. Developing one model for all tasks could simplify deployment and improve knowledge transfer. We evaluated whether small recurrent-depth language models can meet this need through joint fine-tuning and compared their performance with standard transformers. We fine-tuned Ouro1.4B-Thinking, Ouro-2.6B-Thinking, HRM-Text-1B, Qwen3-1.7B, and Qwen3-8B on 9 datasets covering information extraction, classification, inference, and question answering, and evaluated them on four additional datasets excluded from training. Across these 13 datasets, Ouro-1.4B-Thinking achieved higher performances than Qwen3-1.7B on 11 datasets; Ouro-2.6BThinking was competitive with Qwen3-8B on most tasks. Ouro models showed the largest gains on inference and question answering, whereas HRM-Text-1B showed mixed results. These findings support further study of recurrent-depth models for clinical tasks through controlled experiments to isolate architectural effects and investigate how recurrent computation influences reasoning, knowledge acquisition, and generalization.

cs.CL↗

arXiv2026

SEHFS: Structural Entropy-Guided High-Order Correlation Learning for Multi-View Multi-Label Feature Selection

In recent years, multi-view multi-label learning (MVML) has attracted extensive attention due to its close alignment to real-world scenarios. Information-theoretic methods have gained prominence for learning nonlinear correlations. However, two key challenges persist: first, features in real-world data commonly exhibit high-order structural correlations, but existing information-theoretic methods struggle to learn such correlations; second, commonly relying on heuristic optimization, information-theoretic methods are prone to converging to local optima. To address these two challenges, we propose a novel method called Structural Entropy Guided High-Order Correlation Learning for Multi-View Multi-Label Feature Selection (SEHFS). The core idea of SEHFS is to convert the feature graph into a structural-entropy-minimizing encoding tree, quantifying the information cost of high-order dependencies and thus learning high-order feature correlations beyond pairwise correlations. Specifically, features exhibiting strong high-order redundancy are grouped into a single cluster within the encoding tree, while inter-cluster feaeture correlations are minimized, thereby eliminating redundancy both within and across clusters. Furthermore, a new framework based on the fusion of information theory and matrix methods is adopted, which learns a shared semantic matrix and view-specific contribution matrices to reconstruct a global view matrix, thereby enhancing the information-theoretic method and balancing the global and local optimization. The ability of structural entropy to learn high-order correlations is theoretically established, and and both experiments on eight datasets from various domains and ablation studies demonstrate that SEHFS achieves superior performance in feature selection.

cs.LG↗

arXiv2026

Mapping AdS$_4$ quantum gravity via the Bethe Ansatz formula

The topologically twisted index of 3d ADHM theory can be written as a sum over solutions to the Bethe Ansatz Equations (BAEs) of the XXZ-type model constructed from quantum toroidal $\mathfrak{g}\mathfrak{l}_1$. Via a novel numerical algorithm, we solve for all $p(N)$ physical solutions at $N\leq 21$, where $p(N)$ denotes the number of integer partitions of $N$. Surveying these solutions, we find that in addition to the known "1-cut" solution dual to dyonic black holes, where the gauge holonomies are distributed along a single curve in the complex plane, there are "dressed" 1-cut solutions dual to combined configurations of the black hole and M2-branes, $m$-cut solutions dual to $\mathbb{Z}_m$ orbifolds of the black hole, a discrete family of solutions interpolating between the 1-cut and 3-cut solutions, together with other exotic solutions, at each value of $N$. The free energy of the interpolating solutions cannot be matched with the on-shell action of any known bulk solution. The $1,2,3$-cut solutions and interpolating solutions all have a well-defined large $N$ limit, and their leading order distribution of gauge holonomies can be computed analytically. Whenever possible, we fit the contributions of the solutions to the index to all orders in the $1/N$ expansion with high-precision numerics, and extract the leading non-perturbative correction. With exceptions, these are matched with the results of equivariant localization in the bulk, and the action of wrapped M2-branes respectively.

hep-th↗

arXiv2026

Rethinking User Encoders: Routing Diffusion States for Scenario-Adaptive User Representation

Existing large language model (LLM)-based user encoders serialize heterogeneous behavioral logs and compress them into static representations through autoregressive (AR) prediction. However, their unidirectional constraint limits bidirectional information flow, hindering cross-modal interactions and the joint preservation of multi-source signals. Consequently, the learned static representations struggle to accommodate the distinct evidence requirements of diverse downstream scenarios. We address these limitations by unlocking the iterative denoising process of diffusion language models (dLLMs) as a multi-granularity reconstruction space. Based on this insight, we propose FINDERS (FactorIzatioN of Diffusion statEs for Routing across Scenarios), a router-equipped dLLM framework comprising four operations: denoise, factorize, route, and compress. Learnable meta-slots organize complementary denoising states into transferable cross-modal semantics and modality-specific behavioral factors. To provide grounded supervision, we construct BehaviorQA, a large-scale corpus containing explicit behavioral rationales. Guided by scenario queries, a lightweight router dynamically emphasizes relevant slot directions via an input-level residual pathway before cross-source fusion and final compression. Extensive experiments and online A/B tests across ten industrial scenarios demonstrate that FINDERS consistently outperforms state-of-the-art methods in both unified and scenario-adaptive settings. This work effectively pushes the boundaries of dLLMs beyond language generation, demonstrating their immense potential as powerful user representation learners.

cs.AI↗

arXiv2026

Act4R: Action-conditioned 4D reconstruction for kinematics-free control

Precise control of movement across various robotic embodiments is a longstanding challenge. Modeling complex robotic structures requires significant time, dedicated sensors, and manual labeling. Recent advancements in 4D representation learning offer an interesting alternative: can we leverage 4D features lifted from 2D observations to densely map a robot's spatial movement based on the action signal? To this end, we propose Act4R, a framework that builds on 3D foundation model to predict a robot's 3D scene flow with any given action. Act4R introduces a Spatial-Action Encoder, which encodes the current 2D robot state and its action signals into embeddings that represent future robot state in 3D. Then, both the current and future spatial embeddings are decoded into point maps to estimate action-conditioned scene flow. To effectively learn the mapping between action and future state, Act4R introduces spatial-action fusion in its transformer-based encoder, and a spatial alignment loss that guides the encoder for future flow prediction. After training on some samples of a given embodiment, Act4R can perform inverse control by optimizing the signal to achieve the desired scene flow. We evaluate Act4R on a public real-world dataset and synthetic data collected in simulator. Experiments on various robotic platforms show that Act4R outperforms the previous SoTA in scene flow prediction. Our method also performs better in inverse control, both on the action accuracy metric and in closed-loop control in simulation.

cs.RO↗

arXiv2026

DIDO: Distilling Interaction-Centric Dynamics into One-Step Denoising for World Action Models

World Action Models (WAMs) use video generation models to predict future visual dynamics for robotic manipulation, but iterative denoising introduces additional latency for closed-loop control. We empirically find that visual content converges at different rates during denoising. Static background structure forms early, whereas the gripper and manipulated object remain blurry after the first step, with their interaction dynamics emerging only through subsequent denoising. Consequently, naively truncating a multi-step video model to one step preserves scene structure but loses the interaction-centric dynamics most critical for manipulation. To address this issue, we propose DIDO, which distills the converged dynamics of a multi-step video model into a single denoising step. DIDO combines distribution matching distillation with interaction-centric representation guidance. Beyond compressing multi-step generation into one forward pass, DIDO explicitly models the gripper, manipulated object, and their interaction using supervised bounding-box visual reasoning tokens. Additionally, DIDO aligns the target object's representations across multiple model layers with features from a pretrained DINOv3 encoder. This interaction-centric guidance helps the distilled model preserve both the relevant entities and their future dynamics in a single step, while substantially reducing inference latency. DIDO achieves an average success rate of 99.0\% on LIBERO, 76.6\% on LIBERO-Plus, and 92.0\% on RoboTwin, while also demonstrating effective transfer to long-horizon and generalization tasks in real-world robotic manipulation.

cs.RO↗

arXiv2026

TileSkipper: Region-Adaptive Tile Pruning for 3D Gaussian Splatting

Tiled 3D Gaussian Splatting rasterizers often use one scene-wide contribution cutoff for tile enumeration, although content differs in its sensitivity to support truncation. TileSkipper selects a static per-Gaussian cutoff policy for a frozen checkpoint. Calibration renders measure candidate pair savings and an isolated-removal distortion proxy that accounts for front transmittance and background color. The method allocates cutoffs across 64 Gaussian groups and accepts policies only after complete renders on disjoint selection views. The exported policy uses one byte per Gaussian, with no parameter updates, additional kernel, or per-frame policy inference. Across 13 scenes from Mip-NeRF 360, Tanks & Temples, and Deep Blending, a fixed-policy AccuTile sweep gives dataset-macro speedups of $1.088\times$ at standard resolution and $1.238\times$ at 3840 pixels wide, with $-0.007/-0.023$ dB mean PSNR change. Six integrations with existing opacity-aware bounds yield $1.009\times$--$1.121\times$ compiler-only speedups. For four ports from $3σ$ rasterizers, we separately attribute the prior exact-bound transition and our incremental gain. Matched-quality ablations show modest gains over scene-global calibration and parity with per-Gaussian control; the standalone comparison with AdaGScale is regime-dependent.

cs.CV↗

arXiv2026

Less Context, Better Geometry: Masked Geometric Encoder for Robust 3D Foundation Models

Recent progress in 3D foundation models has enabled rapid 3D reconstruction and camera calibration by leveraging learned 3D priors from vast amount of spatial data. However, the all-to-all global attention design leads to quadratic complexity and limits long-sequence inference; unconstrained cross-view interactions also can propagate unreliable evidence from occluded or visually similar but geometrically distant views. In this paper, We introduce a Masked Geometric Encoder (MGE), which promotes the learning of robust geometric representations under incomplete cross-view context. During training, MGE strategically drops frame tokens from global attention and distills from a pretrained full-context teacher model. This allows the model to learn an intrinsically richer per-frame representation while providing sufficient intermediate supervision to avoid performance degradation. Through extensive experiments, we show that MGE leads to much stronger performance under occlusion and doppelganger views while retaining high performance on standard benchmarks. Such a richer frame representation also leads to more effective token reduction during inference. To this end, we develop a novel Anchor-Guided Adaptive token merging technique that preserves representative anchor frames while jointly merging redundant tokens from the remaining views. Compared to other efficient inference approaches, we can achieve inference speedup while consistently maintaining higher reconstruction quality, particularly in limited-view settings.

cs.CV↗

arXiv2026

Towards Trustworthy Physical Intelligence: From Theory to Practice Across Life Cycle

Physical intelligence refers to intelligence systems that understand, reason about, and act in accordance with the physical world and its underlying laws, dynamics, and constraints. Unlike conventional AI systems, physical intelligence interacts continuously with uncertain physical environments, and its actions produce consequences that are physically irreversible. As existing trustworthy AI frameworks have been developed primarily for digital AI systems, they do not fully capture the distinctive challenges of Physical Intelligence, such as physical safety, cyber-physical security, and physical manufacturing process. To address this gap, we present a survey of trustworthy physical intelligence principles. First, we characterize the core capabilities and challenges of physical intelligence. Second, we examine the role of physics in AI. Third, we trace the end-to-end physical intelligence life cycle across five core stages and introduce Trustworthy Physical Intelligence Operationalization (T-PAIO). Fourth, we develop the Trustworthy Physical Intelligence (T-PAI) framework, a theoretical framework that organizes key trustworthiness principles and provides a foundation for governing trustworthy physical intelligence systems.

cs.AI↗

arXiv2026

Wrivinder: Towards Spatial Intelligence for Geo-locating Ground Images onto Satellite Imagery

Aligning ground-level imagery with geo-registered satellite maps is crucial for mapping, navigation, and situational awareness, yet remains challenging under large viewpoint gaps or when GPS is unreliable. We introduce Wrivinder, a zero-shot, geometry-driven framework that aggregates multiple ground photographs to reconstruct a consistent 3D scene and align it with overhead satellite imagery. Wrivinder combines SfM reconstruction, 3D Gaussian Splatting, semantic grounding, and monocular depth--based metric cues to produce a stable zenith-view rendering that can be directly matched to satellite context for metrically accurate camera geo-localization. To support systematic evaluation of this task, which lacks suitable benchmarks, we also release MC-Sat, a curated dataset linking multi-view ground imagery with geo-registered satellite tiles across diverse outdoor environments. Together, Wrivinder and MC-Sat provide a first comprehensive baseline and testbed for studying geometry-centered cross-view alignment without paired supervision. In zero-shot experiments, Wrivinder achieves sub-30\,m geolocation accuracy across both dense and large-area scenes, highlighting the promise of geometry-based aggregation for robust ground-to-satellite localization.

cs.CV↗

arXiv2026

A baby universe from a large family: booklet cosmology states and quantum error correction

We propose an extension of the cosmological-state construction of Antonini, Sasieta and Swingle (AS$^2$) to three or more holographic CFTs. Each CFT resides on the asymptotic boundary of an AdS page, and all pages are glued along a common interface by imposing the multiway junction conditions. The associated states are prepared by Euclidean evolution with a multilinear insertion. In appropriate heavy insertion limit the bulk geometry develop a closed universe in the center and we refer the state in this limit as a booklet cosmological state. Extending the effective Gaussian assumption used in AS$^2$, we model the three-page insertion by a circular complex Gaussian random tensor, yielding a tripartite Haar state in flat energy windows. In this Gaussian model of the cosmology-to-boundary map, each arm is a network branch associated with one boundary CFT. In the simplest example of three pages with three equal output Hilbert-space dimension $b$, we show that a prescribed code of dimension $K$ is approximately recoverable from any two arms, with vanishing error and high probability as $K/b\to0$.

hep-th↗

arXiv2026

Learning Latent Algebraic Structure from Ambiguous Set Observations

We study when statistically learnable latent structure can also be recovered efficiently, and how membership queries change the answer. An unknown support $A\subseteq\mathbb F_2^n$ has small additive doubling and is observed through a fixed set $B$ satisfying $|A\triangle B|\leη|A|$. We seek one linear subspace $V$ such that every compatible support $A$ is covered by few $V$-cosets and satisfies $|V|\le|A|$. For every $η<1$, polynomially many uniform samples suffice statistically, with cost polynomial in the doubling constant and proportional to $(1-η)^{-1}$; this radius dependence is sharp. Under a specified hardness assumption for learning parities with noise (search-LPN), however, no polynomial-time sample-only learner achieves even constant covering cost, including when the latent support is unique. At fixed structural parameters and the same constant covering budget, adding exact membership queries to $B$ permits polynomial-time recovery. The general query learner constructs a short structural list and uses fresh samples to select one common output through a majority-coverage rule. Persistent structured cores make this candidate construction possible. At doubling one, a complementary distinction appears at $η=1/3$: coarse recovery remains polynomial time, while exact recovery requires exponentially many accesses in the worst case when latent cardinality is unknown.

cs.DS↗

arXiv2026

CriPO: Enhancing Rubric-based RL via Self-Distillation

Rubric-based Reinforcement Learning (RL) has recently shown promise in improving Large Language Models (LLMs) on open-ended tasks. A widely recognized limitation of rubric-based RL is limited exploration: criteria that no rollout manages to satisfy (Unexplored Criteria) receive no optimization signal. Recent methods address this by incorporating rubric information as external guidance during rollout generation, yet they introduce a train-inference mismatch: the policy is optimized on rollouts produced under external guidance while this guidance is absent at inference time, causing error accumulation through autoregressive decoding. Moreover, these exploration-focused approaches overlook a fundamentally different failure mode that we term Suppressed Criteria---criteria that are satisfied by some rollouts yet whose learning signals might be lost during optimization because scalar reward aggregation assigns them non-positive aggregate advantages. Our analysis reveals that suppressed criteria constitute a persistent and non-negligible failure mode---over 25% of samples exhibit this issue throughout training. To simultaneously address both unexplored and suppressed criteria without introducing training-inference mismatch, we propose Criterion-Distilled Policy Optimization (CriPO), which enhances rubric-based RL via on-policy self-distillation. For unexplored criteria, CriPO constructs a behavior-injection teacher and computes a filtered forward-KL loss to inject missing behaviors into the policy. For suppressed criteria, CriPO uses a counterfactual teacher to locate criterion-relevant tokens in negative-advantage rollouts, and corrects their advantages in GRPO to preserve useful patterns. Experiments on medicine and science benchmarks demonstrate that CriPO outperforms existing rubric-based RL methods, e.g., achieving an average gain of 3.3 points over GRPO on Qwen3-4B.

cs.LG↗

arXiv2026

Two Heads Are Better Than One: Aggregating Weaker LLMs for Better Forecasts

Large language models (LLMs) are increasingly used to forecast real-world events, but access to the strongest individual forecaster may be costly or otherwise constrained. We study weak-to-strong forecast aggregation: can individually weaker LLM forecasters be aggregated to outperform a stronger forecaster? Using ForecastBench (Karger et al., 2025), we evaluate 70 LLM forecasters across 16 comparison groups, each with more than 1,000 shared subquestions, yielding 1,121 weaker-model pairs. Within each group, we identify the strongest individual by test Brier score and evaluate aggregates composed exclusively of weaker forecasters, with aggregation weights learned on separate training data. We find substantial evidence of weak-to-strong improvement. Learned linear pooling identifies a weaker pair that matches or outperforms the strongest individual in 11 of 16 groups and comes within 5% of its Brier score in all 16 groups. We also find that these improvements do not rely on having a near-best constituent and are generally accompanied by good calibration. Additional analyses show that adding more models does not consistently improve performance, and competitive weaker-model aggregates also remain available under practical constraints.

cs.LG↗

arXiv2026

MedGPT-oss: Training a General-Purpose Vision-Language Model for Biomedicine

Biomedical multimodal assistants have the potential to unify radiology, pathology, and clinical-text reasoning, yet a critical deployment gap remains: top-performing systems are either closed-source or computationally prohibitive, precluding the on-premises deployment required for patient privacy and PHI compliance. We introduce MEDGPT-OSS, an open-weight, 20B-parameter generalist vision-language model designed to facilitate open research in clinical AI. Rather than relying on architectural complexity, MEDGPT-OSS pairs the GPT-oss language backbone with a visual front-end via a optimized, three-stage training curriculum. By progressively domain-adapting these modules through rigorous data curation and long-context multimodal alignment, we demonstrate that a 20B model can bridge the capacity gap. It successfully outperforms larger open medical models on out-of-distribution (OOD) multimodal reasoning and complex text-only clinical tasks. By unifying diverse modalities under a single instruction-following interface, MEDGPT-OSS maintains a parameter-efficient footprint fully compatible with commodity GPUs. We release the complete training recipe, open-weight checkpoints, and a rigorous evaluation harness to serve as a verifiable foundation for privacy-preserving, institution-specific clinical AI research.

cs.CL↗