Academic Intelligence · Curated Daily

探索全球前沿学术脉络

AcademicHub 汇聚顶级期刊与预印本平台的实时文献。定制您的专属科研雷达,利用大语言模型自动生成交叉领域文献分析简报。

01.
medRxiv (Medicine) 2026-06-15

Filum Terminale Diameter on Routine Pediatric MRI: A Large-Cohort Clinical Reference in 3,406 Children and the Age-Dependent Meaning of the 2-mm Thickened-Filum Threshold

Background. A filum diameter >2 mm is the conventional MRI threshold for a thickened filum, but it derives from small, mostly adult series showing no age dependence; whether one cutoff suits all of childhood is untested. Objective. To build an age-specific filum-diameter reference on routine pediatric MRI and test, adjusting for image resolution, whether the 2-mm threshold is age-stationary. Materials and methods. In this retrospective study an nnU-Net tracer measured the maximal filum diameter on consecutive lumbosacral MRI; versus manual tracing it showed negligible bias but moderate single-measure agreement. After excluding report-confirmed fatty filum, lipoma, or tethered cord, the proportion >2 mm was analysed within one acquisition protocol and by logistic regression adjusting for voxel size and slice thickness. Results. Of 7,245 examinations, 3,869 (53%) were traceable; untraced ones were younger (median 0.75 vs 2.0 years). The presumed-normal cohort had median diameter 1.48 mm. At matched resolution, 2 mm marked the 94th percentile in infants (5.6% exceeded it) but the 83rd by 3-6 years (17.4%); the age effect persisted after adjusting for voxel size and slice thickness (3-6 years vs infants, adjusted OR 4.7; P < .001). Conclusion. Filum diameter clusters near 1.5 mm, and the fixed 2-mm cutoff flags ~5% of infants but ~17% of preschoolers. Caliber should be judged against an age-specific clinical reference, not one fixed cutoff; a thick filum is not itself a diagnosis of tethered cord.

02.
arXiv (CS.LG) 2026-06-16

Overcoming Rank Collapse in Feedback Alignment

arXiv:2606.11123v2 Announce Type: replace Abstract: Backpropagation (BP) is widely viewed as biologically implausible, in part because it requires feedback weights to be the transpose of forward weights for error propagation. Interestingly, when training a network with fixed random feedback weights to circumvent this issue, learning aligns the forward weights with the feedback weights, leading the backpropagated error signal to become an approximation of the standard gradient used by BP. This process, called Feedback Alignment (FA), occurs in MLPs and very shallow CNNs but does not scale well to deeper architectures. In this work, we first investigated differences between BP and FA models, trained on CIFAR10, specifically focusing on the effective rank of the signal. We found that the FA error has a considerably lower rank and hence is constrained to a lower-dimensional subspace compared to BP, limiting exploration of the parameter space. Motivated by this observation, we evaluated two mechanisms for increasing the effective dimensionality of FA: Muon, an optimiser that orthogonalises weight updates; and hidden activity normalisation, which promotes activation orthogonality. Across larger architectures and benchmarks, we find that these methods consistently improve over FA baselines, for example, on CIFAR100 with a Resnet-18, accuracy increases by 9 percentage points. Our results identify low-dimensional gradient dynamics as a key obstacle to scaling FA and suggest that inducing higher-dimensional update geometry is a promising route toward scaling alternatives to backpropagation.

03.
arXiv (CS.AI) 2026-06-17

DRFLOW: A Deep Research Benchmark for Personalized Workflow Prediction

arXiv:2606.18191v1 Announce Type: new Abstract: Deep research (DR) systems are increasingly used for complex information-seeking tasks, but existing works mainly focus on generating reports and summaries. In contrast, many enterprise tasks instead require an agent to identify concrete workflows which is a sequence of action-steps. For example, rather than summarizing budgeting policies, an agent should be able to determine the steps needed to answer a question such as: "How do I request new headcount given a fixed budget?". Therefore, we introduce DRFLOW, a benchmark for evaluating personalized workflows predicted by agents from heterogeneous sources. Each task requires the agent to identify relevant evidence from scattered sources, then use that evidence to predict the correct action-step sequence for the user's task. DRFLOW contains 100 tasks across five domains, with 1,246 reference workflow steps grounded in more than 3,900 sources. We define seven diagnostic metrics covering factual grounding, step recovery, structural ordering, condition resolution, and personalization. We further present DRFLOW-Agent (DRFA), a workflow-oriented reference agent to predict personalized workflow. We show that although DRFA improves over strong baseline agents (upto 10.02% average F1 score), there is substantial room for improvement remains across these workflow metrics, indicating that predicting complete and correct personalized workflows remains a challenging frontier for deep research.

04.
arXiv (CS.AI) 2026-06-11

Improving Generalization and Data Efficiency with Diffusion in Offline Multi-agent RL

arXiv:2307.01472v2 Announce Type: replace Abstract: We present a novel Diffusion Offline Multi-agent Model (DOM2) for offline Multi-Agent Reinforcement Learning (MARL). Different from existing algorithms that rely mainly on conservatism in policy design, DOM2 enhances policy expressiveness and diversity based on diffusion model. Specifically, we incorporate a diffusion model into the policy network and propose a trajectory-based data-reweighting scheme in training. These key ingredients significantly improve algorithm robustness against environment changes and achieve significant improvements in performance, generalization and data-efficiency. Our extensive experimental results demonstrate that DOM2 outperforms existing state-of-the-art methods in all multi-agent particle and multi-agent MuJoCo environments, and generalizes significantly better to shifted environments {(in $28$ out of $30$ settings evaluated)} thanks to its high expressiveness and diversity. Moreover, DOM2 is ultra data efficient and requires no more than $5\%$ data for achieving the same performance compared to existing algorithms (a $20\times$ improvement in data efficiency).

05.
medRxiv (Medicine) 2026-06-16

Preventing postpartum depression through mitigating breastfeeding grief: A convergent parallel mixed methods study

Background: Women who did not meet their breastfeeding goals often experience breastfeeding grief (BG) and may be likely to have postpartum depression (PD). Furthermore, PD is nearly twice as common in African American (AA) women as in Non-Hispanic White women. No research exists on BG and its role in PD. This study examined the BG experiences of AA women and its possible contributions to PD symptoms. Methods: A convergent parallel mixed methods design was used. A purposive sample of 16 AA women with children aged 6 months to 2 years with BG participated in individual semi-structured interviews about their experiences of BG and completed an online survey including the Edinburgh Postnatal Depression Scale (EPDS). Qualitative and quantitative data were analyzed using reflexive thematic analysis and descriptive statistics, respectively. Both data were integrated using joint display of data and side-by-side comparison. Results: The mean age of participants was 29.5 years. Four meaning-based themes about BG were generated including: We looked forward to breastfeeding, But it did not go as expected, So we grieve, and These would have helped. From quantitative results, 87.5% of participants reported a history of PD symptoms and almost 44% had EPDS scores >11. All participants reported that experiencing BG contributed to their PD symptoms. Findings suggest that BG influenced PD symptoms in AA women without prior diagnosis of depression. Conclusions: Qualitative and quantitative findings from this novel exploratory study revealed an overlap that AA women with BG report PD symptoms. Clinicians should support women to achieve their breastfeeding goals to prevent BG and PD. Keywords: African American; Breastfeeding grief; Mental health; Mixed methods; Postpartum depression

07.
arXiv (CS.AI) 2026-06-16

Fast LLM-Based Semantic Filtering: From a Unified Framework to an Adaptive Two-Phase Method

arXiv:2606.08090v2 Announce Type: replace-cross Abstract: Evaluating a natural-language yes/no predicate over a document corpus under an accuracy target - the semantic filter - is a cornerstone of LLM-based data processing. Calling the LLM on every document (the oracle) is prohibitive, so cascades pair the oracle with a fast proxy. As deployed today, they leave four limitations on the table. (1) Each cascade family - model-free clustering, prebuilt small-LLM proxies, online-trained proxies - commits to a single representation and pipeline, and wins on only a narrow query regime. (2) The strongest online proxy invests in a custom training scheme on a bi-encoder over dense embeddings, missing the token-level evidence richer predicates require. (3) The proxy is trained against binary yes/no labels, wasting the LLM's per-document confidence at the boundary documents it most needs to learn. (4) Existing calibrations add a uniform safety margin, conflating genuine proxy uncertainty with small-sample noise and inflating cascade cost. We address these by (1) composing families adaptively - model-free clustering first, online proxy only when needed, with oracle calls shared across phases; (2) replacing the cosine bi-encoder with a hybrid of off-the-shelf token-aware models; (3) training the proxy with the oracle's per-document confidence as a soft label; and (4) a calibration that adds the safety margin only where the labeled sample is sparse. We are also the first to use the oracle's per-document confidence for three purposes: a query-level difficulty compass, a lower bound on the minimum oracle calls any proxy-based cascade can make, and the proxy's soft training label. At a 90% accuracy target on three 10K-document corpora, our methods are 1.6-2.0x faster than the best prior method per corpus and meet the target on 95% of queries; the BER-derived lower bound indicates a further ~4-20x of headroom for future work.

08.
arXiv (CS.CL) 2026-06-15

CORA: Analyzing and bridging thinking-answer gap in Multimodal RLVR via Consistency-Oriented Reasoning Alignment

Reinforcement learning with verifiable rewards (RLVR) has successfully elicited the reasoning capabilities of large language models, motivating its extension to multimodal scenarios. Existing methods primarily focus on improving the visual coverage of reasoning traces and mitigating visual hallucinations, but underestimate the semantic inconsistency between the reasoning process and the final answer. In this paper, we delve into thinking-answer inconsistency in RLVR for large vision-language models (LVLMs), showing thorough analyses of rollouts collected throughout Group Relative Policy Optimization (GRPO) training process and post-RLVR evaluation outputs that this issue persists during training and remains present during inference. Motivated by the analysis, we propose Consistency-Oriented Reasoning Alignment (CORA), which introduces thinking-answer semantic consistency into RLVR through a lightweight plug-and-play consistency reward model, and further incorporates Hybrid Reward Advantage Splitting (HRAS) to stably coordinate task and consistency optimization. Extensive experiments across representative multimodal reasoning benchmarks and mainstream LVLMs show that CORA improves task performance while effectively mitigating thinking-answer inconsistency, leading to more faithful reasoning traces.

09.
arXiv (CS.CL) 2026-06-12

Multi-Turn Reasoning When Context Arrives in Pieces: Scalable Sharding and Memory-Augmented RL

When a user reveals task-critical information across several conversation turns, LLM accuracy drops by up to 65% despite full context availability. We show that this Lost in Conversation degradation can be substantially mitigated by training models to maintain a compact rolling memory instead of attending to a growing history. To make such training scalable, we introduce a low-cost sharding pipeline that converts single-turn QA datasets into multi-turn fragmented-information episodes, eliminating the need for hours of manual annotation. Training only on sharded GSM8K, our memory-augmented policy significantly improves multi-turn accuracy and generalises zero-shot to harder math and out-of-domain long-context QA. Moreover, memory-trained models outperform full-history baselines even when given the full history at test time, suggesting that learning to compress induces more robust incremental reasoning than full-context exposure alone.

10.
arXiv (quant-ph) 2026-06-17

Demultiplexing Generalized Information via Quantum Transmission Lines

arXiv:2606.17894v1 Announce Type: new Abstract: Demultiplexers are the fundamental primitives of network architecture, enabling perfect routing of an input classical signal to a designated one, among multiple output ports. Quantum transmission lines, having access to the quantum systems directly, are able to transmit both the classical and quantum information encoded in quantum systems. A natural question therefore emerges that whether the scrambled classical and quantum information in a quantum system can be perfectly demultiplexed in the designated classical and quantum output ports? Here we answer this question by introducing a quantum to quantum-classical device, namely the quantum demultiplexer (Q-DEMUX). We characterize the class of Q-DEMUXs enabling perfect routing of both the classical and the quantum information along with their simple circuit realizations. Our results highlight an explicit connection between the strength of a Q-DEMUX with the incompatibility of quantum instruments. Finally, we extend the notion in a stronger variant where the sender is oblivious regarding the nature of the data to be transmitted through the Q-DEMUX.

12.
arXiv (CS.CL) 2026-06-16

DEEPRUBRIC: Evidence-Tree Rubric Supervision for Efficient Reinforcement Learning of Deep Research Agents

Deep research agents synthesize long-form reports by searching and reasoning over retrieved evidence. Reinforcement learning with rubric-based rewards improves these agents by optimizing them against checkable criteria that translate report quality into reward signals, but its efficiency depends on whether those criteria reliably capture the task scope and evidence needs. Most existing studies ask an LLM to generate rubrics for a given query, but when the model fails to infer the underlying information needs, the generated rubrics may be incomplete and reduce RL efficiency. To obtain more reliable query–rubric supervision, we introduce DeepRubric, a data construction framework that reverses this process: instead of inferring evaluation criteria for a given query, it first determines what an evidence-backed report should be evaluated on and then synthesizes aligned query–rubric pairs from those evaluation targets. Starting from a sampled seed topic, DeepRubric builds an evidence tree by recursively expanding evidence-backed sub-questions, whose leaves serve as atomic and verifiable evaluation targets. It then uses the evidence tree to synthesize the training query and rubrics, ensuring that the reward evaluates exactly the information requested by the query. Using DeepRubric, we construct 9K query–rubric supervision examples and train DeepRubric-8B with rubric-based GRPO, achieving comparable performance to prior open state-of-the-art deep research models across three benchmarks with roughly 13x fewer RL GPU-hours.

13.
arXiv (CS.LG) 2026-06-12

Positional Encoding in the Context of Memristor-Based Analog Computation for Automatic Speech Recognition

arXiv:2606.13379v1 Announce Type: new Abstract: Memristors provide a new chance for resource-efficient computation of neural models for natural language processing by enabling analog execution of vector-matrix-multiplication. Yet, computations on these devices are currently subject to larger distortion, both in weight programming and execution. In this work, we identify large output values of transformed positional encodings to cause major degradation within analog-to-digital conversion (ADC) as part of memristor-based computation. By adjusting the proportion of weight and precision bits of the ADC of specific memristor layers, we reduce the degradation of the execution by ~50% relative, while keeping the estimated energy consumption stable. Additionally, we investigate scenarios where the ADC cannot be modified. In that case the degradation can be reduced by ~30% relative after removing encoding-related linear transformations.

14.
arXiv (CS.CL) 2026-06-17

Precision Is Not Faithfulness: Coverage-Aware Evaluation of Grounded Generation with a Complete Oracle

Reference-free faithfulness metrics verify each atomic claim a model makes against ground truth, and are increasingly used to evaluate grounded generation. We show they share a blind spot: they measure only precision – are the stated claims supported? – and therefore reward abstention, since a model can score near-perfect faithfulness by saying almost nothing. We make this measurable using Formula 1 telemetry, a domain where strategic ground truth is derived deterministically and, crucially, completely: for each decision we know the full set of facts that mattered. This completeness – absent in open-domain faithfulness benchmarks – lets us measure recall (coverage of the relevant facts) exactly, alongside precision. On a multilingual (EN/ES/PT) benchmark of 7,253 decision instances spanning 157 races, the most precise frontier model covers under half of the relevant facts and ranks last by F1, so requiring coverage reorders the systems; the same effect reappears in a second complete-oracle domain (NOAA weather forecasts). Fine-tuning small models (1B-7B) on the complete oracle closes the precision-recall gap entirely (F1 ~0.98), beating every zero-shot frontier system regardless of scale. We pair faithfulness with coverage into a single score, validate the metric (controlled perturbation; agreement across a model-free regex extractor and a cross-family LLM extractor, system-level Spearman 1.0), and give a verifier-guided generation method that improves precision and recall without references. We release the benchmark, structured annotations, metric, baselines, and an interactive demo.

15.
arXiv (CS.LG) 2026-06-15

MAD: Manifold Attracted Diffusion

arXiv:2509.24710v3 Announce Type: replace-cross Abstract: Score-based diffusion models are a highly effective method for generating samples from a distribution of images. We consider scenarios where the training data comes from a noisy version of the target distribution, and present an efficiently implementable modification of the inference procedure to generate noiseless samples. Our approach is motivated by the manifold hypothesis, according to which meaningful data is concentrated around some low-dimensional manifold of a high-dimensional ambient space. The central idea is that noise manifests as low magnitude variation in off-manifold directions in contrast to the relevant variation of the desired distribution which is mostly confined to on-manifold directions. We introduce the notion of an extended score and show that, in a simplified setting, it can be used to reduce small variations to zero, while leaving large variations mostly unchanged. We describe how its approximation can be computed efficiently from an approximation to the standard score and demonstrate its efficacy on toy problems, synthetic data, and real data.

16.
arXiv (quant-ph) 2026-06-12

Vacuum photon emission and mean electromagnetic field in pair-creating external backgrounds

arXiv:2606.12547v1 Announce Type: cross Abstract: We develop a perturbative description of vacuum radiative processes in quantum electrodynamics with a prescribed external electromagnetic background capable of producing electron-positron pairs. Since the initial vacuum is then unstable and the in- and out-vacua are inequivalent, radiative observables require a real-time formulation beyond the ordinary in-out approach of vacuum-stable QED. Using the Keldysh-Schwinger-Fradkin nonequilibrium technique, we derive the mean number density of emitted photons through the second nonvanishing order in the fine-structure constant. The leading term, of order $\alpha$, reproduces the known vertex and tadpole mechanisms, while the complete order-$\alpha^2$ correction contains interference, loop, and induced-current contributions. We also give an independent derivation based on the spectral decomposition of the identity operator in the in-Fock space, where the photon number density is represented as a sum of squared transition amplitudes and vacuum-disconnected terms are canceled by the optical theorem generalized to an unstable vacuum. In addition, we compute the mean electromagnetic field through order $e^3$, including the electromagnetic dressing of the induced vacuum current, and verify it using the corresponding Schwinger-Dyson equations. The final formulas are expressed in terms of exact solutions and propagators of the Dirac equation in the external background and apply to general spacetime-dependent field configurations.

17.
arXiv (CS.AI) 2026-06-16

ALCL: An Adaptive Log-Correntropy Loss for Robust Learning under Non-Gaussian Noise

arXiv:2606.16050v1 Announce Type: cross Abstract: Robust deep learning under heavy-tailed and impulsive noise remains challenging because conventional losses such as mean squared error (MSE) exhibit unbounded sensitivity to outliers. Although correntropy-based objectives improve robustness, existing formulations rely on fixed kernel parameters that must be empirically tuned and remain static during training. To address these limitations, we propose an Adaptive Log-Correntropy Loss (ALCL), a heavy-tailed loss formulation that adaptively learns its robustness geometry during optimization. ALCL introduces a logarithmic residual model whose shape and scale parameters are learned jointly with network weights through differentiable reparameterization. This yields a principled maximum likelihood formulation whose influence function is formally bounded and redescending, allowing the loss geometry to adapt dynamically to evolving residual statistics while suppressing extreme outliers. Comparative experiments on four widely used benchmark datasets spanning grayscale and red-green-blue (RGB) image data under mixed heavy-tailed and impulsive noise demonstrate that ALCL consistently outperforms MSE and optimally tuned generalized correntropy losses in both reconstruction fidelity and downstream classification accuracy. While performance differences remain small under low-noise conditions, under high-noise regimes ALCL improves median accuracy by up to 4.75% on grayscale benchmarks and 4.51% on RGB datasets, with reduced variance across runs. These results demonstrate that adaptive robustness through joint learning of loss parameters provides a computationally efficient alternative to static correntropy-based losses for deep learning in non-Gaussian environments.

18.
arXiv (CS.CL) 2026-06-16

Simplifying the Modeling of Arbitrary Conditionals in Natural Language

Causal Transformers model sequences through an autoregressive factorization of the joint distribution, which enables efficient left-to-right decoding and conditional likelihood computation. However, they cannot tractably sample from or evaluate arbitrary conditionals – e.g., a block of text conditioned on past and future tokens. Recent work aims to solve this problem through novel architectures, but they often lead to sub-optimal modeling of such conditionals and degraded generations. We propose Arbitrary Conditionals GPT (AC-GPT) which introduces a simple modification to standard causal Transformers to enable evaluating and sampling from arbitrary conditionals – including past, future, and mixed contexts – within a single forward pass. Unlike prior approaches, our method preserves the standard left-to-right ordering and next-token prediction objective essential for both strong performance and efficient training on natural language. Crucially, this compatibility allows existing LLMs to be fine-tuned for arbitrary conditioning. Our empirical results indicate that our method outperforms baselines on modeling arbitrary conditionals, without degrading standard left-to-right performance.

19.
arXiv (CS.LG) 2026-06-16

Towards Data-Efficient Cross-Device Generalization of Grad-Shafranov Equilibria via Transfer Learning Neural Operator

arXiv:2606.15512v1 Announce Type: new Abstract: Real-time reconstruction of magnetohydrodynamic equilibria is essential for plasma shaping, stability assessment and feedback control in magnetic confinement fusion. However, Grad-Shafranov equilibrium calculations remain largely device-specific and iterative, limiting their use in latency-constrained control settings. Existing neural approaches can accelerate individual equilibrium predictions, but they do not generally provide reusable models across changing plasma boundaries or tokamak geometries. Here we show that equilibrium reconstruction can be recast as a cross-device operator learning problem. We develop a domain-specific neural operator framework that maps geometry and profile parameters directly to the poloidal flux field, replacing repeated solve-on-demand computation with amortized operator inference. Using the analytically tractable Solov'ev family as a controlled Grad-Shafranov testbed, we generate equilibria across eight geometrically distinct tokamak-like configurations and benchmark five neural operator architectures under four transfer-learning strategies. Single-geometry pretraining gives poor transfer to unseen devices, whereas multi-geometry pretraining enables data-efficient adaptation. The Wavelet Neural Operator gives the strongest cross-geometry performance, reaching mean relative L2 errors below 4% with 100 labelled target equilibria and below 2% with full fine-tuning. The predicted magnetic fields satisfy the divergence-free constraint to numerical precision, and four architectures achieve millisecond or sub-millisecond inference. These results identify neural operator pretraining as a route towards reusable, real-time equilibrium inference across fusion device configurations.

20.
arXiv (CS.AI) 2026-06-19

Overcoming Labelled Data Scarcity for Defect Classification in Scanning Tunneling Microscopy

arXiv:2506.01678v2 Announce Type: replace-cross Abstract: Scanning tunnelling microscopy (STM) is a powerful technique for imaging surfaces with atomic resolution, providing insight into physical and chemical processes at the level of single atoms and molecules. A regular task of STM image analysis is the identification and labelling of features of interest against a uniform background. Performing this manually is a labour-intensive task, requiring significant human effort. To reduce this burden, we propose an automated approach to the segmentation of STM images that uses both few-shot learning and unsupervised learning. Our technique offers greater flexibility compared to previous supervised methods; it removes the requirement for large manually annotated datasets and is thus easier to adapt to an unseen surface while still maintaining a high accuracy. We demonstrate the effectiveness of our approach by using it to recognise atomic features on three distinct surfaces: Si(001), Ge(001), and TiO$_2$(110), including adsorbed AsH$_3$ molecules on the silicon and germanium surfaces. Our model exhibits strong generalisation capabilities, and following initial training, can be adapted to unseen surfaces with as few as one additional labelled data point. This work is a significant step towards efficient and material-agnostic, automatic segmentation of STM images.

21.
Nature (Science) 2026-06-10

Daily briefing: Ancient ground squirrels ate like ‘zombies of the Pleistocene’

作者:

Evidence from fossilized poo reveals the diverse diet of ancient ground squirrels. Plus, the science behind the peptide craze and our innate tendency to wander anticlockwise. Evidence from fossilized poo reveals the diverse diet of ancient ground squirrels. Plus, the science behind the peptide craze and our innate tendency to wander anticlockwise.

22.
arXiv (CS.LG) 2026-06-19

Minimal Filling Architectures of Polynomial Neural Networks: Counterexamples, Frontier Search, and Defects

arXiv:2605.09609v2 Announce Type: replace Abstract: We provide counterexamples to the unimodal minimal filling architecture conjecture for polynomial neural networks (PNNs) with power activation functions. Fixing the input and output widths, the conjecture states that any minimal filling architecture has unimodal widths for the hidden layers. We found counterexamples via a frontier search, recursive dimension bounds on neurovarieties, and symbolic computation. Notably, several subarchitectures of our main example exhibit large defect, in contrast with the predominantly small-defect behavior observed in prior literature.

23.
arXiv (CS.AI) 2026-06-12

Prism: Cost-Efficient Multi-LLM Serving via GPU Memory Ballooning

arXiv:2505.04021v3 Announce Type: replace-cross Abstract: Inference providers must maintain availability for many LLMs, including low-volume but essential models, making resource efficiency increasingly important as token prices fall. Analysis of production traces reveals a dynamic bursty-group pattern in which sets of models become active together and shift over time; existing space- and time-sharing approaches lack principled mechanisms to adapt to this variability, forcing trade-offs between SLO adherence and efficiency. We observe that elastic memory allocation can unify spatial and temporal sharing. Based on this insight, we have developed Prism, a memory-centric LLM co-serving framework that applies memory ballooning to reclaim memory across models and support both forms of sharing under a single scheme. Prism's balloon driver, referred to as kvcached, has been open-sourced at https://github.com/ovg-project/kvcached, and deployed in production environments across 10K+ GPUs.

24.
arXiv (CS.CV) 2026-06-18

Mem-World: Memory-Augmented Action-Conditioned World Models for Persistent Robot Manipulation

Action-conditioned world models have emerged as a promising paradigm for robot learning, offering a scalable alternative to costly real-world experimentation by generating action-consistent video rollouts. However, persistent world modeling remains challenging in manipulation: frequent end-effector occlusions and rapid wrist-camera motion make the current observation insufficient for predicting future views, causing models to forget or hallucinate scene details seen in earlier frames. Existing memory retrieval strategies often fail to identify informative history in dynamic manipulation scenarios. To address this limitation, we propose Mem-World, a memory-augmented multi-view action-conditioned world model. At its core, we present W-VMem, a 4D wrist-view-centered surfel-indexed memory that anchors historical observations to temporally evolving surface elements. By explicitly modeling when and where scene elements are observed, W-VMem enables geometry-aware retrieval of relevant history frames conditioned on future actions. During generation, relevant history frames are selected via surfel-based rendering and scoring, providing informative and non-redundant context for prediction. Extensive experiments show that Mem-World generates persistent rollouts in complex manipulation scenarios, enables more reliable policy evaluation than Ctrl-World, improving the Pearson correlation with real-world performance by 14.5\%, and supports effective policy improvement through synthetic data generation, increasing success rates from 58\% to 72\% on long-horizon tasks.

25.
arXiv (CS.CL) 2026-06-16

Long-Context Modeling via GSS-Transformer Hybrid Architecture with Learnable Mixing

Modeling long-range dependencies remains a central challenge in natural language processing. Transformer architectures achieve strong performance via self-attention but scale quadratically ($O(N^2)$) with sequence length, while State Space Models (SSMs) scale linearly ($O(N)$) but suffer from a selective recall bottleneck, struggling to retrieve precise information from compressed states. This creates a fundamental tradeoff between efficiency and perplexity. To tackle these challenges, we propose the Parallel Hybrid Architecture (PHA), which runs Gated State Spaces (GSS), Grouped Query Attention (GQA), and Feed-Forward Networks (FFNs) as independent parallel branches fused by a learnable mixing mechanism. Instead of forcing SSMs to approximate attention or serializing the two paradigms, PHA allows each branch to specialize: GSS captures global context, while attention performs selective retrieval, with FFN providing complementary processing. On WikiText-103, PHA achieves 16.51 PPL at 125M parameters, outperforming Hedgehog (16.70) and H3-125M (23.70). Scaling to 180M parameters yields 16.42 PPL, which gives comparable results with the pure attention baseline while delivering 24\% higher throughput and up to 40\% lower memory usage at long contexts. On OpenWebText, our 125M model achieves 19.72 PPL, outperforming standard Transformers (20.60) and GSS hybrid baselines (19.80). These results demonstrate that separating sequence modeling paradigms into parallel specialists enables Transformer-level perplexity with substantially improved efficiency for long-context language modeling.