Model-based reinforcement learning (MBRL) is sample-efficient but struggles in sparse reward settings. A critical bottleneck arises from the lack of informative gradients in sparse settings, where standard reward models often yield flat landscapes that struggle to guide planning. To address this challenge, we propose Shaping Landscapes with Optimistic Potential Estimates (SLOPE), a novel framework that shifts reward modeling from predicting sparse scalars to constructing informative potential landscapes. SLOPE employs optimistic distributional regression to estimate high-confidence upper bounds, which amplifies rare success signals and ensures sufficient exploration gradients. Evaluations on 30+ tasks across 5 benchmarks and real-world robotic deployments, demonstrate that SLOPE consistently outperforms leading baselines in fully sparse, semi-sparse, and dense rewards. The project code is available at https://github.com/Arya87/SLOPE.
Radiologists interpret computed tomography (CT) scans by jointly determining what a finding represents and where it lies in three-dimensional (3D) anatomy. However, CT models have treated CT interpretation and 3D segmentation as separate tasks, leaving textual findings spatially unverifiable and overlooking potential synergy between clinical understanding and spatial perception. Here we present HoloCT, a generalist 3D foundation model unifying CT interpretation and voxel-level segmentation by jointly modelling whole-volume context and fine-grained anatomical detail, providing 3D masks as evidence. We curated HoloDATA, comprising 195,921 CT scans from 73 public datasets and six clinical cohorts, with 6.2 million instruction dialogues and 2.8 million CT-mask pairs. Across 346 tasks spanning segmentation, lesion screening, visual question answering, and report generation on 61 internal and 12 external datasets, HoloCT outperformed specialist and foundation models. In multicentre validation across 11,455 clinical CT scans from five hospitals, HoloCT outperformed competing models in lesion segmentation and report generation by 18.2 and 15.3 points. It also achieved 85.1% mean accuracy in screening thyroid tumours and colon lesions, exceeding all four participating radiologists in a retrospective reader study. Further analyses revealed bidirectional semantic-spatial synergy: language-guided segmentation delineated unseen anatomical structures and lesions as well as diffuse free-text findings, outperforming the leading approach by 22.1 points, whereas prompting the model to generate corresponding 3D masks alongside its textual outputs significantly improved interpretation performance. HoloCT shows that clinical understanding and spatial perception can be learned together within a single foundation model and offers a path towards more accurate and reliable CT analysis.
Let $d_1,\dots,d_m$ be directions in $\mathbb{R}^k$, where $k\ge 2$. A multiset $X$ of $n$ points of $\mathbb{R}^k$ projects to the $m$ multisets of numbers $\{\{x\cdot d_j : x\in X\}\}$, or equivalently to the $m$ sorted lists of these numbers. How many directions are needed for these projections to determine every multiset of $n$ points? Two simple bounds are known in every dimension: more than $(k-1)\log_2 n$ directions are always needed, and $(k-1)n+1$ directions in general position always suffice. In the plane much more is known: Matousek, Privetivy and Skovron showed that almost every choice of $O(\log n \log\log n)$ directions suffices. Whether anything similar holds in higher dimensions was left open by Dym, Wellershoff, Tsoukanis, Levy and Balan. We observe that two lemmas of Hair and Sahai, proved for list recovery of random linear codes, answer that question, and by tightening their argument we show that in every dimension almost every choice of $(4k-2)\ln(2n)\ln(k\log_2 2n)+k+1$ directions suffices. For $k\le 10$ this beats the linear bound once $n$ is larger than about 100: for example, 144 directions suffice for multisets of 1000 points in the plane, and an exact computation lowers this to 133. Letting some of the random blocks in their argument fail removes a factor $\log k$: almost every choice of $O(k\log n\log\log n)$ directions suffices, uniformly in $k$ and $n$. We also show that $2k-1$ directions never determine three points, so that the logarithmic lower bound is not tight; that every $m$ directions in the plane fail on some pair of multisets of size $O(1.78^m)$, improving the bound $O(1.80^m)$ of Matousek, Privetivy and Skovron; and that for each $k$, $m$ and $n$, either almost every choice of $m$ directions works or almost none does.
We consider a fundamental problem of \emph{mixedness testing}: Given $n$ copies of an $N$-qubit state $ρ$, determine whether $ρ= \mathbb{I}_d/d$ or $\|ρ-\mathbb{I}_d/d\|_1 \geq \varepsilon$ with high probability, where $d = 2^N$. In particular, we focus on performing this task with simple single-qubit measurements, i.e. measurements that are prepared independently on each qubit. We provide a nearly complete picture of single-qubit mixedness testing by showing that its copy-complexity is $\tildeΘ\left(\sqrt{10}^N/\varepsilon^2\right)$. To establish our lower bound, we introduce a new measurement-dependent lower bound framework for adaptive single-copy state certification. For the upper bound, we present a randomized Pauli basis measurement protocol, which relies on a new primitive for computationally efficient uniformity testing of correlation-concentrated distributions on the Boolean hypercube. In conjunction, we provide lower and upper bounds for mixedness testing with fixed Pauli measurement protocols.
David Gabriel Harispe, Matías Gerard, Pablo A. Kler
Microfluidic droplet generation is critical for many applications, yet efficient device design in high-dimensional parameter spaces remains computationally prohibitive due to complex multiphase flow dynamics. This work presents a unified methodology that couples adaptive surrogate enhancement with simulation-driven inverse design. High-fidelity Volume-of-Fluid simulations provide ground-truth data, while neural network ensembles predict flow regimes and droplet diameters. Ensemble-based uncertainty quantification identifies epistemic uncertainty using consensus levels for regime classification and the coefficient of variation for diameter prediction, revealing that uncertainty concentrates near regime transition boundaries where physics is highly sensitive to parameter perturbations. A physics-informed enrichment strategy targets these critical zones and addresses class imbalances by deploying additional targeted simulations and retraining the surrogates forming the ensembles. A framework that integrates Particle Swarm Optimization for constrained inverse design, balancing accuracy with predictive confidence and physical consistency is therefore proposed. Validation on independent test datasets demonstrates that the enriched ensembles achieve higher accuracy, with the fraction of predictions within the 0-3% error range on droplet diameter nearly doubling, and high-error cases (above 10%) dropping substantially. The inverse design procedure reliably identifies physically realizable configurations meeting target droplet sizes. This approach significantly reduces computational costs while ensuring reliable predictions, offering a framework with potential applicability to the optimization of a broad range of microfluidic systems.
Dalton Raphael Harmsen, Swier Garst, Thomas van Osch, Zarè Palanciyan, Joaquin Vanschoren
Cross-lingual transfer describes how knowledge in a source language benefits a target language. Measuring it quantitatively requires broad multilingual pre-training, as prior work has done with cross-lingual transfer matrices. We ask whether transfer is predictable from freely available typological features, and whether the prominence of high-resource source languages reflects typology or data quality and quantity. We show that typological databases contain cheap and dense signals about cross-lingual transfer. Our typology-only random forest on a 24-language prior-work transfer matrix scores leave-one-language-out $ρ{=}0.705$ and $R^2{=}0.49$, against $ρ{=}0.62$ for a tuned non-typological control, an increment that a paired bootstrap cannot separate at this scale. When both languages of a held-out pair are unseen, performance degrades to $ρ{\approx}0.41$, so the model interpolates a ground-truth matrix more reliably than it predicts new languages. The signal survives leave-one-script-out and leave-one-family-out protocols, so script and family confounding do not explain the effect. By decomposing the transfer into a typology term and a resource-and-script bias term, we find the best-source ranking sensitive to this bias. Typological features never observe this bias, which makes them a zero-compute screening tool that replaces hundreds of training runs with a model fit. Our code is available \href{https://github.com/dharmsen/typo-x-ling-transfer}{here}.
Transaction execution remains a major performance bottleneck in blockchains. Existing parallel execution engines are still blocked by slow state accesses on each transaction's critical path, creating a hidden loss of parallelism. This paper presents Weaver, a state-decoupled transaction processing paradigm that enables execution flow to proceed in the absence of concrete state values. Weaver represents states as lightweight symbolic references and records subsequent computation as symbolic expressions while concrete values are fetched on the fly. To handle state-dependent control flow, Weaver predicts the most likely branch direction through hierarchical symbolic pattern matching. Weaver further introduces expression-granular scheduling that evaluates each symbolic expression as soon as its state dependencies are resolved, rather than waiting for the transaction's entire read set, fully exploiting fine-grained intra-transaction parallelism. Evaluation on real Ethereum blocks and synthetic workloads spanning four representative dApps shows that Weaver delivers 41%-63% higher throughput than state-of-the-art parallel execution engines and up to 7.5 times the throughput of serial execution.
While teaching untyped $λ$-calculus to undergraduate students, we were wondering why $α$-equivalence is not directly inductively defined. In this paper, we demonstrate that this is indeed feasible. Specifically, we provide a grounded, inductive definition for $α$-equivalence and show that it conforms to the specification provided in the literature. The work presented in this paper is fully formalized in the Rocq Prover.
We present the first worst-case, sublinear protocol to estimate the fidelity between a known target state and unknown lab-prepared state using Pauli basis measurements. More precisely, for $d$-dimensional states, we estimate the fidelity to $\pm \varepsilon$ using $o(d^{0.9675})/\varepsilon^2$ copies, which applies to all target states. Furthermore, our results establish improved guarantees for Pauli shadow tomography and Distributed Inner Product Estimation.
Filip Blašković, Tim O. F. Conrad, Stefan Klus, Nataša Djurdjevac Conrad
Temporal (or time-evolving) networks provide a natural framework for modeling complex systems with time-dependent interactions, where understanding the evolution of community structures is a central challenge. While random walk-based approaches to community detection in static networks are well established through the spectral analysis of associated transfer operators, extending these ideas to temporal networks is nontrivial due to the inherent time-dependence of the underlying dynamics. In this work, we develop a general framework for community detection in temporal networks that is based on multi-view canonical correlation analysis (mCCA). We show that the proposed formulation admits a spectral characterization via a time-reversible random walk on an augmented space-time network, providing a clear dynamical interpretation of temporal communities as metastable structures of the process. Furthermore, we analyze key spectral properties of the resulting transfer operators and the interplay between spatial and temporal effects, which allows us to distinguish between structural features and artifacts induced by the snapshot coupling. Finally, we derive a reduced-order model, which preserves the essential spectral properties while significantly improving computational efficiency. We show that the proposed approach effectively detects communities in temporal networks and captures their evolution.
NVMe Flexible Data Placement (FDP) reduces SSD write amplification by separating data with different lifetimes to minimize garbage collection (GC) copying. However, on shared servers, multiple tenants generate more lifetime classes than the available placement handles. Existing approaches rely on static mappings or application-specific knowledge, failing to capture runtime lifetimes and adapt to changing workloads. The challenge is to identify which data die together and determine which classes should be separated to maximize GC savings under a limited handle budget. We present eFDP, an eBPF-based system that automatically learns data lifetimes and manages FDP placement for unmodified applications on a stock Linux kernel. By tracking writes and invalidations, eFDP identifies lifetime classes and estimates the GC copying saved by separating them. It first assigns each tenant a handle, then allocates the remaining handles to the separations offering the greatest savings. It also re-plans placement when active tenants change. With four database tenants on an emulated FDP SSD, eFDP eliminates twice as much excess write amplification as tenant-level isolation and improves every writing tenant's throughput by 19-23%. On an enterprise FDP SSD, given RocksDB's levels, it matches a manually optimized placement using three handles instead of five. After a tenant switch, eFDP re-plans within seven minutes and maintains 25% lower write amplification than the original layout even after an hour.
Dynamic text-attributed graphs capture evolving interactions among entities with rich textual information, but future-link prediction remains difficult when nodes have little interaction history or appear as new targets. We propose \method, a generative framework that represents nodes with compact semantic identifiers (SIDs) learned from text and autoregressively predicts a target SID conditioned on the source SID, recent interaction events, and relative time gaps. A lightweight structural calibration further incorporates repetition, co-occurrence, and popularity. At inference time, \method encodes each query context once and reuses SID likelihoods to rank prescribed candidates, providing a candidate-efficient prediction interface. Experiments on nine datasets show broadly competitive performance across domains and graph densities. Source-side inductive evaluations demonstrate that text-derived SIDs provide a useful representation when a test source has no observed interaction history, while two recommendation-oriented case studies evaluate controlled new-user adaptation and first-arrival ranking for inductive new items. Further analyses characterize SID uniqueness, capacity, and temporal stability. Overall, the results support combining text-derived SIDs with temporal histories for dynamic link prediction.
Shristy Shah, Thomas B. Moeslund, Kamal Nasrollahi
Analytics on video archives are bottlenecked by decoding. In a dense pipeline that runs a detector on every frame, decoding is a small fraction of the runtime, but once the detector is restricted to selected frames, decoding takes over as the main cost. We introduce CoDecide, a retrospective analytics system that only decodes the frames where new content appears, using signals from the compressed bitstream to find them. Motion tells us something is moving, but not whether it is new to the scene or has been there the whole time. To make that distinction, we supervise a lightweight compressed-domain student on a label-free novelty signal from a frozen DINOv2 teacher, which fires on semantically new regions rather than on any motion. Consequently, the student model generalizes to new camera deployments without requiring per-camera retraining. At inference, it decides which frames to decode and which regions to reconstruct. Detections from the decoded frames are propagated to the skipped frames using motion vectors. We evaluate CoDecide with a leave-one-dataset-out protocol on VIRAT, MEVA, and ShanghaiTech. Against a dense per-frame detector, CoDecide recovers more moving-object detections than prior compressed-domain selectors at an equivalent decoding budget, outperforming even a competitor trained directly on the target camera. CoDecide runs 15 to 24x faster than a pipeline that decodes and detects every frame. Code is available at https://github.com/shristy51/CoDecide.
Time-series forecasting is foundational to a wide spectrum of web-scale services, spanning personalized recommendations, energy scheduling, and infrastructure capacity planning. Accurate long-term time series forecasting (LTSF) requires models to capture long-range temporal dependencies while adapting to periodic patterns that vary across variables. Although frequency-aware methods offer a global view of temporal dynamics, they typically rely on inflexible heuristic designs, such as predefined period lengths, manually selected or randomly sampled top-k frequency components, and separate spectral transformations. Consequently, they fail to account for time-frequency heterogeneity, where variables that appear synchronized in the time domain can differ substantially in the frequency domain. Meanwhile, Mamba-based forecasters, despite their efficiency in long-range modeling, introduce frequency information only as auxiliary signals, leaving spectral evidence disconnected from state transitions. In this work, we propose AdaMamba, a novel approach that explicitly internalizes input-dependent frequency information into the modeling of temporal dependencies. AdaMamba comprises an interaction learning module and an adaptive frequency-gated state space module: the former captures cross-variable dependencies and multi-scale temporal context, while the latter generates context-dependent frequency bases and unifies temporal forgetting into a single time-frequency gate that directly modulates state evolution. Extensive experiments on seven public LTSF benchmarks and four domain-specific datasets demonstrate that AdaMamba consistently outperforms state-of-the-art baselines while maintaining competitive computational efficiency.
Qi Huang, Furong Ye, Ananta Shahane, Thomas Bäck, Niki van Stein
Large Language Models (LLMs) have already been widely adopted for automated algorithm design, demonstrating strong abilities in generating and evolving algorithms across various fields. Existing work has largely focused on examining their effectiveness in solving specific problems, with search strategies primarily guided by adaptive prompt designs. In this paper, through investigating the token-wise attribution of the prompts to LLM-generated algorithmic code, we show that the embedded code examples and their associated strategy instructions have the strongest influence on the generated algorithms. Building upon this insight, we propose leveraging existing benchmark algorithms as strong priors to guide LLM-driven optimization. We evaluate this approach with three LLMs on two black-box optimization benchmarks: the pseudo-Boolean optimization suite (pbo) and the black-box optimization suite (bbob). Despite relying on a simple (1+1) elitist search, it obtains the best average rank in five of six benchmark--LLM settings against six state-of-the-art LLM-driven methods, including more sophisticated search strategies. Our findings highlight the value of integrating benchmarking studies to enhance both efficiency and robustness of the LLM-driven black-box optimization methods.
Closed-loop driving benchmarks in CARLA rank agents by a driving score that multiplies route completion by an infraction penalty. This score does not distinguish an agent that stops in front of an obstacle from one that pushes past it. In a corridor closed by two parked cars, for example, one agent that drives through with contact averages 52.2 on the blocked widths, while the agents that stop average between 37.8 and 40.6. We propose failure-boundary profiles, an evaluation in which each scenario family varies one measured factor, such as corridor width or the time gap in oncoming traffic, and each variant is certified from measurement as one where the agent should proceed or one where it should stop. Along this axis we estimate two curves per agent, commitment (the probability of attempting the passage) and competence (the probability of passing without contact), and locate where each crosses one half, with intervals that account for the simulator's run-to-run variance. The certified labels also give a summary metric, decision accuracy, which credits clean passes on passable variants and stops on blocked ones. We evaluate seven driving agents on seven CARLA families. The rule-based expert reaches a decision accuracy of 0.74 and TFv6, the highest-scoring learned agent, 0.60. Agents with similar driving scores fail in different ways, and agents that reliably stop at a blocked corridor still rarely wait for a gap in oncoming traffic. We also apply the method, unchanged, to a warehouse robot in Isaac Sim. The navigation policy we test stops at blocked gaps in only one of three families, and two fine-tuned versions of it pass more gaps cleanly but no longer stop at all.
Background/Objectives: CNNs perform well in retinal disease classification from spectral-domain optical coherence tomography (SD-OCT), but whether EfficientNet's natural-image scaling hierarchy transfers to medical-image transfer learning at fixed resolution remains unclear. This study evaluates EfficientNet-B0-B7 for four-class SD-OCT classification and accuracy-efficiency trade-offs. Methods: Eight ImageNet-pretrained variants (4.05-64.10M instantiated backbone parameters) were evaluated at 224x224 on the Kermany dataset. Grayscale images were replicated to RGB and processed using model-embedded preprocessing and augmentation. Training comprised frozen-backbone head training followed by partial fine-tuning. Metrics covered classification, discrimination, efficiency, and visual analyses. Results: Test accuracy varied non-monotonically from 95.93% (B0) to 96.95% (B1); B7 achieved 96.73%. B1 had the highest accuracy and Drusen F1 (0.8780), while B7 had Drusen F1 of 0.8720 and the highest Drusen recall (0.9177). Drusen had the lowest F1 across all variants and contributed 82.2% of the summed squared deviation in class-specific F1; class ranking remained Normal > CNV > DME > Drusen. The OCT accuracy ranking differed from the reported ImageNet hierarchy, where B7 ranked first. The accuracy-to-parameter ratio fell from 19.82 to 1.48 percentage points per million total parameters. The final train-test gap had a weak, non-significant association with test accuracy (r = 0.26, p = 0.53). Conclusions: Larger EfficientNet variants did not consistently improve SD-OCT classification. B1 achieved the highest accuracy and Drusen F1, whereas B7 showed a secondary accuracy recovery at greater model size and CPU latency, indicating task and protocol dependent scaling effects. As cross-partition duplicates have not been independently audited, results are conditional on the supplied split.
Image characterization relates coding budgets to the probability carried by source cells and their output images. At second order, refining a cell can reveal unavailable information, while a change of reference measure can alter the relevant likelihood fluctuations. We establish image bounds that account for both effects. A finite-block partition theorem bounds the cost of merging component images by the joint conditional information spectrum of their label. It preserves continuous second-order profiles when that cost is negligible on the square-root blocklength scale. For reconstruction constraints, cell-dependent reference measures retain the actual distortion deficits in a single likelihood. A Gaussian evaluation gives matching second-order rates for Heegard--Berger coding with one uninformed decoder in the strict both-active regime, common reconstruction under a joint agreement and distortion criterion, and two-stage refinement with side information absent at the first decoder. A structured binary source illustrates partition stability even when its global density and conditional-divergence deficits are linear in blocklength. Posterior-mass images also sharpen extraction after a public refinement by charging its residual information rather than its alphabet size. These results preserve the joint constraints and the observations available for choosing the local coding experiment.
We establish upper bounds on the Shannon entropy $\mathbb H(\widehat f^{\,2})$ of the Fourier distribution of Boolean functions. We prove that $$ \mathbb H(\widehat f^{\,2}) =O\!\left(\log\bigl(1+\|\widehat f\|_{1,1/3}\bigr)\right), $$ where $\|\widehat f\|_{1,1/3}$ denotes the minimum spectral norm of a pointwise $1/3$-approximant to $f$. This implies $\mathbb H(\widehat f^{\,2})=O(\mathsf R^\oplus(f))$, where $\mathsf R^\oplus(f)$ is bounded-error randomized parity-query complexity. We also bound Fourier entropy by the smaller of the two one-sided average unambiguous parity-certificate complexities. More generally, for any exact representation of indicator of $f$ as $\sum_C w_C\mathbf1_C$, with nonnegative weights and affine subspaces $C$, we prove $$ \mathbb H(\widehat f^{\,2}) =O\!\left(\sum_C w_C\,2^{-k_C}k_C\right), \qquad k_C=\operatorname{codim}(C). $$ The quantity on the right is the average conical junta degree of the representation. These results strengthen earlier bounds of (Arunachalam et al. 2021) by extending approximation-based entropy estimates to Shannon entropy and allowing one-sided and overlapping exact representations. The proofs use an entropy bound from a sequence of increasingly accurate approximants of $f$ with \emph{small} spectral norm. Our techniques also find an application in quantum communication complexity, where it improves a lower bound for the bounded-error quantum communication complexity of XOR functions.
Classical Timed Games formulations may be unsuitable, or require substantial modeling effort, to capture complex interaction patterns between a controller and its surrounding environment, in which non-determinism must be resolved after the controller has chosen which action to perform. In this paper, we develop the theory of Timed CLTLoc Games (TCGs), a Timed Game variant designed to facilitate the modeling of such patterns. Unlike classical Timed Games formulations, TCGs partition locations rather than transitions, and use Constraint Linear Temporal Logic over clocks (CLTLoc) to specify the controller's objectives. We develop algorithms for solving TCGs and synthesizing winning controller strategies for games whose objectives include formal properties such as reachability and safety. For reachability objectives, our synthesis algorithm yields winning controller strategies that minimize the number of moves required to reach a goal state. Solving reachability TCGs is EXPTIME-complete, matching the complexity of solving reachability Timed Games. All of our algorithms operate backwards and are implemented in our C++ region-based library TARZAN, leveraging OpenMP for efficient parallelization. Experimental results demonstrate the feasibility of TARZAN.
Boyu Niu, Yingjie Xiong, Shurui Li, Linhan Pang, Yuntong Zhang, Han Wang, Hao Wu, Kai Zuo 等 11 位作者
Traditional document parsing systems use Markdown as their output representation, which is effective for linear text but does not fully preserve fine-grained spatial structure, table geometry, visual hierarchy, or rendering semantics. We show that HTML is advantageous for heterogeneous and layout-sensitive documents whose representations must remain editable, inspectable, and reusable, and introduce AmbiHTMLBench to systematically study this setting. In this benchmark, given a real-world document, a model must recover executable HTML that preserves its content, organization, and visual appearance. The benchmark contains 1,115 diverse pages with structured annotations and evaluates predictions in two different ways: The objective protocol aligns ground-truth and predicted rendered blocks and measures table fidelity, page-level text error, and a two-dimensional layout consistency; a complementary reference-anchored protocol measures perceived fidelity by aligning LLM judges with human preferences. Across ten primary multimodal systems, the best objective score is 0.8092, while the subjective winner is different, demonstrating that automatic decomposition and human-perceived fidelity capture complementary properties. We further fine-tune a small-parameter expert model and find that task-specific adaptation yields good performance on document-to-HTML reconstruction task. These results position AmbiHTMLBench as a diagnostic testbed for building smarter and more logical document parsing systems.
Vision-Language-Action Models (VLAs) enable robots to perform manipulation tasks by leveraging Vision-Language Models (VLMs) to process visual observations and language instructions. However, the action prediction objective itself lacks fine-grained task-specific perceptual supervision, making it prone to relying on task-irrelevant visual shortcuts, leading attention away from target objects and key interaction areas, thereby limiting its operational performance and generalization ability. To address this, we propose Automatic Perception-Guided Attention Alignment for Robot Manipulation ($A^3$-VLA), a framework that derives task-specific visual guidance from the manipulation process itself. Specifically, $A^3$-VLA first generates heuristic candidate regions for target objects and the robot from interaction-induced visual changes, and refines them with a visual segmentation model offline. These regions serve as explicit perceptual guidance through two complementary objectives: an attention allocation loss that encourages sufficient focus on task-critical regions when language or robot state tokens query the visual input, and a contrastive learning loss that aligns the representations of candidate target regions with language instructions while separating them from task-irrelevant regions. Extensive experiments on simulated and real-world manipulation tasks demonstrate that $A^3$-VLA consistently improves multiple VLA backbones, with an average success rate gain of 6.5% - 25.2% across benchmarks.
Long-horizon language-model agents accumulate increasingly large interaction histories, making context compaction essential for efficient execution. Existing learned compaction methods often rely on downstream task outcomes or solver behavior, coupling training to specific tasks or agents. We introduce SumOne, a standalone summarizer trained to preserve information in agent histories through masked reconstruction. Given a summary of the expired history, a frozen reconstructor recovers randomly masked portions of the original trajectory; reconstruction quality, combined with a length penalty, provides a task-agnostic reward for optimizing the summarizer with GRPO. This yields self-supervision directly from raw trajectories, without reference summaries or downstream task execution. At inference time, only the lightweight summarizer is retained, enabling plug-and-play context compression for otherwise unchanged agents. We conduct a systematic evaluation of SumOne in long-horizon search and software-engineering tasks. SumOne reduces estimated inference cost by approximately 78\% on DeepSearchQA and 75\% on SWE-bench Verified relative to operating with full context, with absolute decreases of 0.04. The same checkpoint also transfers across downstream models and agent harnesses without additional fine-tuning.
Skills let language agents reuse experience across tasks as natural-language documents. Recent work evolves skills from interaction, but existing methods have limitations when skills evolve online. Prior methods update a skill offline from a batch of trajectories collected over many tasks. Some then adopt the new skill without any check, which can degrade the skill over time. Others keep it only if it improves performance on a held-out validation set, which is too slow and costly to run after every update. We introduce EvoGuard, a pairwise skill selector that mitigates skill degradation at low latency and cost. We first run the agent iteratively to build pools of skills and score each skill on a private test set to obtain utility labels. We pair the skills and use rejection sampling to keep only pairs with a clear utility gap. The selector is trained on the logits of two label tokens to pick the better skill. At deployment it decides in a single forward pass, which is significantly faster than evaluation on the private test set. Compared with adopting every new skill, EvoGuard improves success by 11.0% on ALFWorld, 5.9% on ScienceWorld, and 2.7% on LiveCodeBench.
Francesco Ferrini, Veronica Lachi, Antonio Longa, Bruno Lepri, Andrea Passerini
Graph Neural Networks (GNNs) often struggle to capture the link-specific structural patterns crucial for accurate link prediction, as their node-centric message-passing schemes overlook the subgraph structures connecting a pair of nodes. Existing methods to inject such structural context either incur high computational cost or rely on simplistic heuristics (e.g., common neighbor counts) that fail to model multi-hop dependencies. We introduce SP4LP (Shortest Path for Link Prediction), a novel framework that combines GNN-based node encodings with sequence modeling over shortest paths. Specifically, SP4LP first applies a GNN to compute representations for all nodes, then extracts the shortest path between each candidate node pair and processes the resulting sequence of node embeddings using a sequence model. This design enables SP4LP to capture expressive multi-hop relational patterns with computational efficiency. Empirically, SP4LP achieves state-of-the-art performance across link prediction benchmarks. Theoretically, we prove that SP4LP is strictly more expressive than standard message-passing GNNs and several state-of-the-art structural features methods, establishing it as a general and principled approach for link prediction in graphs.
Ning Yang, Chengzhi Wang, Yibo Liu, Baoliang Tian, Haijun Zhang
Prefill-only KV compression freezes a token subset at the end of prefill and decodes from it without further eviction. The retention decision is therefore irreversible, yet existing methods estimate the corrective signals it relies on, per-head reliability and prompt-level compression sensitivity, online from a single noisy prompt. We argue this is the wrong statistical unit: these signals exhibit far higher cross-prompt regularity than within-prompt signal-to-noise. We introduce \textsc{CompilerKV}, a KV-retention policy whose corrective tables are compiled offline from a calibration corpus, reducing online correction after the standard observation-window scan to $O(1)$ lookups plus a budget clamp. We find that compiled retention tables behave as portable architectural priors: rankings transfer across disjoint corpora on four backbones (mean Spearman $\barρ{=}0.90$), and direct model-to-model table transfer costs only $0.4$--$0.8$ LongBench points on average. At a 512-token budget, \textsc{CompilerKV} attains compressed-SOTA on all four backbones, improving over the strongest prefill-only baseline by $+1.67$ points on average (task-bootstrap 95\% CI $[+1.08,+2.37]$). Pressure regimes amplify the gap: under a fixed $512/32k$ cache ratio, \textsc{CompilerKV} remains the strongest compressed method through 128k RULER ($\sim\!73$ vs.\ FullKV $\sim\!79$, SnapKV $\sim\!38$); on 32k NIAH it reaches $0.89$ vs.\ SnapKV $0.42$; and at 32k input, retaining only $1.56\%$ of the prefill KV, batch-16 serving remains feasible where FullKV is OOM.
Shu'ang Sun, Yibo Qiu, Haoliang Ye, Zan Huang, Ao Tian, Kan Ding, Ronald Xu, Mingzhai Sun
Online reinforcement learning offers an effective approach to precise laboratory manipulation, yet costly real-world demonstrations and limited access to dense process rewards hinder its practical application. We present TwinBridge, a digital twin-driven framework that supplies policy initialization and dense visual rewards for real-world online learning. We construct digital twins by combining 3D Gaussian splatting with physics simulation. Here, simulator states provide task events and geometric progress under physical prerequisites and ordering constraints, yielding rewards for policy pretraining with simulated demonstrations and reinforcement learning. We distill this supervision into a visual reward model. During real-world interaction, its predictions are combined with execution history to compute dense rewards for human-in-the-loop policy refinement. Across five real-world laboratory tasks, TwinBridge achieves a final mean autonomous success rate of 98% with at most 25 minutes of real-world interaction per task. Experiments show that dense rewards improve interaction efficiency over sparse rewards, and simulation-based policy initialization further accelerates learning.
Multivariate time series anomaly detection (MTSAD) relies on modeling both temporal dynamics and inter-sensor dependencies to identify complex system faults. Although graph-based detectors explicitly model sensor topologies, they train task-specific networks from scratch on limited target-domain data, remaining vulnerable to spurious correlations and unable to reuse transferable priors. While pretrained Graph Foundation Models (GFMs) offer transferable topological priors and anomaly discriminability, directly adapting them to MTSAD is hindered by cross-modal feature-topology mismatch, unreliable pseudo-anomaly supervision, and decision distortion during node-to-window aggregation. To fill these gaps, we propose a novel adaptation framework, Gfm4Mtsad, which unleashes pretrained GFMs for MTSAD. We introduce the Structure-and-Information Contract Alignment (SICA) module to bridge continuous time series and the pretrained GFM space. This module harmonizes principal subspace token projection, normal-prior-guided graph construction, and residual recovery to suppress spurious edges while retaining fine-grained temporal details. We propose a Mechanism-Aligned Pseudo-Anomaly Screening strategy (MAPS) to steer unsupervised learning. It applies localized dependency breaks while preserving empirical residuals, selecting hard negatives near the normal boundary for targeted supervision. We develop a Tail-Gated Evidence Calibration mechanism (TGEC) to eliminate node-to-window aggregation distortion. It calibrates multi-granular graph responses against normal operational references, strictly preserving benign score rankings while selectively refining extreme anomaly candidates in the tail distribution for reliable window detection and interpretable sensor evidence. Extensive evaluations with two GFM backbones demonstrate that Gfm4Mtsad achieves superior detection performance over existing baselines.
Chain-of-Thought (CoT) reasoning substantially improves Large Language Models (LLMs), yet its out-of-distribution (OOD) generalization mechanism remains theoretically underexplored. We study CoT as compositional OOD generalization: the target task lacks complete reasoning trajectories or direct input-output pairs, while training provides only in-distribution (ID) atomic step data. We show that CoT generalizes by repeatedly applying a shared single-step predictor to compose reusable ID transitions. Theoretically, let $T$ denote the number of reasoning steps in the target task, $n$ the number of training samples per sub-task, and $k$ the number of output tokens per step; then the OOD task risk is controlled by the sum of the ID sub-task risks, scaled by a coefficient of order $\mathcal{O}(k)$ that is independent of $T$, although the summed risk itself may still grow with $T$. A Rademacher-complexity analysis further shows that the estimation term is of order $\mathcal{O}(\sqrt{T/n})$, rather than linear in $T$, due to predictor sharing under a uniform capacity control on the shared class. For autoregressive Next-Token Prediction, the final-task risk is bounded by the best-in-class step-level risk scaled by the same coefficient of order $\mathcal{O}(k)$, plus an estimation term of order $\widetilde{\mathcal{O}}\big(k\sqrt{T/n}\big)$. A parity-coupled construction further shows that, for the sub-task empirical risk minimizer, the expected final-task risk is at least the best in-class risk plus $Ω(\sqrt{T/n})$ when $T\le n$. Experiments on Atom-Task and GSM8K support the theory: CoT enables short-to-long and compositional generalization, while explicit state tracking is crucial. Our code is available at https://github.com/maxliang114514/CoT-OOD.
Aleksandros Sobczyk, Gioele Gottardo, Christos K. Matzoros, Mirko De Vita, Filip Skogh, Anastasios Zouzias, Jiawei Zhuang
Linear attention has emerged as a cornerstone for efficient long-context architectures, as evidenced by its integration into state-of-the-art open-source models including Qwen3.5/3.6, Kimi Linear/K3, and RWKV-7. Models that incorporate linear attention layers with the so-called Delta-Rule involve the inversion of triangular matrices as a core sub-routine. This operation often forms a performance bottleneck, and, due to its high-sensitivity to numerical errors, it can significantly deteriorate end-to-end model accuracy if it is not carefully implemented. This work provides a systematic analysis of both direct and iterative triangular inversion algorithms, targeting methods that are rich in matrix products, and, therefore, have the potential to efficiently utilize modern hardware. To that end, our analysis covers a broad spectrum of mathematical and practical aspects, with a heavy focus on numerical stability, computational complexity, and, ultimately, hardware efficiency and practical considerations. We provide a rigorous experimental evaluation to verify these properties in practical scenarios, and in low-precision floating-point representations, highlighting the strengths and limitations of each method. Performance benchmarks on NPUs reveal up to $4.3\times$ speed-up against the state-of-the-art implementations of SGLang for triangular matrix inversion, leading to significant performance improvements on the entire layer level, while maintaining end-to-end model accuracy.