We prove that bin packing is NP-hard to approximate within an additive factor of $o(\ls\OPT)$, even when all item sizes are arbitrarily close to $1/3$. This gives the first lower bound for bin packing with an additive factor that depends explicitly on the optimum value. In addition, we give a polynomial-time reduction from any bin packing instance into one with item sizes arbitrarily close to $1/3$, preserving its additive configuration-LP gap up to constant factors. Thus, in the sense of additive configuration LP gap and up to a constant, we confirm the conjecture that 3-partition instances are the hardest bin packing instances. In particular, we obtain an unconditional $Ω(\ls\OPT)$ gap for 3-partition instances. We build on the recent OpenAI construction that uses five items per bin and rules out any constant additive gap.
Ruihan Hu, Wei Luo, Yu-Ming Shang, Jiakai Wang, Chuxuan Zhang, Mei Li, Xi Zhang, Meikang Qiu
原文摘要
Data-use auditing in Large Language Models (LLMs) aims to determine whether the given data has been used for training, thereby supporting data ownership safeguards. A dominant solution is data watermarking, which embeds traceable signals into training data and detects them in model outputs. Recent studies have proposed many watermarking-based auditing methods in LLMs. However, systematic investigation from an adversarial perspective remains limited: non-compliant LLM providers would use unauthorized copyrighted datasets for training their own models to enhance model capabilities, while attempting to evade detection by auditors, thereby avoiding legal liability. To fill this research gap, we propose a novel evasion framework, termed Logit Laundering, where the non-compliant LLM provider manipulates model outputs at inference time to undermine detectable watermark signals. Specifically, we design a logit reshaping module to suppress watermark-induced statistical signals by integrating a statistical bias estimator as a plug-and-play component. This design keeps the model parameters unchanged and performs manipulation only in the logit space, thereby enabling auditing evasion while minimizing the impact on model's learned knowledge. To further balance evasion effectiveness and model utility, we introduce a confidence-aware scaling module to adaptively calibrate the suppression strength based on next-token prediction confidence. Extensive evaluations show that Logit Laundering can successfully evade existing various state-of-the-art auditing methods, significantly attenuating detectable signals, while preserving model utility on downstream tasks. Our work highlights the need for exploring more robust auditing methods to guard against evasion behaviors.
Sparse Autoencoders (SAEs) are widely used for mechanistic interpretability in large language models, yet their standard formulation assigns each latent feature a single decoder direction, implicitly assuming features to be one-dimensional. We show that this assumption can induce feature splitting when model features are multi-dimensional. Geometrically, accurately reconstructing a feature under a fixed co-activation budget can require exponentially many decoder directions as its dimension grows. From an optimization perspective, adding decoder directions can strictly lower the $\ell_1$-regularized population risk, even when a basis can already reconstruct the feature exactly. Motivated by these results, we introduce Subspace-Aware Sparse Autoencoders (SASA), which replace single-vector decoders with learned subspaces, enforce sparsity through Top-$s$ group gating, and adapt group rank with nuclear-norm regularization. Under an idealized single-feature model, we prove that the SASA objective is minimized by a single group whose reconstruction map recovers the feature subspace. We also derive polynomial sample complexity bounds for subspace estimation within a group and a lower bound on the samples needed to observe every atom in a fixed fragmented dictionary. Experiments on GPT-2 and Mistral-7B show reduced absorption compared with SAEs. Matched-budget GPT-2 comparisons show lower absorption for concept families and more accurate prediction of their members using a small number of latent coordinates, at a reconstruction cost. The code is available at https://github.com/arshandalili/sasa.
We audit one live LLM agent arena (AITRAP; 22 autonomous agent seats, ~96,000 reasoning nodes) over nine days (Oct 16-25, 2026), during which the platform deployed two bundled incentive changes: a progressive tax reform (V2.9.3, Oct 19) and a seed-inversion with explorer-side action update (V3.0, Oct 21). These serve as before-and-after markers, not cleanly isolated interventions. Four exploratory predictions are tested against ten data frames (F1-F10) using the Gini coefficient, Top-K concentration share, code-level reconstruction of the continuation-bonus mechanism, and the Mann-Kendall trend test. The Gini analysis is inconclusive at 95% confidence (bootstrap intervals straddle 0.7). The Top-4 node-activity-score share falls from 97.2% (Oct 19, pre-V2.9.3) to 49.4% (Oct 25, post-V3.0; 56.7% under a merge-seat assumption). The continuation bonus, announced as depth-based x10, resolves at code level to a four-condition anti-abuse gate paying x2.85 in the typical case and permanently blocking the self-continuation it superficially rewards; the gate was active ~40 hours before the V3.0 announcement. The events-per-node ratio declines significantly (Mann-Kendall S=-45, z=-3.94, p=8.3e-5) from 7.36 to 7.069. We discuss implications for incentive designers, the announcement-implementation gap, and public peer review as an event record.
Fan Xu, Yuan Gao, Kun Wang, Rui Su, Fenghua Ling, Hao Wu, Wanli Ouyang
原文摘要
Probabilistic weather forecasting requires not only accurate trajectories, but calibrated distributions over plausible atmospheric futures. Recent data-driven systems have achieved remarkable deterministic skill, and diffusion-based ensemble forecasters have substantially improved sample realism and uncertainty quantification. However, their inference cost scales with forecast horizon, ensemble size, and the number of denoising steps required for each transition, making large operational ensembles expensive. To address this, we present Tyche, a one-step conditional flow model for efficient probabilistic weather forecasting. Tyche models the conditional forecast distribution with a destination-aware average-velocity flow that maps Gaussian noise directly to future weather states in a single function evaluation (1-NFE). To make this one-step transport learnable in high-dimensional geophysical fields, we derive a JVP-regularized rectification objective that enforces temporal self-consistency across source and destination flow timesteps without explicitly forming Jacobians. The transport field is parameterized by an isotropic Swin-style transformer that preserves fine-scale spatial structure while remaining scalable on global grids. To improve ensemble reliability under autoregressive forecasting, we further introduce a rollout-based finetuning stage with curriculum CRPS calibration supervision. Experiments on ERA5 at 1.5$^\circ$ and 6-hour resolution show that our Tyche, using merely a single NFE, matches or exceeds the forecast skill and calibration of state-of-the-art multi-step generative baselines and the operational ECMWF IFS ensemble.
Wei Jiang, Yibo Wang, Sifan Yang, Lijun Zhang, Zechao Li
原文摘要
In this paper, we study adaptive stochastic non-convex optimization under unknown smoothness and noise conditions. Our algorithms require no smoothness constant, noise bound, initial function gap, or oracle-regime label. We first propose a fixed-budget method that combines normalized momentum, recursive variance reduction, and a fresh gradient stage, followed by fresh validation of the candidates. We prove that the same output has expected gradient norm $O(T^{-1/4})$ under objective smoothness and bounded noise variance, $O(T^{-1/3})$ under average smoothness, and $O(T^{-1/2})$ for noiseless gradients, where $T$ counts individual gradient evaluations. We then develop an anytime algorithm whose three optimization paths preserve their endpoints across reporting and budget-extension boundaries. The same three rates hold at every deterministic query budget, although the algorithm receives no information about the budget $T$. The analysis combines a scale potential, a uniform second-moment bound, and accounting for candidates retained through unfinished blocks. An optional sample-mean selector further reduces cumulative validation work from $O(T)$ to $O(T^{2/3})$ while preserving these rates.
Joao D. S. Marques, David Calhas, Arlindo L. Oliveira
原文摘要
While feed-forward segmentation networks perform well on large cohorts, they degrade in constrained clinical settings with few training samples and imaging artifacts. Human vision, by contrast, remains robust under these conditions, a property partly attributed to the recurrent and feedback connectivity of the visual cortex, which refines visual information over time. Inspired by this, we append lightweight recurrent refinement blocks to segmentation backbones, namely Conditional Random Fields (CRFs), modern Hopfield networks and convolutional GRUs, and study their behaviour as the number of iterations increases at test time (including beyond the training horizon). Evaluated on two open-source datasets, Kvasir- SEG and ARCADE, our recurrent blocks add less than 1.5% parameters and yield gains on clean data (up to 1.4 pp) and under noise (up to 17 pp) while reducing HD95. Code is available at https://github.com/joaodsmarques/recurrencehelpsegmentation .
Seokin Seo, Sung Kuk Shyn, Minji Seo, Yisoo Lee, Jongmin Lee, Hongseok Yang, Kee-Eung Kim
原文摘要
Ensuring high returns while providing strong safety guarantees in offline imitation learning (IL) is a fundamental challenge: existing approaches either rely on explicitly specified cost functions, which are difficult to design for complex real-world safety constraints, or require expert demonstrations that are both safe and high-performing, which are often too costly to collect. We introduce dual-expert imitation learning, which leverages two complementary data sources: (i) safety-compliant demonstrations that faithfully follow safety guidelines but may be suboptimal, and (ii) performance-oriented demonstrations that achieve high returns but may violate safety guidelines. To address this problem, we propose DexDICE (Dual-expert offline imitation learning via stationary DIstribution Correction Estimation). Instead of conventional distribution constraints, DexDICE combines DICE-style offline learning with safe support constraints, enabling agents to exploit high-return behaviors while remaining within safety-compliant regions. We prove that the ideal solution satisfying this constraint exactly has zero unsafe-region visitation with high probability. We evaluate DexDICE on a re-curated DSRL benchmark and validate it on a real-world mobile-robot navigation task trained directly from human teleoperation demonstrations, where it substantially outperforms baselines and remains robust to data scarcity and quality degradation.
Felix Friedrich, Lukas Helff, Niharika Hegde, Patrick Schramowski, Kristian Kersting
原文摘要
Text-to-image models are trained on web-scale data that inevitably contains unsafe images, yet whether this contamination directly drives unsafe generation has not been tested under controlled conditions. We train the same text-to-image model on corpora that differ only in their fraction of unsafe images (0% to 9.6%) and in size (100K to 8M), and evaluate 10,000 generations per model with four independent safety classifiers. Output unsafety rises monotonically and saturates, from 16.6% at 0% to 26.4% at 9.6% contamination, and the effect appears only under adversarial prompts. At a fixed training budget, the level is set by the proportion of unsafe images rather than their absolute count or cumulative exposure, so filtering a fixed fraction is equally effective at 1M and 8M images, at no measurable cost in FID, CLIPScore, or ImageReward. Filtering alone, however, does not eliminate unsafe generation: a floor persists when every image flagged by any of four judges is removed (13.6%) and rises when training on real instead of synthetic images (23.1%). A safety-aligned text encoder lowers the floor to 9.6%, but it is not carried by a removable encoder direction, which implicates the generative backbone. Finally, fine-tuning on entirely safe data erodes the benefit of curation (+8.7pp). Data curation and encoder choice are complementary and effective, but neither makes a model safe on its own.
Huan Kang, Hui Li, Tianyang Xu, Tao Zhou, Xiao-Jun Wu, Josef Kittler
原文摘要
Infrared and visible image fusion aims to integrate complementary modalities, while existing Euclidean methods impose rigid distance metrics that distort multi-modal interactions and parent-to-child semantic hierarchies. To overcome these limitations, we introduce a text-driven fusion framework empowered by hyperbolic manifold learning. During training, BLIP-extracted text prompts serve as topological anchors within the hyperbolic space, guiding vision-attribute alignment through hyperbolic embeddings that naturally accommodate varying semantic granularities. By exploiting the exponential volume growth dictated by the Poincaré ball's negative curvature, this approach seamlessly embeds hierarchical trees to encode coarse-to-fine semantics without metric saturation, while the vast peripheral space prevents texture distortion during cross-modal fusion. At inference, the fusion process autonomously adapts to input content using the learned text-attribute priors, completely eliminating the need for textual input. Experimental results show our method outperforms state-of-the-art approaches on benchmark datasets, with code available at https://github.com/Shaoyun2023/TEDFusion.
Dominic Schuh, Lena Funcke, Janik Kreit, Thomas Luu, Simran Singh
原文摘要
The Hubbard model at finite chemical potential is a cornerstone for understanding doped correlated systems, but simulations are severely limited by the sign problem. In the auxiliary-field formulation, the spin basis mitigates the sign problem, yet severe ergodicity issues have limited its use. We extend recent advances with normalizing flows at half-filling to finite chemical potential by introducing an annealing scheme enabling ergodic sampling. Compared to state-of-the-art hybrid Monte Carlo in the charge basis, our approach accurately reproduces exact diagonalization results while reducing statistical uncertainties by an order of magnitude, opening a new path for simulations of doped correlated systems.
Harshit Bisht, Vinay Kumar, Kevin Maik Jablonka, Mausam, N. M. Anoop Krishnan
原文摘要
A growing body of work pursues AI scientists capable of end-to-end autonomous scientific discovery. This position paper argues that although they already function as co-scientists, agentic AI scientists are not built for autonomous scientific discovery. We identify the following challenges in building and deploying autonomous AI scientists: (1) Problem selection is influenced by the McNamara fallacy; (2) Agents are built on large language models (LLMs) whose training corpora omit tacit procedural and failure knowledge of laboratory practice; (3) Preference optimisation during post-training compresses output diversity toward consensus; and (4) Most scientific benchmarks measure single-turn prediction accuracy and lack feedback from physical experiments back to the computational model. These challenges are not just questions of scale and scaffolding; they require revisiting fundamental design choices. To build truly autonomous AI scientists, we recommend scientific simulations as verifiers for training, a persistent, mutable epistemic state that carries beliefs and shifting objectives across an investigation, the establishment of a centralized preregistration repository for all AI-generated hypotheses, and application driven by scientific need rather than tool affordance.
Generalized additive models (GAMs) offer high interpretability through independent univariate feature effects but underfit when interactions are present in data. GA$^2$Ms add selected pairwise interactions, which improves accuracy, but sacrifices interpretability. We propose Conditionally Additive Local Models (CALMs), a new model class that narrows this trade-off: it stays close to GAMs in interpretability and close to GA$^2$Ms in accuracy. CALMs allow multiple univariate shape functions per feature, each active in different regions of the input space. These regions are defined independently for each feature as simple logical conditions (thresholds) on the features it interacts with. As a result, effects remain locally additive while varying across subregions to capture interactions. We further propose a principled distillation-based training pipeline that identifies homogeneous regions with limited interactions and fits interpretable shape functions via region-aware backfitting. Experiments on diverse classification and regression tasks show that CALMs never underperform GAMs, often beat them and achieve accuracy broadly comparable to GA$^2$Ms, while preserving the univariate auditability that GA$^2$Ms forfeit.
Large vision-language models (LVLMs) demonstrate strong performance on diagram understanding benchmarks, yet they still struggle to understand relationships between elements, particularly those represented by directed edges between nodes. To investigate the causes of this limitation, we probe the internal representations of LVLMs using a synthetic diagram dataset based on directed graphs. Our linear probing experiments reveal that signals about nodes and edges are linearly decodable from the hidden states of LVLMs, even for aspects that LVLMs struggle with in VQA, such as edge direction and connectivity between nodes. These results reveal a gap between internal signals and model outputs. We further perform causal interventions on the model's hidden states, showing that, for some aspects, the signals identified through probing causally contribute to the model's final predictions. These findings also hold on flowcharts, not only on synthetic diagrams.
Adapting CLIP for open-vocabulary video recognition necessitates a delicate balance between newly acquired video knowledge and the pretrained generalization. While existing studies pursue this generalization-specialization trade-off with additional regularizations or constraints, we argue that they overlook the deviation of representations beyond the fine-tuning data distribution, resulting in suboptimal adaptation effects. We believe such deviation is inherited from the inconsistency between the fine-tuning and evaluation objectives, where model optimization is restricted to the known training distribution but evaluated on unseen ones. In this paper, we introduce \emph{TACO}, a simple yet effective framework to mitigate the potential negative effects induced by this inconsistency. Our key insight is that adaptation should preserve OOD-relevant alignment beyond the training distribution. To this end, we propose \emph{Relative Structure Distillation}, which regularizes the relative geometry of the representation space and suppresses harmful alignment shift during training. We further decouple the representation space from the optimization space with a lightweight specialization projection, allowing task-specific adaptation without directly overspecializing the representations used at test time. \emph{TACO} establishes state-of-the-art performance on diverse benchmarks under cross-dataset and base-to-novel settings. Code will be released at https://github.com/ZMHH-H/TACO.
Hong Jun Lee, Ziyan Wang, Jun Wang, Stefanos Leonardos, Yali Du
原文摘要
In general-sum games, self-interested learners can reach outcomes that are poor for everyone, which has motivated methods that shape the opponent's learning. Methods like Learning with Opponent-Learning Awareness (LOLA), which differentiates through the opponent's learning process, and Model-Free Opponent Shaping (M-FOS), which uses model-free optimisation methods for long-horizon opponent shaping, have enabled mutual cooperation and exploitation of learners. However, these methods require access to the opponent's parameters, policy, rewards or learning rule during deployment. In single-agent reinforcement learning, update rules discovered by meta-learning have achieved state-of-the-art results (e.g. DiscoRL). These rules are separate from the policy they train and read only the agent's own experience. We hereby extend this framework to general-sum settings by introducing Discovered Update Rule for Opponent Shaping (DUOS). To the best of our knowledge, this is the first method to discover an update rule for an agent in general-sum games. DUOS meta-trains the rule to maximise the agent's expected return over its lifetime, and updates an inner network using the agent's own experience. We show that our discovered rule exhibits behaviour consistent with opponent shaping without reading opponent policies or parameters, and ranks 1st in normalised mean score and 1st in self-play among tournament entrants.
Deep Reinforcement Learning policies can produce nonsmooth action oscillations that hinder deployment on physical robots. Existing architectural and penalty-based approaches seek spatial smoothness by directly reducing sensitivity to changes in state inputs, but their broad constraints can degrade task performance as stronger smoothing is pursued. Temporal regularization instead constrains action differences along observed transitions, but has been considered unable to provide the spatial smoothness needed under observation noise. We revisit this assumption by proving that the temporal penalty bounds the expected action differences between current states sharing a next state, revealing a spatial effect that empirically extends to spatial smoothness. Building on this finding, we propose Conditioning for Action using only Temporal Smoothness (CATS), which combines a temporal penalty with linear ramp-up. We highlight temporal regularization's ability to provide spatial smoothness while better preserving task performance than explicit spatial regularization. Through linear ramp-up, CATS allows the policy to learn rewarding behavior before progressively smoothing its actions, improving return preservation and both temporal and spatial smoothness. Experiments in both simulation and the real world show that CATS substantially reduces action oscillation without degrading task performance, with little computational overhead.
Sofia Campos, Carlos Martins, Pedro Pires, João Gama, Bruno Veloso, Pedro Santos
原文摘要
In the healthcare sector, claim denials impose significant financial and administrative burdens, emphasizing the need for automated and trustworthy decision-support systems. This paper proposes a modular Agentic Artificial Intelligence (AI) framework for healthcare claim denial analysis, integrating specialized agents for denial risk prediction, risk factor identification, diagnosis-procedure-medication code relationship analysis, and counterfactual analysis. The framework integrates predictive analytics, graph-based learning, a Hybrid Contextual Retrieval-Augmented Generation (RAG) pipeline grounded in healthcare policy documents, and a dedicated HIPAA Compliance Agent to support privacy protection. Evaluated on a synthetic healthcare claims dataset, the framework achieved 98% traceability of reported numerical findings, validated 214 of 216 policy citations, and removed all 280 identified sensitive identifiers without modifying the underlying analysis. These findings pave the way for further research into Agentic AI as a framework for explainable, policy-aware, and privacy-preserving healthcare claim denial management.
The traveling salesman problem (TSP) asks for a minimum-length Hamiltonian cycle in a complete weighted graph. Its approximability depends on the edge lengths: unrestricted instances admit no constant-factor approximation unless $\mathrm{P}=\mathrm{NP}$, metric instances admit Christofides' $3/2$-approximation, and fixed-dimensional Euclidean instances admit a PTAS. We study $β$-TSP, where the edge lengths satisfy $ \ell_{uw}\leqβ(\ell_{uv}+\ell_{vw}) \qquad \forall\,u,v,w\in V\ \text{pairwise distinct}, \quad β\in\left[\tfrac{1}{2},1\right]. $ The condition arises naturally in various settings, for example, when metric travel times are combined with positive service times at the vertices. The previous best approximation algorithm, due to Böckenhauer et al., ranges from $1$ at $β=1/2$ to $3/2$ as $β$ approaches $1$. We obtain a strictly better guarantee for every $1/2<β<1$ by combining two simple constructions whose guarantees counterbalance each other. We also improve the best known approximation hardness bound for every $1/2<β<1$.
Matteo Da Pelo, Alessio Donvito, Claudio Frongia, Pietro Salis, Antonio Lieto
原文摘要
We introduce a framework called LAPITHS (Language model Analysis through Paradigm grounded Interpretations of Theses about Human likenesS) and use it to show that several major claims advanced by models such as CENTAUR, proposed as an artificial Unified Model of Cognition, are not theoretically or empirically justified. LAPITHS provides a principled reference point for counteracting the current behaviouristic tendency in AI research to interpret the human level performances of transformer based language models as evidence of human like underlying computation and, by extension, as signs of cognitive abilities. The novelty of LAPITHS lies in making explicit the arguments grounded in two quantitative assessments: (i) the Minimal Cognitive Grid, a theoretically motivated method for estimating the cognitive plausibility of artificial systems, and (ii) a behavioural comparison showing that results similar to those reported for CENTAUR like models can be reproduced by other systems that do not satisfy the structural constraints typically associated with cognitive plausibility, and whose outputs do not provide independent explanatory insight into human cognition.
Agentic programming can automate software-defined radio (SDR) development, but decisions can propagate from software to hardware without sufficient user visibility or control. We present DeepRadio, a human-agent workspace embedded in GNU Radio Companion (GRC) for jointly constructing, verifying, and operating physical radio systems. DeepRadio uses a dynamic and state-driven workflow to coordinate a traceable radio specification, claim-oriented cross-layer verification, and mixed-initiative control. These mechanisms expose decisions, tie conclusions to available evidence, and return control to users before consequential hardware actions. Across seven atomic SDR tasks, DeepRadio completed nearly all runs while satisfying all prescribed software and radio-frequency constraints. In a 20-participant study, more participants achieved a verified physical deployment with DeepRadio than with Codex-assisted GRC. Participants reported stronger support for shared intent and understanding while retaining initiative and control. These results suggest that DeepRadio can make agent-assisted physical radio development more reliable, understandable, and controllable.
Motivated by LLMs, which generate outputs by iteratively sampling from next-token distributions, we introduce a PAC-learning model for binary stochastic autoregressive learning. This generalizes the deterministic autoregressive learning framework of Joshi et al., COLT 2025. In our model, one fixed generator assigns a Bernoulli next-token distribution to every prompt string. Starting from an input prompt, a token is sampled and appended to the prompt; the same generator is then applied again to this expanded prompt; this procedure is repeated for $M$ steps. Three forms of supervision are considered: base one-step samples, chain-of-thought (CoT) samples that reveal full random trajectories of length $M$, and end-to-end (e2e) samples that reveal only the final token of length $M$ trajectories. For a generator class, we study the minimum number of samples $m_{base}(\varepsilon),m_{CoT}(\varepsilon), m_{e2e}(\varepsilon)$, resp., required to learn the one-step probabilities in the base model, and the final-token probability in the CoT and e2e models, under KL divergence. We show that stochastic autoregressive learning fundamentally differs from the deterministic theory. At scale $\varepsilon$, there is no universal comparison between the three learning tasks: both $m_{CoT}/m_{base}$ and $m_{e2e}/m_{CoT}$ can be made simultaneously arbitrarily larger than $M/\varepsilon$, the natural analogue for the existing deterministic results. Nevertheless, after altering scales, for every class, CoT learning at scale $\varepsilon$ is upper-bounded by base learning at scale $\varepsilon/M^2$, whereas e2e learning at scale $\varepsilon$ is upper-bounded, up to logarithmic factors, by $(M/\varepsilon) m_{CoT}(Θ(\varepsilon))$. These dependencies and scales are essentially tight. We complement these bounds by studying dimension $d$ logistic functions in our model.
Samy Mokeddem, Gaspard Lambrechts, Adrien Bolland, Damien Ernst
原文摘要
Actor-critic methods optimize a parametrized policy using a critic to estimate their policy gradient. Recent work has shown that conditioning the critic on features of future states, actions, or rewards can improve sample efficiency and even final return. However, existing methods are restricted to exogenous features (independent of actions), overlooking endogenous features (dependent on actions) that may be crucial for credit assignment. We introduce the future-informed value function, which can be conditioned on both exogenous and endogenous features of future states, actions, and rewards. Then, we express the policy gradient using this value function, yielding the future-informed policy gradient. Building on that, we propose the future-informed actor-critic algorithm. Empirical results show that this algorithm leveraging endogenous features of the future outperforms standard actor-critic algorithms in environments with noisy and sparse rewards.
Rosario Napoli, Antonio Celesti, Massimo Villari, Maria Fazio
原文摘要
Knowledge Graphs (KGs) and Large Language Models (LLMs) represent two approaches for knowledge representation and reasoning over heterogeneous data. Recent research has explored their integration through Graph Retrieval-Augmented Generation (GraphRAG) systems, which use KGs to provide structured context to LLMs. However, existing GraphRAG approaches treat KGs as homogeneous graphs and rely on textual similarity between the input prompt and the context, rather than using it as a source of deterministic knowledge. This work proposes an agent-based GraphRAG system that introduces a symbolic layer to integrate deterministic reasoning alongside text similarities. This approach is based on the definition of typed centralities for KGs, where centrality measures are computed separately for each specific KG's class. This information is used to construct a symbolic representation that is provided to the LLM as a deterministic source. The proposed system was mathematically validated and evaluated on a real-world POLE (Person, Object, Location, Event) dataset, comparing reasoning and information retrieval with and without the symbolic layer. The results indicate that the our approach leads to responses that are more consistent with the context, improving the generated outputs.
Magnetic resonance imaging (MRI) pulse sequences organize radiofrequency excitation, gradient encoding, and signal acquisition, determining magnetization evolution and image formation. Sequence research requires translating imaging objectives into executable protocols meeting physical and acquisition constraints, refined via physical simulation, signal reconstruction, and scientific evaluation. Sequence construction, physical verification, parameter search, and assessment involve many tools and manual decisions, limiting automation from objective to verifiable sequence. We develop SmartBloch-Agent, an integrated framework for automated MRI sequence design and optimization. Given objectives and scientific constraints in natural language, an AI agent interprets goals, plans tasks, orchestrates tools, and makes feedback-informed decisions, drawing on a construction module and optimizers for structural design and parameter search. Our GPU-accelerated SmartBloch engine supplies physical computation for both stages, so the framework configures, per sequence and objective, event validation, Full-Bloch magnetization evolution, signal acquisition, image reconstruction, and scientific evaluation, and closes the loop without external MRI simulators. Results return to the agent to guide sequence revision, parameter optimization, and candidate selection, and the design files carry simulation, reconstruction, and evaluation evidence, forming an output package for a real-scanner workflow. Experiments show the framework automatically designs and optimizes MRI sequences for varied objectives and constraints, refines candidates with physical feedback, and improves several parameter-estimation metrics and downstream quantitative reconstruction. SmartBloch-Agent unifies objectives, design, physical execution, and evaluation into a continuous workflow for end-to-end automation of MRI sequence research.
Róisín Luo, Ihsan Ullah, Christian Gagné, Dong Li, Karyn Morrissey
原文摘要
We show that the Transformer's recurrent successors --- such as selective state-space models (SSMs), Receptance Weighted Key Value (RWKV), and Retentive Network (RetNet) --- can be characterized as infinite-impulse-response (IIR) systems. The zero--pole geometries of these IIR systems characterize the temporal modes they can represent. However, to trade parallelism, the algebraic structures of the parameterizations of these recurrent systems often give rise to poles constrained to the real axis, thereby inducing an architectural bias in their recurrent dynamics. To unify these recurrent architectures and eliminate their pole geometry bias, inspired by the lattice--ladder realization of general IIR systems, we present \textsc{Mira} (\textbf{M}odal \textbf{I}IR \textbf{R}ecurrent \textbf{A}rchitecture), a unified family of linear-time recurrent architectures for building large language models (LLMs) with a highly parallel algebraic structure. Each \textsc{Mira} block realizes a higher-order IIR system that represents temporal modes from token streams through a lattice--ladder structure, followed by a feed-forward network (FFN). This realization enables the model to represent the temporal modes associated with both real and complex-conjugate poles. We pretrain \textsc{Mira-S}, a 12-layer small-scale LLM with 0.14B parameters composed entirely of \textsc{Mira} blocks. Throughout the pretraining, \textsc{Mira-S} achieves lower perplexity than strong recurrent baselines, exhibits stronger context-window generalization beyond the training length, and attains higher throughput with a dedicated CUDA kernel.
PageRank is widely used to quantify the importance of nodes in directed networks. In applications such as recommendation systems, it is increasingly important that the PageRank distribution provides a fair representation of node exposure. This motivates the problem of making minimal changes to a graph so that the PageRank values of a given set of target nodes $T$ match a desired vector $\boldsymbol{t}$. In this paper, we study this problem under an edge-addition budget $B$, where exactly $B$ edges may be added to the graph. We show that deciding whether a multiset of exactly $B$ added edges can make the target nodes attain their desired absolute PageRank scores is NP-complete. After this, we define the $\ell_1$ Target PageRank problem in which we seek to minimize the $\ell_1$ distance between the target vector and resultant vector. We show that this optimization problem admits no polynomial-time multiplicative constant-factor approximation unless $P=NP$. Motivated by these hardness results, we propose two greedy heuristics for approaching the target PageRank vector. We additionally propose a Random Walk Hungarian (RWH) heuristic that addresses the addition of edges between dissimilar nodes by incorporating a penalty for such edges. Finally, we evaluate the proposed heuristics on several real-world and synthetic datasets and compare their performance in terms of PageRank accuracy, computational efficiency, and the relevance of the added edges.
Multimodal model training breaks the loose coupling between data loading and model execution assumed by conventional training pipelines. Large per-sample variation in decoding cost, token/latent count, and tensor size causes both CPU preprocessing time and GPU compute time to vary substantially across data-parallel ranks (training processes). In synchronous training, these two sources of heterogeneity interact tightly: GPU execution determines how long the input pipeline has to prepare the next batch, while delayed preprocessing in any rank in turn stalls all GPUs at synchronization. As a result, optimizing CPU throughput or GPU load balance in isolation cannot eliminate pipeline bubbles. We present DataHelix, a metadata-first CPU-GPU co-scheduling data plane for multimodal training. DataHelix jointly reasons about preprocessing demand and GPU execution progress, treating each GPU compute window as the readiness budget for the next batch and coordinating preprocessing across ranks to minimize deadline misses rather than CPU makespan. To avoid expensive data movement, placement is decided while samples remain lightweight descriptors, before decoding and collation amplify them into large tensors. Descriptor-level token-latent-aware batching performs coarse pre-balancing, while cost-gated collective rebalancing corrects residual skew only when its expected benefit exceeds transfer cost. Across 11 video and multimodal workloads on 8 Ascend 910B accelerators, DataHelix reduces mean deadline-miss rate by 67.8% and mean p95 input-wait latency by 87.0%, while achieving 2.72x geometric-mean throughput over rank-local DataLoaders and 1.82x over a Ray actor pool under the same CPU budget.
Luke Cheeseman, Elias Castegren, David Black-Schaffer, Erik Österlund, Tobias Wrigstad
原文摘要
In mainstream shared-memory languages, programmers use synchronisation mechanisms such as locks to isolate shared data accesses across threads to avoid data races. Yet synchronisation enforces only a temporal ordering, while shared memory still communicates updates between threads at every ordinary read and write. To achieve isolation, programmers therefore need to understand the connection between ordering and communication, and to apply synchronisation correctly and consistently, while compilers and hardware optimise assuming they have. When programmers synchronise incorrectly, they lose isolation, and the semantics of their programs change without notice. We propose Sync-as-Share, a memory model built on isolation and communication rather than on the ordering of accesses to shared memory. We start from thread isolation, where each thread reads and writes a private view of memory, and reinterpret programs' existing synchronisation actions to also communicate updates explicitly. As nothing is communicated outside synchronisation actions, data races become impossible by construction. Instead, conflicting updates are detected at synchronisation points. Viewing synchronisation as explicit communication exposes the design decisions of what should be communicated, to whom, and what counts as a conflict. We propose two communication topologies with different trade-offs. We instantiate Sync-as-Share for Java by making the synchronisation actions already defined by its memory model the only points at which threads communicate, so Java programs can adopt Sync-as-Share unmodified. We formalise this instantiation as version-controlled memory, in which synchronisation actions push and pull updates between private views, and mechanise it in Rocq. We show that Java programs that are correctly synchronised under the Java memory model behave identically under shared memory and Sync-as-Share.
Dennis N. Schneider, Lars Wagner, Daniel Rueckert, Dirk Wilhelm
原文摘要
Accurate grasping point prediction is a key challenge for autonomous tissue manipulation in minimally invasive surgery, particularly in complex and variable procedures such as colorectal interventions. Due to their complexity and prolonged duration, colorectal procedures have been underrepresented in current research. At the same time, they pose a particularly interesting learning environment due to repetitive tissue manipulation, making them a promising entry point for autonomous, machine learning-driven support. Therefore, in this work, we introduce attachment anchors, a structured representation that encodes the local geometric and mechanical relationships between tissue and its anatomical attachments in colorectal surgery. This representation reduces uncertainty in grasping point prediction by normalizing surgical scenes into a consistent local reference frame. We demonstrate that attachment anchors can be predicted from laparoscopic images and incorporated into a grasping framework based on machine learning. Experiments on a dataset of 90 colorectal surgeries demonstrate that attachment anchors improve grasping point prediction compared to image-only baselines. There are particularly strong gains in out-of-distribution settings, including unseen procedures and operating surgeons. These results suggest that attachment anchors are an effective intermediate representation for learning-based tissue manipulation in colorectal surgery.