让下一篇重要论文，不再错过。

Recent work on robot manipulation has focused on specialist policies trained on demonstrations. We show that generalist AI models can control manipulators more reliably than specialist policies such as vision-language-action models (VLAs) and world-action models (WAMs). We systematically analyze agentic manipulation policies (AMPs), in which a generalist AI model commands the manipulator directly, adapting its actions as the task unfolds. We present AMP1, a reference harness that runs unchanged across single-arm, bimanual, and mobile manipulation. AMP1 achieves state-of-the-art results on the RoboLab-120 and CaP-Gym benchmarks, with no demonstrations or task-specific training. On RoboLab-120, AMP1 reaches an 85.6\% success rate zero-shot, compared to 49.2\% and 42.9\% for leading VLA and WAM policies. The agent succeeds zero-shot in 104 of 108 real-world trials spanning twelve tasks and three robot instances, reducing failure rates by 20x and 15x relative to strong VLA and WAM baselines. The basic harness supplies no grasping, pushing, or placing skills; the agent develops its own strategies, such as searching for hidden objects, using objects as tools, and calibrating its cameras. With the addition of a persistent memory store, the agent can also learn from experience: an augmented harness with a memory representation that is carried across episodes lowers the failure rate on RoboDojo bimanual manipulation tasks by 4.6x relative to zero-shot deployment. Our analysis suggests that AMPs are substantially more general and robust than specialist manipulation policies, providing a promising foundation for robot manipulation.

We prove that the set $\mathcal{N}_N$ of ReLU networks with one hidden layer and at most $N\in\mathbb N$ neurons is weakly closed in $W^{1,p}(Ω)$, for a bounded connected open set $Ω\subset\mathbb R^d$. Consequently, every coercive, weakly lower semicontinuous energy on this space attains its minimum over $\mathcal{N}_N$. This covers the Deep Ritz method with convex integrands, Sobolev training and Lipschitz learning. Moreover, as the weight decay parameter tends to zero, minimizers of the regularized problem converge, up to subsequences, to minimizers with minimal parameter norm. We also prove that the set of ReLU networks with two hidden layers of bounded widths is not closed in $W^{1,p}(Ω)$ or in $C(\overlineΩ)$. This disproves a conjecture of Petersen, Raslan and Voigtlaender.

Human motion prediction combines the tasks of trajectory forecasting and human pose prediction. For each of the two tasks, specialized models have been developed. Combining these models for holistic human motion prediction is non-trivial, and recent methods have struggled to compete on established benchmarks for individual tasks. To address this, we propose a simple yet effective transformer-based model for human motion prediction. The model employs a stack of self-attention modules to effectively capture both spatial dependencies within a pose and temporal relationships across a motion sequence. This simple, streamlined, end-to-end model is sufficiently versatile to handle pose-only, trajectory-only, and combined prediction tasks without task-specific modifications. We demonstrate that this approach achieves state-of-the-art results across all tasks through extensive experiments on a wide range of benchmark datasets, including Human3.6M, AMASS, ETH-UCY, and 3DPW.

The Gaussian jet of an image is a graded vector space: a dilation by $λ$ multiplies derivatives of order $k$ by $λ^{-k}$. Local graded convolutions are exactly homogeneous differential operators, and a local feature is dilation-invariant if and only if it factors through the scale-normalized jet or, without the scale, through the retraction onto the weighted sphere of the grading. Under unknown Gaussian blur the retraction changes only by a change of scale, whereas the scale-normalized jet acquires a grading factor that no radial normalization removes. At the zero jet the retraction separates image families that radial normalizations and continuous classifiers cannot. Networks built on the retraction are universal for invariant features. On MNIST such a network keeps 88.4% accuracy under a two-pixel blur, against 66.2% for the same network on scale-normalized derivatives; in Lindeberg's Gaussian derivative network the retraction raises accuracy at that blur from 91.5% to 95.1%.

Loco-manipulation, which requires mobile manipulators to coordinate locomotion and manipulation over long horizons, is a key capability for large-object rearrangement in applications such as service robotics and warehouse automation. Despite recent progress, reliably executing these tasks remains challenging because performance depends on complex robot-object and object-environment interaction dynamics. Existing approaches address this challenge either through model-based methods, which often require accurate dynamics models, or through learning-based methods, which reduce this requirement but typically assume a fixed or predefined initial grasp. However, different geometrically stable grasps can lead to substantially different downstream manipulation performance. Therefore, we propose Grasp-TAGS, a task-aware grasp selection framework for goal-conditioned large-object rearrangement. We train a unified PPO-based reinforcement learning policy across multiple initial grasp configurations and distill the learned critic into a lightweight grasp utility network that ranks candidate grasps using only the candidate grasp configuration and goal information. At inference time, we select the grasp with the highest predicted downstream utility for the downstream policy to complete the task, before even making a contact. Experiments demonstrate that Grasp-TAGS improves rearrangement success over both fixed single-grasp policies and random grasp selection, achieving an average success rate of 89.84% compared with 79.87% for the best-performing single-grasp policies, an improvement of 9.97 percentage points. Grasp-TAGS also reduces final orientation errors across all evaluated objects. For more qualitative results, refer to the supplementary webpage: https://rpm-lab-umn.github.io/Grasp-TAGS-web-page/

Many multimodal alignment problems require mapping a source modality into a fixed space defined by a target modality, where the target space is not merely an intermediate embedding but is used directly for downstream analysis. In this setting, successful alignment should preserve both the global structure of the target distribution and the instance-level correspondence observed in paired source--target training data. Existing objectives typically address only part of this goal: contrastive losses encourage relative similarity, distribution-matching losses improve marginal agreement, and pointwise regression fits each source sample to its paired target without explicitly modeling target-space path geometry. We propose FlowAlign, a flow-guided target-space alignment method that repurposes conditional flow matching as a training-time alignment signal rather than an inference-time transport model. FlowAlign uses the learned flow field to guide the alignment of source-side representations with the fixed target space, encouraging them to preserve paired correspondence while remaining compatible with the global target-space structure. We further provide theoretical analysis showing how FlowAlign helps improve paired correspondence by controlling endpoint accuracy and capturing local target-space geometry. Experiments on synthetic data, paired single-cell multiome benchmarks, and image--text alignment demonstrate that FlowAlign improves instance level correspondence while preserving target-space structure across different modality settings.

In many areas of AI, general-purpose models have increasingly replaced handcrafted methods and specialized architectures. We argue that embodied AI is poised for a similar shift. Our key claim is that embodied tasks can be formulated as interaction with digital software. Specifically, we build software that renders an agent's view of a physical environment on a screen and provides controls for acting in it. A computer-use agent (CUA) can then perform embodied tasks through the same screenshots and keyboard and mouse actions that it uses to operate ordinary software. Using a single CUA design and changing only the software, our entries win first-, second-, third-, and sixth-place prizes across four ECCV workshop competitions, without any competition-specific agent tuning. These results suggest that the skills CUAs learn on software, such as visual grounding, precise clicking, and exploration, already carry over to embodied tasks, giving CUAs a far broader scope in embodied AI than previously realized. We therefore advocate that computer-use agents can be embodied generalists. Further, we advocate that, by following the training strategy that made computer-use agents general, embodied AI can converge on a single general model as well, ushering in a new era of general embodied AI. Website: https://pranjal2041.github.io/cua-embodied

For the combinatorial graph alignment problem (GAP), finding the node correspondence that maximizes the number of common edges (nce) between two unlabeled graphs, the classical Frank-Wolfe heuristic FAQ remains a strong baseline once it is properly initialized, while existing GNN approaches struggle in the purely structural setting. We introduce a chaining procedure: a sequence of Folklore-type (2-FWL) GNNs in which each network is trained with cross-entropy after decoding the previous network's similarity matrix and ranking nodes by their current alignment quality. This non-differentiable ranking step injects discrete combinatorial feedback at every stage; at inference, we iterate the final network and keep the candidate with highest observed nce. On sparse Erdos-Renyi graphs at noise level 0.25, chained FGNNs used to initialize FAQ reach 85% accuracy versus 13% for FAQ initialized from the convex relaxation, and essentially 0% for prior GNN methods. On correlated regular graphs, where MPNNs with constant features produce identical node embeddings (1-WL fails to refine) and FAQ's convex initialization is degenerate, chaining is the only method among those we evaluate, including Gromov-Wasserstein solvers, that recovers a non-trivial alignment. On three real-world benchmarks (yeast PPI, coauthorship, and road networks), recent comparisons underestimate FAQ by initializing it from a uniform doubly stochastic matrix; initialized from the convex relaxation it already surpasses prior reported numbers, and dataset-specific chained FGNNs further improve on this strengthened baseline. Code and pretrained chains available at: https://github.com/mlelarge/chaining-gnn-graph-alignment.

The pigs were bleeding. That was the point of the experiment: could airway pressure reveal hemorrhage before it became critical? In the UCR Time Series Classification Archive, the de facto benchmark for Time Series Classification (TSC), the same recordings ask a different question: "which of 52 pigs produced this trace?". The dataset PigAirwayPressure is not an outlier. Although the archive is the standard benchmark, whether its tasks resemble the problems TSC is meant to solve has rarely been examined. We find that several datasets are repurposed from their original task, built from toy setups, or have flawed test splits. We further question how representative the archive's ordered non-temporal sequences (e.g. spectra, shape outlines) are of real TSC, and the potential downsides from the decomposition of multivariate problems. Our findings call for more careful interpretation of archive-wide scores and greater attention to deployment-relevant evaluation. The audit is at https://github.com/automl/ucr-provenance-audit.

Radiology AI has developed along two paths: specialized models that perform well on narrow tasks but cover few clinical findings, and generalist models that address a broader range of tasks but remain prone to hallucination and inaccurate measurements. Combining the two through agentic systems offers a route towards comprehensive radiology reporting, but published systems are tied to one modality or body part. We describe a radiology agentic framework that standardizes how an orchestrator calls modular tools to produce a structured report, and use it to measure what such a system can do with the tools from FDA-cleared radiology AI devices. Report generation is simulated on a large multi-modality report corpus, with the orchestrator given either a catalog of finding classes consolidated from the cleared devices, each available on the modalities it is cleared for, or an unlimited catalog in which any class can be requested, and planning either from the clinical indication or from the target report. Tool outputs are simulated from the reference report, so the comparison isolates tool availability and orchestration from tool error. With the cleared catalog and the indication alone, the orchestrator recovers under a third of the clinically actionable findings; recovery improves substantially with the unlimited catalog, and further when the plan is written from the target report. Attributing every missed finding to a cause shows that most are classes no cleared device offers for the study's modality, and that a third are classes or tools that were offered and not called. Adding the most frequently missing tools closes most of the gap to the unlimited catalog on held-out reports, whereas a smaller orchestrator with the unlimited catalog gives up the whole gain. Tool availability and effective orchestration are thus complementary constraints on comprehensive AI-assisted radiology reporting.

Reinforcement learning (RL) rewards are notoriously difficult to design and control, often leading to the model learning unintended behaviors such as reward hacking. One potential solution is to monitor for reward hacking and penalize it when detected; however, training against a monitor could lead to evasive behavior, and our general understanding of how to apply monitors effectively during training is limited. To study how best to use monitors to mitigate reward hacking, we introduce and open source three realistic environments where Qwen3-4B reward hacks: a coding environment hackable via test overwriting, a medical chat environment hackable via sycophancy, and a biography generation environment hackable via hallucination. We first focus on the coding environment, where we find that: (1) models can learn to evade highly accurate monitors by exploiting systemic flaws in probes and LLM judges; (2) monitors that leak more learning signal during RL suppress reward hacking but are more often evaded; and (3) including easier problems in training can decrease reward hacking. We apply our findings to build better reward hacking monitors for the medical chat and biography generation environments that improve upon naive baselines to reduce reward hacking rates across seeds to before-training levels in some interventions. Our results demonstrate that our takeaways translate to new settings and that better monitor intervention designs are possible.

Machine learning (ML) methods have been explored as a means of optimizing lattice field theory calculations, with increasing interest in recent years. This chapter gives an introduction and summary to the use of ML methods to optimize lattice field theory without introducing any uncontrolled approximations. We focus on those directions which are clearly established as of late 2025, marking the 50th anniversary of the beginning of the study of lattice gauge theory. This chapter is a contribution to the online book ``Lattice QCD at 50 years,'' (LQCD@50), edited by Tanmoy Bhattacharya, Maarten Golterman, Rajan Gupta, Laurent Lellouch, and Steve Sharpe.

This paper presents a comprehensive comparison of four representative Reinforcement Learning (RL) methods in autonomous vehicle (AV) decision-making, including Fitted Value Iteration (FVI), Rollout Classification Policy Iteration (RCPI), Deep Q-Network (DQN), and a discrete-action adaptation of Deep Deterministic Policy Gradient (D-DDPG). The four methods are applied to the same AV decision-making problem at an unsignalized intersection and are compared in terms of optimality, safety, training stability, and training and decision cost. Observations suggest that RCPI provides the closest estimation to the exact solution and has the most stable training curve; however, it is data-intensive and requires the largest volume of data. D-DDPG is the most data-efficient one, but at the price of the training stability. Based on the observations, we provide suggestions for the algorithm selection for AV decision making.

Public benchmark scores reflect skill, prior exposure to the questions, or both, and for most models the training data are unknown. CleanScore is a black-box audit: the estimator needs only a correct-or-incorrect score for each form, not training data or weights. Each benchmark question becomes a parent item with one public form and two independently written fresh forms that keep its numbers, facts and answer. The audit reports an interval for the public-form advantage, not a verdict, and a private control bank with an explicit transport radius provides a proxy for the ordinary form gap that a rewrite alone can produce. A registered controlled-exposure experiment, a preliminary validation, detects planted exposure in one of two model families. A registered audit of five open models on 200 GSM8K and 200 ARC-Challenge items (ARC with stem-only rewrites, scored by option log probabilities) finds no exposure-consistent advantage, and two reruns declared before they were run (GSM8K at a longer token budget, ARC scored from generated letters) leave that conclusion unchanged. For four of the ten model-benchmark pairs the upper end of the interval is about five points or less, and for none is it above 21 points. The registered claim rule is much less sensitive than the primary interval: it flags a planted ten-point advantage in 5% to 11% of audits. Registered positive controls show what a null can mean. In four models, each trained under four seeds, items leaked into fine-tuning gained 25 points on average over unleaked items on the leaked wording and 19 points on paraphrases the model never saw, so a substantial share of the effect, estimated at 76% (95% interval 50% to 113%), is invisible to a paraphrase audit. A null on surface form says much less about contamination than the name suggests.

Uncertainty bands on wind forecasts help inform reserve sizing and commitments, but often fail to deliver their stated coverage. We apply conformal prediction to native bands to see what it adds and what it costs, comparing six forecasters, from ARIMA to a tabular foundation model, at two wind farms in Great Britain (GB) under operational conditions. We show that calibration repairs empirical marginal coverage with little or no Winkler Score penalty, but this apparent success conceals under-coverage in high winds. Calibrating by forecast wind regime turns marginal coverage into regime-level coverage, cutting day-ahead out-of-band volume in high winds by over 40% relative to pooled calibration while narrowing bands overall.

Chemical reaction datasets such as USPTO suffer from substantial incompleteness, frequently missing byproducts, co-reactants, and stoichiometric coefficients. This limits their applicability and reliability in downstream applications. Here, we introduce CompleteRXN, a large-scale supervised benchmark for reaction completion under realistic missing-data conditions. We construct a dataset of aligned incomplete and atom-balanced reactions by mapping USPTO records to curated mechanistic reactions. We evaluate representative baselines, including a novel encoder-decoder reaction completion model with constrained decoding, the Constrained Reaction Balancer (CRB), and a recent algorithmic method, SynRBL. On our CompleteRXN benchmark, the CRB achieves high performance across splits of increasing difficulty, reaching 99.28% equivalence accuracy on the random split and 91.71% on the extreme out-of-distribution split. SynRBL produces many balanced and chemically plausible completions, but with lower accuracy on the benchmark test splits. Across all methods, performance degrades with increasing incompleteness. We observe a substantial drop when evaluating on reactions outside the benchmark (full uncurated USPTO), highlighting the gap between benchmark performance and practical robustness and motivating future work. Data, code and models can be found at: https://github.com/Intelligent-molecular-systems/CompleteRxn-Benchmarking.

Representations are routinely used across machine learning, psychology, and neuroscience to draw inferences about the computations of biological and artificial systems. Such inferences presume that the geometry of these representations reflects the computation being performed. For artificial neural networks, however, the extent to which the function a network computes constrains this geometry remains unclear. One key obstacle is that these networks admit parameter symmetries: changes in parameterization that preserve function exactly while reshaping representational geometry. Here, we show that a broad class of parameter symmetries acts on representations through just three primitive feature transformations: addition, duplication, and scaling. This feature-level characterization yields a closed-form decomposition of representational geometry into essential and auxiliary components, revealing how representational degeneracy can exist without and grows with overparameterization even when function is held fixed. Finally, we establish conditions under which implementation-level selection rules resolve this degeneracy, yielding identifiable geometries in which features are weighted according to their contributions to the network's function. Together, our results delineate when representational geometry can support inferences about computation, and when it cannot.

We study planning for finite-horizon partially observable Markov decision processes (POMDPs) with finite state, action, and observation spaces. Both exact and approximate planning are known to be computationally hard in this setting, so prior work has explored natural assumptions that make planning easier. Recent work [Golowich et al., 2022] gave quasipolynomial-time approximate planning algorithms for observable POMDPs. These are POMDPs whose observation channels satisfy an $\ell_1$-observability condition, but notably, no assumptions are required on the transition dynamics. In a similar vein, for every fixed $ρ>0$, we give quasipolynomial-time approximate planning algorithms for $ρ$-decodable POMDPs. We show that decodability is implied by both $\ell_1$- and $\ell_2$-observability and is strictly weaker than observability. Moreover, we show that the optimal decodability constant can be computed in polynomial time, whereas computing the optimal $\ell_1$-observability constant is NP-hard in the worst case. We also show that the fast filter-forgetting guarantee underlying the approach of [Golowich et al., 2022] does not extend directly from $\ell_1$-observability to constant $\ell_2$-observability. Thus, decodability is an easy-to-check, unifying assumption under which quasipolynomial-time approximate planning for POMDPs is possible.

Evaluating large language models increasingly relies on LLM-as-a-judge protocols, but such evaluations remain costly: different judges have different prices and reliabilities, and the difficulty of each prompt-response pair can vary substantially. This raises a basic allocation question: under a fixed budget, how should one distribute evaluation queries across heterogeneous judges and instances to obtain the most accurate score estimates? We formalize this question as budgeted heteroskedastic multi-judge estimation under independent, unbiased score observations. Given $K$ prompt-response pairs and $J$ judges with known costs but unknown query-judge variances, our goal is to estimate the bounded score vector while minimizing an $\ell_p$-error. Our first contribution is to analyze the inverse-variance weighted estimator (IVWE) and to derive the oracle allocation that minimizes its leading error bound. Since this allocation depends on the unknown variances, we then address the unknown-variance setting by proposing EST-IVWE, an adaptive algorithm that constructs and uses optimistic variance estimates to stabilize the empirical allocation. We prove that EST-IVWE matches the oracle IVWE rate up to lower-order terms in the budget. Our second and central theoretical contribution is a local minimax lower bound matching the leading allocation-dependent rate, establishing asymptotic instance-optimality up to logarithmic factors. An Assouad-type construction based on local, variance-preserving perturbations yields this bound both in expectation and with constant probability. A complementary Fano-type change-of-measure argument captures the logarithmic dependence on the confidence level with a different (suboptimal) allocation geometry. Finally, we numerically validate our approach on synthetic data and resampled HelpSteer2 evaluations, demonstrating improvements over naïve uniform allocation.

Probably Approximately Correct (PAC) bounds on the generalization error (the difference between the empirical prediction error computed on the training data and its expectation) are an established tool in statistical machine learning, and there have been several extensions of these bounds to dynamical systems. In parallel, in recent years, there has been growing interest in finite-sample bounds for parameter estimation error in system identification. The latter bounds tend to be algorithm-specific, relying on reformulating the identification algorithm as a least-squares regression problem. This is in contrast to classical PAC bounds, which are algorithm-agnostic and can be applied to any identification algorithm based on minimizing a cost function involving the empirical loss. In this paper, we show how oracle bounds derived from PAC bounds can be used to derive finite-sample bounds for dynamical systems. As an illustration of this approach, we present finite-sample parameter-estimation bounds for linear and bilinear systems with bounded inputs and noise. These bounds apply to identification algorithms based on minimizing a regularized empirical loss function. The key ingredient is to establish an oracle inequality, i.e., an inequality between the true prediction error of the learned model and the minimum achievable true prediction error, and then use either the convexity of the true loss or system-theoretic properties of the loss function to derive bounds on the parameter estimation error. We also analyze the rates of the obtained bounds and compare them with those of classical asymptotic results.

Chain-of-thought (CoT) reasoning has emerged as a powerful paradigm for extending the computational capabilities of transformer-based models. A growing body of recent work has begun to uncover the fine-grained computational power of transformers augmented with CoT, as well as their connections to classical complexity theory. In this work, we advance this line of research by establishing the first hierarchy theorem for CoT-based computation. Our main result shows that for any reasonable resource bound $r(n) \geq n$ and any $\varepsilon > 0$, \[ \mathrm{CoT}(r(n)) \subsetneq \mathrm{CoT}(r(n)^{1+\varepsilon}), \] where $\mathrm{CoT}(r(n))$ denotes the class of decision problems solvable by transformers using $O(r(n))$ CoT reasoning steps on inputs of length $n$. Our result shows that increasing the number of CoT reasoning steps provably yields strictly greater computational power, establishing CoT steps as a fundamental resource in transformer-based computation. Unlike classical hierarchy theorems, which are typically proved via diagonalization, our approach leverages known coarse-grained simulations between CoT-augmented transformers and Turing machines in both directions. Building on these connections, we adapt the classical padding technique from complexity theory to this setting to obtain a tight hierarchy. A key technical contribution of our work is the implementation of padding within the transformer-based CoT model.

Equivariant Graph Neural Networks (GNNs) have significantly advanced the modeling of 3D molecular structure by leveraging group representations. However, message passing built only from exactly equivariant operations can have restricted practical expressiveness at a fixed representation degree. We introduce the Equivariant Spherical Transformer (EST), a plug-and-play module that applies Transformer attention to spherical spatial representations obtained from group representations. EST explicitly mixes information across sampled orientations, unlike prior orientation-wise spherical operations, while retaining a strong approximate equivariant inductive bias through near-uniform spherical sampling. As demonstrated by our experiments on challenging benchmarks like OC20, MPtrj and QM9, EST-based models achieve state-of-the-art performance. For the complex molecular systems within OC20, small models empowered by EST can outperform some larger models and those using additional data. We characterize the ideal closed-domain condition under which the continuous operation is exactly equivariant and empirically quantify the finite-sampling accuracy--equivariance--memory trade-off used in practice.

OpenAI's agents breached Hugging Face in July 2026, and the same company released more than seven hundred AI-generated manuscripts of mathematical proofs in October 2026. The first event is read as a failure in security, while the second is read as a success in mathematics. We read both through the hybrid framework of our paper (Kim and Kang, 2026, arXiv:2611.01914), in which coding, proving, and learning from data are derivations by one and the same deductive mechanism, distinguished by the kinds of premises from which they are derived. Under this framework, the AI systems behind the two events ran the same loop, in which a statistical proposer produces candidate exploits or proofs and a deductive verifier settles each one by a derivation. What separates the events is the logical form of the statement wanted by each of the two domains, security and mathematics. The safety of a program is a $Π_1$ statement, for it says that no exploit exists, which is not assured by any finite list of tests that have been passed by the program. By contrast, the provability of a sentence is a $Σ_1$ statement, for it says that a proof exists, which is assured by a single instance, namely a proof that has been checked by the deductive verifier. The theorem behind both is Turing's Undecidability Theorem: the halting set is $Σ_1$ but not decidable, so its complement is $Π_1$ and not $Σ_1$, which is to say that membership has a finite witness but non-membership has none. We prove that attaching a statistical proposer to a deductive verifier does not change this undecidability result, however good the proposer is. We then discuss what does assure safety and where the residual risk goes.