Multimodal large laboratory models (MLLMs) still struggle with spatial understanding under the dominant perspective-image paradigm, which inherits the narrow field of view of human-like perception. For navigation, robotic search, and 3D scene understanding, 360-degree panoramic sensing offers a form of supersensing by capturing the entire surrounding environment at once. However, existing MLLM pipelines typically decompose panoramas into multiple perspective views, leaving the spherical structure of equirectangular projection (ERP) largely implicit. In this paper, we study pano-native understanding, which requires an MLLM to reason over an ERP panorama as a continuous, observer-centered space. To this end, we first define the key abilities for pano-native understanding, including semantic anchoring, spherical localization, reference-frame transformation, and depth-aware 3D spatial reasoning. We then build a large-scale metadata construction pipeline that converts mixed-source ERP panoramas into geometry-aware, language-grounded, and depth-aware supervision, and instantiate these signals as capability-aligned instruction tuning data. On the model side, we introduce PanoWorld with Spherical Spatial Cross-Attention, which injects spherical geometry into the visual stream. We further construct PanoSpace-Bench, a diagnostic benchmark for evaluating ERP-native spatial reasoning. Experiments show that PanoWorld substantially outperforms both proprietary and open-source baselines on PanoSpace-Bench, H* Bench, and R2R-CE Val-Unseen benchmarks. These results demonstrate that robust panoramic reasoning requires dedicated pano-native supervision and geometry-aware model adaptation. All source code and proposed data will be publicly released.

As language models scale, the amount of data they require grows -- yet many target data sources, such as low-resource languages or specialized domains, are inherently limited in size. A common strategy is to mix this scarce but valuable target data with abundant generic data, which presents a fundamental trade-off: too little target data in the mixture underexposes the model to the target domain, while too much target data repeats the same examples excessively, yielding diminishing returns and eventual overfitting. We study this trade-off across more than 2,000 language-model training runs spanning multiple model and target dataset sizes, as well as several data types, including multilingual, domain-specific, and quality-filtered mixtures. Across all settings, we find that repetition is a central driver of target-domain performance, and that mixture training tolerates much higher repetition than single-source training: scarce target corpora can be reused 15-20 times, with the optimal number of repetitions depending on the target data size, compute budget, and model scale. Next, we introduce a repetition-aware mixture scaling law that accounts for the decreasing value of repeated target tokens and the regularizing role of generic data. Optimizing the scaling law provides a principled way to compute effective mixture configurations, yielding practical mixture recommendations for pretraining under data constraints.

The development of separate-encoder Unified multimodal models (UMMs) comes with a rapidly growing inference cost due to dense visual token processing. In this paper, we focus on understanding-side visual token reduction for improving the efficiency of separate-encoder UMMs. While this topic has been widely studied for MLLMs, existing methods typically rely on attention scores, text-image similarity and so on, implicitly assuming that the final objective is discriminative reasoning. This assumption does not hold for UMMs, where understanding-side visual tokens must also preserve the model's capabilities for editing images. We propose G$^2$TR, a generation-guided visual token reduction framework for separate-encoder UMMs. Our key insight is that the generation branch provides a task-agnostic signal for identifying understanding-side visual tokens that are not only semantically relevant but also important for latent-space image reconstruction and generation. G$^2$TR estimates token importance from consistency with VAE latent, performs balanced token selection, and merges redundant tokens into retained representatives to reduce information loss. The method is training-free, plug-and-play, and applied only after the understanding encoding stage, making it compatible with existing UMM inference pipelines. Experiments on image understanding and editing benchmarks show that G$^2$TR substantially reduces visual tokens and prefill computation by 1.94x while maintaining both reasoning accuracy and editing quality, outperforming baselines on almost all benchmarks. Code is at: https://github.com/lijunxian111/G2TR.

LLM-powered multi-agent systems can now automate the full research pipeline from ideation to paper writing, but a fundamental question remains: automation for whom? Researchers operate under different resource configurations, hold different methodological preferences, and target different output formats. A system that produces uniform outputs regardless of these differences will systematically under-serve every individual user, making personalization a precondition for research automation to be genuinely usable. However, achieving it requires three capabilities that current systems lack: accumulating reusable procedural knowledge across projects, retaining user-specific experience across sessions, and internalizing implicit preferences that resist explicit formalization. We propose NanoResearch, a multi-agent framework that addresses these gaps through tri-level co-evolution. A skill bank distills recurring operations into compact procedural rules reusable across projects. A memory module maintains user- and project-specific experience that grounds planning decisions in each user's research history. A label-free policy learning converts free-form feedback into persistent parameter updates of the planner, reshaping subsequent coordination. These three layers co-evolve: reliable skills produce richer memory, richer memory informs better planning, and preference internalization continuously realigns the loop to each user. Extensive experiments demonstrate that NanoResearch delivers substantial gains over state-of-the-art AI research systems, and progressively refines itself to produce better research at lower cost over successive cycles.

We introduce an automatically generated benchmark for predicting hidden text in technical papers. A paper supplies visible context $X$ and a hidden continuation $Y$; the evaluated model writes an auxiliary forecast string $Z$, and a separate scorer assigns next-token probability to $Y$ both with and without conditioning on $Z$. This gives a label-free test of whether $Z$ transmits information about the continuation, compared against controls where $Z$ is recent context rather than a forecast. Our main testbed is equation-suffix prediction: the predictor sees context and the first part of a displayed equation, then forecasts the rest. The task mixes surface-level arXiv/TeX text modeling with reasoning-sensitive inference; the suffix is one of many roughly equivalent continuations, so the benchmark is read statistically rather than item-by-item. On 1363 equation continuations from 138 recent physics and mathematics papers, forecasts from GPT-5.5, Opus 4.7, and GPT-5.4 nano all improve clipped likelihood over the context control under both Qwen3-8B and Kimi K2.6 scorers, distinguishing model families and reasoning-effort settings without human labels. To emulate shortcuts where $Z$ further primes the scorer rather than making a useful forecast, we also fine-tune the scorer on context-only prompts and apply it to held-out papers as a stronger control. GPT-5.5 forecasts still beat this fine-tuned control; GPT-5.4 nano forecasts do not. Longer prose/TeX continuations show positive but noisier lift over controls, concentrated near the beginning of the target. These results support cross-model likelihood scoring as a static benchmark and as a setup for probing shortcut vulnerabilities before reinforcement learning or model-selection optimization is applied.

Architectural choices inside the Transformer feedforward network (FFN) block do not merely affect the block itself; they reshape the computations learned by the rest of the model. We study this effect in one-layer Transformers trained on digit addition with carry, modular arithmetic, and histogram counting. Comparing dense FFNs, gated linear units (GLUs), mixture-of-experts (MoE), and MoE-GLUs, we find that sparse MoE routing can shift computation from FFN to attention, with the strongest ablation-visible effect on carry-based addition. We decompose this redistribution into reduced per-token FFN capacity and sparse partitioning across experts. Critically, frozen random routing nearly matches learned routing, suggesting that redistribution is driven largely by architectural sparsity rather than router-learned specialization. As a secondary finding, GLU-style multiplicative gating rotates task-relevant Fourier structure out of the per-neuron basis and into distributed subspaces, making neuron-level interpretability less informative while preserving structured computation. We validate these conclusions with random-routing, narrow-FFN, and top-2 MoE controls, plus parameter-matching, activation-function, and width-scaling analyses. Together, these results show that local FFN design choices can have nonlocal consequences for Transformer computation.

It is becoming increasingly necessary to have monitors check for harmful behaviors during language model interactions, but text-only monitoring has not been sufficient. This is because models sometimes exhibit strategic deception and sandbagging, changing their behavior during evaluation. This motivates the use of white-box monitors like linear probes, which can read the model internals directly. Currently, such probes can fail under distribution shift, limiting their usefulness in real settings. We study whether there exists a low-dimensional subspace of the model internals that captures harmful behaviors more robustly, while leaving out spuriously correlative features. Inspired by the Assistant Axis and Persona Selection Model, we construct persona axes for deception and sycophancy using contrastive persona prompts. The first principal components, obtained by unsupervised PCA of the persona-specific vectors, cleanly separate harmful and harmless personas. Across 10 evaluation datasets, we show that persona-derived directions transfer non-trivially and probes trained on persona-PC projections generalize better than probes trained on raw activations. We also find that a unified axis consisting of multiple harmful and harmless behaviors improves generalization across behaviors and datasets. Overall, persona vectors provide a useful inductive bias for building more transferable behavior probes.

Deep generative models provide flexible frameworks for modeling complex, structured data such as images, videos, 3D objects, and texts. However, when applied to sequences of human skeletons, standard variational autoencoders (VAEs) often allocate substantial capacity to nuisance factors-such as camera orientation, subject scale, viewpoint, and execution speed-rather than the intrinsic geometry of shapes and their motion. We propose the Elastic Shape - Variational Autoencoder (ES-VAE), a geometry-aware generative model for skeletal trajectories that leverages the transported square-root velocity field (TSRVF) representation on Kendall's shape manifold. This representation inherently removes rigid translations, rotations, and global scaling of shapes, and temporal rate variability of sequences, isolating the underlying shape dynamics. The ES-VAE encoder maps skeletal sequences to a low-dimensional latent space incorporating the Riemannian logarithm map, while the decoder reconstructs sequences using the corresponding exponential map. We demonstrate the effectiveness of ES-VAE on two datasets. First, we analyze skeletal gait cycles to predict clinical mobility scores and classify subjects into healthy and post-stroke groups. Second, we evaluate action recognition on the NTU RGB+D dataset. Across both settings, ES-VAE consistently outperforms standard VAEs and a range of sequence modeling baselines, including temporal convolutional networks, transformers, and graph convolutional networks. More broadly, ES-VAE provides a principled framework for learning generative models of longitudinal data on pose shape manifolds, offering improved latent representation and downstream performance compared to existing deep learning approaches.

Zeroth-order (ZO) optimization has become increasingly popular and important in fine-tuning large language models (LLMs), especially on edge devices due to its ability to adjust the model to local data without the need for memory-intensive back-propagation. Recent works try to reduce ZO variance through low-dimensional subspace search, but subspace restriction alone leaves key optimization geometry under-exploited, motivating additional acceleration. In this work, we focus on the hidden layer training problem in which spectral optimizers like Muon outperform AdamW due to its ability to exploit weak spectral directions by orthogonalization. However, we have discovered that unlike in the first-order setting, full orthogonalization works poorly in the ZO setting since the gradient estimates are highly noisy and unreliable. To address this issue, we propose applying partial spectral orthogonalization to accelerate ZO optimization. To do so, we replace the iconic Newton-Schulz procedure in Muon with the faster, more concentrated power-iteration method so that it only amplifies dominant spectral directions. Furthermore, to improve the efficiency and generalization of the algorithm, we adopted a streaming variant of power-iteration that requires low variance in gradients, which was achieved through constraining our search inside a subspace obtained through the projection of momentum, echoing recent advances. Experiments on LLM fine-tuning show that our method can achieve from 1.5x to 4x the convergence speed of ZO-Muon, the current SOTA algorithm, across SuperGlue datasets in the OPT-13B model. Across different models, we also reach competitive final accuracies with less time in most cases compared with strong ZO baselines such as MeZO, LOZO and ZO-Muon. Code is available at https://github.com/MOFA-LAB/ZO-MOPI.git.

Autoregressive neural simulators now match classical solvers on short-horizon prediction of physical systems, yet their accuracy degrades rapidly when rolled out over long horizons. In this work, we identify transient amplification of perturbations around rollout trajectories as a structural mechanism driving rollout error. Using a linearization analysis we show that when the Jacobians along an autoregressive trajectory are non-normal and non-commuting, the model amplifies errors transiently, resulting in model rollout drift even when the overall system is asymptotically stable. Building on the analysis, we propose commutativity regularization: a combination of two penalties designed to reduce the normality defect of individual Jacobians and the commutator norm of Jacobians across steps. The penalties are estimated with Jacobian-vector products and have no inference-time cost. We show a propagator bound that quantifies rollout error under approximate commutativity and normality. We evaluate UNet and FNO variants with commutativity regularization on 1D and 2D spatio-temporal data in synthetic and real settings, showing successful long-horizon rollouts over thousands of steps. Further, we show that the method improves FourCastNet climate forecasts on ERA5 without using any new data. The gain is most pronounced out-of-distribution: trained on trajectories of a few hundred steps, regularized models remain in-distribution for thousands of rollout steps on initial conditions where baselines diverge.

The manifold hypothesis (MH) is often used to explain how machine learning can overcome the curse of dimensionality. However, the MH is only applicable in regimes where the training data provides a sufficiently dense sample of the underlying low-dimensional data manifold, or where such a low-dimensional manifold is conceivably present. We describe the regimes where the MH is not applicable as sparse. In this paper, we demonstrate that models succeed in the sparse regime by exploiting a highly structured local geometry, a property we formalize as normal alignment. We prove that normal-aligned classifiers -- whose input-output Jacobians are rank-one and align perfectly with the training data -- minimize the training objective under norm constraints and achieve maximal local robustness under a non-zero Jacobian constraint. For continuous piecewise-affine deep networks, normal alignment manifests geometrically as centroid alignment within the network's induced power diagram partition and results from the feature-learning regime. Motivated by these theoretical insights, we introduce GrokAlign, a regularization strategy that actively induces normal alignment. We demonstrate that GrokAlign significantly accelerates the training dynamics of deep networks relevant to the grokking phenomenon. Furthermore, we apply the principle of normal alignment to Recursive Feature Machines (RFMs) to introduce Recursive Feature Alignment Machines (RFAMs). We show that RFAMs exhibit greater adversarial robustness compared to RFMs when trained on tabular data.

Recent advances in large language models (LLMs) have demonstrated remarkable reasoning capabilities, largely stimulated by Reinforcement Learning with Verifiable Rewards (RLVR). However, existing RL algorithms face a fundamental limitation: their exploration remains largely constrained by the inherent capability boundary of the policy model. Although recent methods introduce external expert demonstrations to extend this boundary, they typically rely on complete trajectory-level guidance, which is sample-inefficient, information-sparse, and may confine exploration to a static guidance space. Inspired by the potential of multi-agent systems, we propose $\textbf{AIPO}$, an enhanced reinforcement learning framework that improves LLM reasoning through active multi-agent interaction during exploration. Specifically, AIPO enables the policy model to proactively consult three functional collaborative agents, $\textit{Verify Agent}$, $\textit{Knowledge Agent}$, and $\textit{Reasoning Agent}$, when encountering reasoning bottlenecks, thereby receiving fine-grained and targeted guidance to actively expand its capability boundary during training. We further introduce a tailored importance sampling coefficient together with a clipping strategy to mitigate the off-policy bias and gradient vanishing issues that arise when learning from agent-provided feedback. After training, the policy model performs reasoning independently without relying on collaborative agents. Extensive experiments on diverse reasoning benchmarks, including AIME, MATH500, GPQA-Diamond, and LiveCodeBench, show that AIPO consistently improves reasoning performance, generalizes robustly across different policy models and RLVR algorithms, and effectively expands the reasoning capability boundary of the policy model.

We introduce a probabilistic approach for dating historical manuscript pages from visual features alone. Instead of aggregating centuries into classes as is standard in the previous literature, we pose dating as an evidential deep regression problem over a continuous year axis, allowing our neural network to output a full predictive distribution with decomposed aleatoric and epistemic uncertainty in a single forward pass. Our architecture combines an EfficientNet-B2 backbone with a Normal-Inverse-Gamma (NIG) output head trained with a joint negative-log-likelihood and evidence-regularization objective. On the DIVA-HisDB benchmark (150 pages, 3 medieval codices, 151,936 patches), our model scores a test MAE of 5.4 years, well below the 50-year century-label supervision granularity, with 93\% of patches within 5 years and 97\% within 10 years. Our approach achieves \textbf{PICP=92.6\%}, the best calibration among all compared methods, in a single forward pass, outperforming MC Dropout (PICP=88.2\%, 50 passes) and Deep Ensembles (PICP=79.7\%, 5 models) at $5\times$ lower inference cost. Uncertainty decomposition shows aleatoric uncertainty is a strong predictor of dating error (Spearman $ρ=0.729$), and a selective prediction about the most certain 20\% of patches can provide \textbf{0.5 years MAE}. We show that predicted uncertainty increases as image degradation worsens, spatial decomposition maps explain which script regions cause aleatoric uncertainty, and page-level aggregation reduces MAE to 4.5 years with $ρ=0.905$ between uncertainty and page-level error.

Natural-language instance navigation becomes challenging when the initial user request does not uniquely specify the target instance. A practical agent should reduce the user's burden by actively asking only the information needed to distinguish the target from similar distractors, rather than requiring a detailed description upfront. Existing approaches often fall short of this goal: they may stop at the first plausible candidate before sufficiently exploring alternatives, or, even after collecting multiple candidates, ask about the target's attributes derived from individual candidates rather than questions selected to distinguish candidates in the pool. As a result, despite the dialogue, the agent may still fail to distinguish the target from distractors, leading to premature decisions and lengthy user responses. We propose Proactive Instance Navigation with Comparative Judgment (ProCompNav), a two-stage framework that first constructs a candidate pool and then identifies the target through comparative judgment. At each round, ProCompNav extracts an attribute-value pair that splits the current pool, asks a binary yes/no question, and prunes all inconsistent candidates at once. This reframes disambiguation from open-ended target description to pool-level discriminative questioning, where each question is chosen to narrow the candidate set. On CoIN-Bench, ProCompNav improves Success Rate over interactive baselines with the same minimal input and non-interactive baselines with detailed descriptions, while substantially reducing Response Length. ProCompNav also achieves state-of-the-art Success Rate on TextNav, suggesting that comparative judgment is broadly useful for instance-level navigation among similar distractors. Code is available at https://github.com/tree-jhk/procompnav.

Simulation-based inference (SBI) provides amortized Bayesian parameter inference from simulator-generated data without requiring explicit likelihood evaluation. Its reliability can degrade under model misspecification, where real-world observations are not well represented by the simulator used for training. Existing methods using unlabeled real-world data often align simulated and real-world data distributions, but marginal alignment alone does not directly preserve parameter-relevant information needed for posterior inference. We propose SPIN, an SBI framework with parameter-relevant information-preserving domain transfer using unlabeled, unpaired real-world observations. During training, SPIN translates labeled simulator observations toward the real-world domain and back to the simulator domain, using the original simulator labels to encourage domain transfer that preserves parameter-relevant mutual information. At test time, the learned real-to-simulator transport maps real-world observations into the simulator domain for posterior inference, without requiring real-world parameter labels or paired real--simulator observations. Across controlled synthetic and physical real-world benchmarks, SPIN improves real-world posterior inference, with the improvement becoming clearer as misspecification increases.

Agentic reinforcement learning (RL) for software engineering spends much of its compute on stateful trajectories whose grouped binary rewards are highly skewed and weakly contrastive. We frame this as pass-rate control and show that the binary reward-side signal is strongest near a 50% rollout pass rate under four criteria: reward entropy, group-filtering survival, leave-one-out (RLOO) advantage energy under Group Relative Policy Optimization (GRPO), and success-failure pair count. We propose Prefix Sampling (PS), which replays self-generated trajectory prefixes to steer skewed groups toward this regime: successful prefixes give mostly failing groups a head start, while failing prefixes handicap mostly passing groups. Replayed states are reconstructed through the existing rollout path, and replayed tokens are masked from the loss so optimization applies only to current-policy continuations. On SWE-bench Verified, PS reaches the baseline high-score regime within evaluation variability while delivering 2.01x and 1.55x end-to-end wall-clock speedups on Qwen3-14B and Qwen3-32B; the 14B peak improves from 0.274 to 0.295. AIME 2025 experiments on 4B and 8B show the same pass-rate-control pattern, and 4B ablations attribute gains to replay, bidirectional coverage, and adaptive control.

Training machine learning interatomic potentials (MLIPs) for reactive chemistry is often bottlenecked by the high cost of quantum chemical labels and the scarcity of transition state configurations in candidate pools. Active learning (AL) can mitigate these costs, but its effectiveness hinges on the acquisition rule. We investigate whether the latent space of a pretrained MLIP already contains the information necessary for effective acquisition, eliminating the need for auxiliary uncertainty heads, Bayesian training and fine-tuning, or committee ensembles. We introduce two acquisition signals derived directly from a pretrained MACE potential: a finite-width neural tangent kernel (NTK) and an activation kernel built from hidden latent space features. On reactive-chemistry benchmarks, both kernels consistently outperform fixed-descriptor baselines, committee disagreement, and random acquisition, reducing the data required to reach performance targets by an average of 38% for energy error and 28% for force error. We further show that the pretrained model induces similarity spaces that preserve chemically meaningful structure and provide more reliable residual uncertainty estimates than randomly initialised or fixed-descriptor-based kernels. Our results suggest that pretraining aligns latent-space geometry with model error, yielding a practical and sufficient acquisition signal for reactive MLIP fine-tuning.

Large language models often fail at simple counting tasks, even when items to count are in the prompt. We investigate whether this failure occurs because transformers do not represent counts internally, or because they cannot convert representations to the correct output tokens. Across three model families: Pythia, Qwen3, and Mistral, ranging from 0.4B to 14B parameters, we find evidence for the second explanation. Linear probes recover the correct count from intermediate layers with $R^2>0.99$, showing that the information is present. However, the internal directions that encode counts are nearly orthogonal to digit-token output-head rows ($|\cos| \leq 0.032$). In other words, the model stores the count in a form that the digit logits do not naturally read out. We localize this failure with two interventions. Updating only the digit rows of the output head (36,864 parameters) substantially improves constrained digit prediction (60.7--100.0% on four tasks), but it does not fix unconstrained generation (0%); we do not claim that digit-row repair fixes open-ended text. By contrast, small LoRA on attention Q/V (7.67M parameters) improves upstream routing and achieves 83.1%$\pm$7.2% in true greedy autoregressive generation (deployable fix). Logit-lens at layer 35 (entity counting; correct-digit rank): (i) median over 3 seeds drops from order-$10^4$ to 1; (ii) seed 42 shows $54{,}332 \to 838$ (median top-1 while one seed stays far below). Norm, logit-lens, and cross-task analyses generalize the bottleneck to counting, addition, and list length; nulls on MMLU and GSM8K and limited DROP transfer. These results identify counting failure as a geometric readout bottleneck, not an internal-representation failure: the model knows the count but the output pathway is misaligned with tokens needed to express it.

Long-term time series forecasting requires models that simultaneously capture rapid oscillations, medium-range periodicities, and slowly evolving macro-trends from a fixed look-back window. Existing lightweight MLP-based models typically operate on a single temporal resolution, limiting their ability to explicitly model patterns at multiple scales. We propose MSMixer, a channel-independent multi-scale MLP architecture that addresses this limitation through three complementary innovations: (i) three parallel scale branches at down-sample factors {1x, 4x, 16x} with independent MLP blocks, (ii) a learnable softmax gate that dynamically weighs branch outputs, and (iii) a DLinear complementary shortcut that provides full-window trend and seasonality context. MSMixer contains only 112K parameters at H=96 and runs at O(T) complexity. Evaluated on four ETT benchmarks with standard chronological splits and three random seeds, MSMixer achieves the lowest average MSE (0.357) among lightweight models, outperforming DLinear (0.386, -7.4%) and NLinear (0.365, -2.1%), winning 12 of 16 configurations. Against five Transformer-based baselines from the literature, MSMixer achieves best or second-best MSE in 9 of 16 configurations while using 5x fewer parameters than PatchTST. Ablation and sensitivity analyses confirm the complementary contributions of the multi-scale branches and the DLinear shortcut.

The effectiveness of active learning hinges on the choice of the acquisition criterion by which a learning algorithm selects potentially informative data points whose label is subsequently queried. This paper proposes a novel gradient-based acquisition criterion, derived from a generalization bound introduced by Luo et al. (2022). This criterion can be applied in lieu of uncertainty measures in uncertainty sampling, or incorporated into diversity-based methods that consider the spread of sampled points in addition to the uncertainty of their labels. We provide a theoretical justification of the proposed acquisition criterion, and demonstrate its effectiveness in an empirical evaluation.

Coherent Point Drift (CPD) is a representative probabilistic framework for unsupervised non-rigid point set registration. Its standard non-rigid M-step, however, relies on a point-indexed Gaussian-kernel system whose size grows with the number of moving points, making deformation estimation computationally heavy for large point sets and difficult to control in complexity during registration. To address these limitations, we propose Analytic-CPD, a new unsupervised non-rigid registration framework that gives CPD a structured analytic reformulation. Analytic-CPD preserves the CPD posterior correspondence layer, but lifts the M-step from point-indexed kernel displacement estimation to structured analytic mapping estimation. By coupling the Gaussian-mixture posterior mechanism of CPD with Structured Analytic Mappings (SAM), the method obtains a deformation model whose coefficient dimension is governed by the ambient dimension and analytic order rather than by the number of moving points. More importantly, deformation estimation is organized over an interpretable hierarchy of analytic function spaces, so the analytic order can be increased progressively as posterior correspondences become more reliable. We implement this idea through an increasing-degree continuation strategy with decreasing stage lengths: low-order analytic maps first stabilize the posterior correspondence structure, while higher-order modes later refine nonlinear residual deformation. Experiments on controlled model-matched, smooth model-mismatch, and registered human-shape data demonstrate the effectiveness and favorable accuracy--efficiency performance of Analytic-CPD.

Large language models (LLMs) are becoming increasingly capable mathematical collaborators, but static benchmarks are no longer sufficient for evaluating progress: they are often narrow in scope, quickly saturated, and rarely updated. This makes it hard to compare models reliably and track progress over time. Instead, we need evaluation platforms: continuously maintained systems that run, aggregate, and analyze evaluations across many benchmarks to give a comprehensive picture of model performance within a broad domain. In this work, we build on the original MathArena benchmark by substantially broadening its scope from final-answer olympiad problems to a continuously maintained evaluation platform for mathematical reasoning with LLMs. MathArena now covers a much wider range of tasks, including proof-based competitions, research-level arXiv problems, and formal proof generation in Lean. Additionally, we maintain a clear evaluation protocol for all models and regularly design new benchmarks as model capabilities improve to ensure that MathArena remains challenging. Notably, the strongest model, GPT-5.5, now reaches 98% on the 2026 USA Math Olympiad and 74% on research-level questions, showing that frontier models can now comfortably solve extremely challenging mathematical problems. This highlights the importance of continuously maintained evaluation platforms like MathArena to track the rapid progress of LLMs in mathematical reasoning.

End-to-end (E2E) autonomous driving aims to directly map sensory observations to driving actions, but its real-world deployment is hindered by evolving data distributions and the high cost of continual annotation. While combining imitation learning (IL) and reinforcement learning (RL) is a common strategy for policy improvement, conventional RL training relies on delayed, event-based rewards, where policies learn only from catastrophic outcomes such as collisions, leading to premature convergence to suboptimal behaviors. To address these limitations, we propose GSDrive, a framework that uses a differentiable 3D Gaussian Splatting (3DGS) environment for future-aware trajectory probing and reward shaping in E2E driving. GSDrive first learns a multi-mode trajectory probe via IL and then uses RL to evaluate multiple candidate futures in the 3DGS environment, converting their simulated returns into dense shaping rewards for policy optimization. This yields a cyclic hybrid IL-RL training loop, where IL supplies structured future priors and RL provides interactive feedback for iterative refinement. Evaluated on the reconstructed nuScenes dataset, our method outperforms other simulation-based RL approaches in closed-loop experiments. Code is available at https://github.com/ZionGo6/GSDrive.

Live future prediction refers to the task of making predictions about real-world events before they unfold. This task is increasingly studied using large language model-based agent systems, and it is important for building agents that can continually learn from the real world. It can provide a large number of prediction questions grounded in diverse real-world events, while preventing answer leakage. To leverage the advantages of future prediction, we present FutureWorld, a live agentic reinforcement learning environment that closes the training loop between prediction, outcome realization, and parameter updates. Specifically, we modify and extend verl-tool, resulting in a new framework that we call verl-tool-future. Unlike standard reinforcement learning training frameworks that rely on immediate rewards, verl-tool-future stores prediction-time rollouts, backfills rewards after real-world outcomes become available, and then replays the completed trajectories for policy update. Across three open-source agents, successive FutureWorld training rounds lead to consistent improvements in prediction accuracy, probabilistic scoring, and calibration, demonstrating that delayed real-world outcome feedback can serve as an effective reinforcement learning signal.

This paper presents a pipeline designed to transform raw Wikimedia dumps into quality textual corpora for seven South Slavic languages. The work is divided into two major phases. The first involves extracting and cleaning text from raw dumps of Wikipedia, Wikisource, Wikibooks, Wikinews, and Wikiquote. This step requires careful handling of raw wiki markup to isolate, first of all, textual articles, and then usable natural language text within them. The second phase addresses the challenge of questionable or low-quality articles, which are often generated from databases or structured knowledge bases. These articles are characterised by repetitive patterns, generic phrasing, and minimal to no original content. To mitigate their impact, a n-gram-based filtering strategy was employed to detect high levels of textual redundancy between articles and then remove such articles from the corpora entirely. The resulting datasets aim to provide linguistically rich texts suitable for training language models or conducting comparative research across South Slavic languages. By combining systematic extraction with quality control, this work contributes to the creation of reliable, high-information corpora that reflect the authentic cultural contexts of languages. While focused on the South Slavic case in the paper, the approach is mostly language-agnostic and can be generalised to other languages.

Large language models (LLMs) trained on unfiltered corpora inherently risk retaining sensitive information, necessitating selective knowledge unlearning for regulatory compliance and ethical safety. However, existing parameter-modifying methods face fundamental limitations: high computational costs, uncontrollable forgetting boundaries, and strict dependency on model weight access. These constraints render them impractical for closed-source models, yet current non-invasive alternatives remain unsystematic and reliant on empirical experience. To address these challenges, we propose the Controllable Alignment Prompting for Unlearning (CAP) framework, an end-to-end prompt-driven unlearning paradigm. CAP decouples unlearning into a learnable prompt optimization process via reinforcement learning, where a prompt generator collaborates with the LLM to suppress target knowledge while preserving general capabilities selectively. This approach enables reversible knowledge restoration through prompt revocation. Extensive experiments demonstrate that CAP achieves precise, controllable unlearning without updating model parameters, establishing a dynamic alignment mechanism that overcomes the transferability limitations of prior methods.

Failures in complex systems often emerge through gradual degradation and the propagation of stress across interacting components rather than through isolated shocks. Democratic systems exhibit similar dynamics, where weakening institutions can trigger cascading deterioration in related institutional structures. Traditional reliability and survival models typically estimate failure risk based on the current system state but do not explicitly capture how degradation propagates through institutional networks over time. This paper introduces a trajectory-aware reliability modeling framework based on Dynamic Causal Neural Autoregression (DCNAR). The framework first estimates a causal interaction structure among institutional indicators and then models their joint temporal evolution to generate forward trajectories of system states. Failure risk is defined as the probability that predicted trajectories cross predefined degradation thresholds within a fixed horizon. Using longitudinal institutional indicators, we compare DCNAR-based trajectory risk models with discrete-time hazard and Cox proportional hazards models. Results show that trajectory-aware modeling consistently outperforms Cox models and improves risk prediction for several propagation-driven institutional failures. These findings highlight the importance of modeling dynamic system interactions for reliability analysis and early detection of systemic degradation.

As terminal agents scale to long-horizon, multi-turn workflows, a key bottleneck is not merely limited context length, but the accumulation of noisy terminal observations in the interaction history. Retaining raw observations preserves useful environment feedback, but also leads to context saturation and high token cost; conversely, naive compression may discard task-critical signals needed for subsequent actions. Because terminal environments are highly heterogeneous across repositories, commands, and execution states, heuristic-based or fixed-prompt compression methods are difficult to generalize. We propose TACO, a plug-and-play, training-free, self-evolving Terminal Agent Compression framework for existing terminal agents. TACO automatically discovers, refines, and reuses structured compression rules from interaction trajectories, enabling workflow-adaptive filtering of low-value terminal outputs while preserving task-relevant observations. Experiments on TerminalBench (TB 1.0 and TB 2.0) and four additional terminal-related benchmarks, including SWE-Bench Lite, CompileBench, DevEval, and CRUST-Bench, show that TACO consistently improves task performance and token efficiency across agent scaffolds and backbone models. On TerminalBench, TACO yields 1%-4% accuracy gains across strong agentic models and improves accuracy by around 2%-3% under the same token budget. On additional terminal-related benchmarks, it reduces total token consumption while maintaining or improving task success rates. These results suggest that self-evolving, workflow-adaptive observation compression is an effective path toward more reliable and efficient long-horizon terminal agents. The code is publicly available at https://github.com/multimodal-art-projection/TACO.

Quantization has become a standard tool for efficient LLM deployment, especially for local inference, where models are now routinely served at 2-3 bits per parameter. The state of the art is currently split into simple scalar quantization techniques, such as GPTQ or AWQ, which are widely deployed but plateau in accuracy at 3-4 bits per parameter (bpp), and "second-generation" vector- or trellis-quantized methods, such as QTIP, GPTVQ and AQLM, which push the accuracy frontier but are notoriously hard to implement and to scale. In this paper, we ask whether this gap is fundamental, or whether a carefully optimized $\textit{scalar}$ quantizer can recover most of it. We answer in the affirmative, by introducing GSQ (Gumbel-Softmax Quantization), a post-training scalar quantization method which jointly learns the per-coordinate grid assignments and the per-group scales using a Gumbel-Softmax relaxation of the discrete grid. GSQ matches the cardinality of the relaxation to the small number of levels available in the target bit-width regime (e.g., 3-8 levels for ternary and 3 bpp, respectively), making optimization tractable. Practically, on the standard Llama-3.1-8B/70B-Instruct models, GSQ closes most of the gap between scalar quantization and the QTIP frontier at 2 and 3 bits, while using a symmetric scalar grid with group-wise quantization, and thus remains compatible with existing scalar inference kernels. We further show that the same discrete-assignment optimization can be applied to practical GGUF K-Quant checkpoints: starting from publicly released GGUF models, GSQ improves accuracy while projecting the result back into the same deployment format. Finally, GSQ scales to trillion-scale Mixture-of-Experts models such as Kimi-K2.5, where vector-quantized methods are difficult to apply. The source code is publicly available at https://github.com/IST-DASLab/GSQ.

Automated medical report generation for 3D PET/CT imaging is fundamentally challenged by the high-dimensional nature of volumetric data and a critical scarcity of annotated datasets, particularly for low-resource languages. Current black-box methods map whole volumes to reports, ignoring the clinical workflow of analyzing localized Regions of Interest (RoIs) to derive diagnostic conclusions. In this paper, we bridge this gap by introducing VietPET-RoI, the first large-scale 3D PET/CT dataset with fine-grained RoI annotation for a low-resource language, comprising 600 PET/CT samples and 1,960 manually annotated RoIs, paired with corresponding clinical reports. Furthermore, to demonstrate the utility of this dataset, we propose HiRRA, a novel framework that mimics the professional radiologist diagnostic workflow by employing graph-based relational modules to capture dependencies between RoI attributes. This approach shifts from global pattern matching toward localized clinical findings. Additionally, we introduce new clinical evaluation metrics, namely RoI Coverage and RoI Quality Index, that measure both RoI localization accuracy and attribute description fidelity using LLM-based extraction. Extensive evaluation demonstrates that our framework achieves SOTA performance, surpassing existing models by 19.7% in BLEU and 4.7% in ROUGE-L, while achieving a remarkable 45.8% improvement in clinical metrics, indicating enhanced clinical reliability and reduced hallucination. Our code and dataset are available on GitHub.