The paper tackles the frozen-context bottleneck in outcome-based reinforcement learning for forecasting: prior work either pastes fixed research snippets into the prompt before training or lets a frozen model search only at test time, so the policy is never rewarded for gathering better evidence. The authors close this gap by building a multi-turn agentic harness from 2,100+ resolved Polymarket questions, filtering every search and page-read to pre-cutoff information, and training Qwen3.5-35B-A3B with single-epoch GRPO under the Brier-score reward $r=1-(p-y)^2$. The core advance is that the model learns when and what to search for as a temporally grounded skill, rather than optimizing a frozen context window. This matters because it moves forecasting from prompt engineering to genuinely learned research discipline, and the trained open model reportedly beats several frontier systems at roughly 5% of the inference cost.

Multimodal reinforcement learning with verifiable rewards (RLVR) is widely assumed to improve visual reasoning when models are rewarded for correct answers. This paper exposes a critical gap: benchmark gains can be achieved without learning from images at all, and prolonged training can even erode visual grounding. The authors propose "visual resolvability"—designing tasks where visual evidence is strictly necessary for reward—and show that standard GRPO on carefully constructed counterfactual coordinate scenes drives genuine visual discovery, unlike conventional benchmark training.

Relational learning with tabular foundation models currently relies on Deep Feature Synthesis (DFS), which flattens multi-table schemas into single tables via column-wise aggregations along meta-paths. This paper identifies two critical limitations of DFS—feature explosion and interaction blindness—and proposes RelICL, a training-free method that instead propagates row embeddings through the schema graph using the foundation model itself. By iteratively pooling and fusing embeddings, RelICL keeps the TFM-agnostic benefits of DFS while scaling to complex schemas and capturing cross-column interactions. The experimental study on 21 RelBench tasks shows RelICL matches state-of-the-art DFS approaches, though it does not dominate them outright.

Many applications require sampling from high-dimensional categorical distributions with local dependence, but discrete diffusion models lack end-to-end statistical guarantees. This paper introduces a pinning decomposition of the discrete score under uniform noising, showing that time dependence separates multiplicatively from target dependence for Markov random fields of order $d$. Building on this, the authors design a weight-sharing neural score learner and prove a sample-complexity rate of $\widetilde O_{\mathbb P}(S^d/n)$, replacing the ambient dimension $D$ with the interaction order $d$.

Graph positional encodings (PSEs) are computed from pairwise edges and inherently miss higher-order interactions such as triangles and cliques. The paper proposes to enrich standard graph models by computing positional encodings on lifted higher-order domains (e.g., clique complexes) and projecting them back to nodes, without altering the backbone. Its core theoretical insight is that lift-induced symmetric node operators can mix the eigenmodes of the original graph Laplacian $L$, placing them outside the class of scalar spectral filters $g(L)$. This yields a principled design rule---the commutator $[L,M] \neq 0$---and an experimental pipeline using Hodge-Laplacian encodings.

Monitoring contact wire geometry is essential to prevent catastrophic failures like pantograph destruction and contact wire breakage in railway systems. Existing vision-based methods track wire height and stagger, but they process these measurements without knowing where along the route they occur. This paper proposes spatially aligning pantograph-catenary telemetry to a GPS master grid, partitioning the route into geographic segments, and training a localized bidirectional LSTM-autoencoder per segment to detect collective anomalies. It targets the blind spot of geographic agnosticism in current industrial monitoring, arguing that location-aware models reduce false positives caused by natural topological variation.

LoRA adapters often suffer from adaptation imbalance, where a handful of singular directions hoard the update energy and hurt both target-task specialization and off-task capability retention. This paper proposes LoRA-Norm, a data-free post-training edit that preserves the learned singular directions but rebalances their gains with a fixed hybrid Newton-Schulz transform and per-module nuclear-norm restoration. The authors argue that learning where to adapt does not ensure that adaptation gains are well balanced, and they evaluate the method on 18 adapters spanning two 8B backbones and three tasks.

Time series foundation models post strong scores on large benchmarks, yet their grasp of basic temporal logic remains unclear. This paper introduces SimpleTimeBench, a diagnostic suite of univariate and multivariate unit tests for primitives such as linear trends, periodic signals, and deterministic leading-indicator covariates. The authors show that prominent zero-shot TSFMs—including Chronos-2, Moirai, and Toto—fail on several trivial tasks, and that targeted fine-tuning on the full suite yields only marginal gains, suggesting that pre-training scale alone has not bought robust fundamental reasoning.

This paper studies fixed-confidence best-arm identification for Bernoulli bandits using the top-two $\beta$-EB-TCI rule, where the challenger is selected by a penalized Bernoulli GLR index. Its core contribution is a sharp non-asymptotic analysis: once the true best arm becomes the permanent leader and its sampling fraction stays near $\beta$, the stopping time is $T_\beta^\star(\mu)\log(1/\delta)$ up to lower-order terms. The paper also identifies exactly where the analysis stalls for the unguarded rule when suboptimal means are tied, and closes the gap with a mild $O(\sqrt{Kt})$ exploration guard.

This paper targets few-shot ransomware detection by pairing an Autoencoder Feature Extractor (AFE) with a Model-Agnostic Meta-Learning (MAML) classifier. The idea is to compress 72 tabular behavioral features into a 64-dimensional latent space via the autoencoder, then let MAML adapt a binary classifier from just a handful of labeled malware-versus-benign samples. The authors pitch this as a rapid-deployment defense for healthcare and critical infrastructure settings where new ransomware variants appear with little warning.

Pan and Yang study stochastic nonconvex–strongly-convex bilevel optimization when only unbiased stochastic gradient estimates are available. They propose MRT-FD, a single-loop algorithm that simultaneously tracks the upper-level variable, lower-level solution, and an auxiliary response variable, using order-$p$ finite differences to reconstruct hypergradient actions from first-order queries. For lower-level objectives with order-$p$ smoothness, the method achieves $\mathcal{O}(\varepsilon^{-4-2/p})$ oracle complexity, and the authors prove a matching $\Omega(\varepsilon^{-4-2/p})$ lower bound, settling the optimal $\varepsilon$-dependence for any fixed finite $p$.

Varda-single-1.0 is a deterministic, data-driven weather forecasting system built for the Alpine domain, delivering hourly forecasts at 1 km resolution via a stretched-grid Graph Transformer architecture. It combines a 6-hourly autoregressive forecaster with a temporal downscaler to balance forecast horizon and temporal fidelity. The work demonstrates that machine learning models can match or exceed operational numerical weather prediction baselines in complex topography, and it openly releases model weights and software to enable operational adoption and downstream research.

This paper asks whether deep self-attention networks can exactly interpolate arbitrary finite datasets using only depth, without adapting parameters to the task. The authors prove that universal interpolation is possible with as few as two frozen Gaussian-initialized single-head residual attention blocks, composed with data-dependent signs, durations, and ordering. They also characterize when causal masking prevents interpolation. The results matter because they provide a theoretical foundation for looped and recurrent-depth transformers that share parameters across layers.

STORM is a single-sample stochastic recursive momentum method that avoids periodic full-gradient checkpoints. Prior analyses required a stronger average smoothness assumption and produced suboptimal or logarithmically degraded rates. This paper shows that the original STORM recursion alone achieves the optimal $O(T^{-1/4})$ nonconvex rate, $O(\sigma R/\sqrt{T})$ convex rate, and $O(\sigma^2/(\lambda T))$ strongly convex rate under standard smoothness, by introducing auxiliary sequences and shifted iterates to cancel the estimator bias.

Judgment models that return option probabilities without reasoning text are attractive as fast agent policies, but their competence boundary is unclear. This paper argues that such models excel at evaluation (judging given options against described criteria) yet fail at simulation (predicting missing consequences like opponent moves or physical dynamics). By cleanly separating the two, the authors show that supplying simulated outcomes via code restores strong performance across matrix games, text adventures, and robot control. The result is a sharp diagnostic of where fast judgment models can be trusted and a principled architecture for combining them with external simulators.

Asynchronous LLM post-training mixes stale rollouts from earlier policies, making GRPO-style optimization unstable. This paper derives a convergence bound that exposes a second-moment versus bias tradeoff, then proposes Group-Mass Capping GRPO (GMC-GRPO), which rescales groups when their average importance ratio exceeds a cap. The result is tighter delay dependence than per-trajectory clipping and stronger empirical robustness under large rollout delays on Qwen3 models.

This paper dissects the group attention mechanism in the time series foundation model Chronos-2, which is advertised to serve both multivariate (MV) and in-context learning (ICL) forecasting. The authors decompose attention into its value/output ($V$/$O$) pooling pathway and its query/key ($Q$/$K$) learned weighting pathway by editing the attention matrix $\alpha$ at inference time. Their core finding is that uniform pooling helps nearly all tested configurations, while learned $Q$/$K$ weighting helps MV forecasting but systematically degrades ICL—sometimes leaving forecasts worse than univariate inference. The work matters because it identifies a cheap, retraining-free fix: forcing $\alpha$ uniform in block 0 alone fully recovers ICL performance across every configuration tested.

This paper asks how continuous latent reasoning in chemistry can be improved by training thoughts to anticipate future outcomes without verbalizing every intermediate step. The authors introduce Latent-JEPA, which augments a recurrent latent reasoner with joint-embedding prediction of both textual continuations and molecular structures using pretrained encoders. Evaluated on ChemCoTBench, the framework shows the strongest task gains in molecular optimization and yields latent representations that are more chemically structured.

This paper tackles validation congestion in scientific discovery, where enormous virtual libraries overwhelm scarce experimental capacity. Instead of relying on data-hungry regressors, it proposes PRISMS, a framework that curates candidates using multi-fidelity pairwise rankings aggregated through a regularized Bradley-Terry model, escalating queries to expensive experts via a Fisher-information criterion. The central claim is that comparative judgments from heterogeneous experts can slash the number of slow experimental rounds needed to find top-performing designs.

Selecting the best large language model for a live stream of prompts is difficult because benchmark scores are only coarse proxies for real-world performance and oracle annotations are costly. This paper proposes Online LLM Picker, the first active model selection framework designed specifically for LLMs in streaming settings: it uses an Exponential Weights posterior and an adaptive query probability based on hypothetical response variance scaled by posterior entropy to decide which prompts are worth annotating. Evaluated across over 130 models and 10 datasets spanning question answering, dialogue, grammar correction, and math, the method claims to cut annotation cost by up to 71.67% and reduce regret by up to 2.51× compared to baselines.