Reinforcement learning from verifiable emotion rewards RLVER has produced language models with strong empathetic performance, evaluated on benchmarks that assume cooperative, honest users. Yet real emotional interactions systematically violate this assumption: users gaslight, escalate, and pressure AI systems for unconditional validation, dynamics that cooperative benchmarks cannot surface. We construct the Adversarial Empathy Benchmark AEB and introduce the Emotional Consistency Score ECS to evaluate empathetic robustness under adversarial conditions. AEB comprises six psychologically grounded adversarial trajectory types with discriminative reward structures that penalize formulaic responses; ECS formally disentangles a model's capacity to track user emotional states from its capacity to improve them. In a controlled experiment across eight scenario-matched conditions (think and no-think conditions on 2 RLVER models, and 2 base models (Qwen 1.5B and 7B) with 480 adversarial dialogues), RLVER-PPO-Think substantially outperforms the same-scale untuned baseline (0.963 vs. 0.761, \(p<0.001, r=0.688\)), with zero dialogue collapses and 47\% higher hidden-intention detection. However, ECS remains nearly flat and is not significantly different for RLVER-PPO-Think versus Base-7B-Think (\(p=0.650\)): RL training improves emotional responsiveness without measurable gains in observable state tracking. We interpret the ECS--FS (Final Score) gap as a behavioral/legibility dissociation inside this simulator family, not as evidence about internal understanding or clinical readiness.

Reinforcement learning with verifiable rewards (RLVR) has become a highly effective method for improving the reasoning abilities of Large Language Models (LLMs). Recent research shows that Negative Sample Reinforcement (NSR) -- which focuses on penalizing incorrect steps rather than simply rewarding correct ones -- can match or even exceed the performance of more complex frameworks like PPO and GRPO across the entire Pass@k spectrum. However, current NSR techniques usually apply a fixed penalty throughout the training process and treat every incorrect response with the same weight. To address these limitations, we propose two extensions to the NSR framework: Adaptive Negative Sample Reinforcement. Rather than using a fixed update rule, A-NSR uses time-dependent scheduling functions. In the initial training phases, the system focuses heavily on correcting errors to stabilize the model. As training continues, it shifts toward more subtle and controlled updates. We also introduce Confidence-Weighted Negative Reinforcement, which operates on the principle that different mistakes carry different levels of importance. CW-NSR assigns specific penalty weights based on the model's normalized sequence likelihood. If the model is highly confident in a wrong path, it receives a larger penalty and for uncertain errors -- where the model is effectively exploring -- are penalized less strictly. Our formal analysis shows how these mechanisms govern token-level updates, allowing the model to leverage prior-guided probability redistribution while providing a natural defense against overfitting. We evaluated these methods on difficult reasoning datasets, including MATH, AIME 2025, and AMC23, using the Qwen2.5-Math-1.5B architecture.

Web agents perceive web pages through an observation space, yet its granularity has remained an underexamined design choice. Existing work treats observation at the same element-level granularity as the action space, leaving the page's functional organization implicit and forcing the agent to infer it from element-level signals at every step. We argue observation should instead operate at the granularity of functional regions, parts of the page that each serve a distinct purpose. We propose Region4Web, a framework that reorganizes the AXTree into functional regions through hierarchical decomposition and semantic abstraction, exposing the page's functional organization as the basis for page state understanding. Moreover, we propose PageDigest, a web-specific inference pipeline that delivers this region-level observation to the actor agent as a compact per-page digest that persists across steps. On the WebArena benchmark, PageDigest substantially reduces observation length while improving overall task success rate across diverse backbone large language models (LLMs) and established agent methods, regardless of backbone capacity. These results show that operating at the granularity of functional regions delivers a more compact and informative basis for the actor agent than element-level processing alone.

Graph Anomaly Detection (GAD) is a critical task in graph machine learning with vital applications in financial fraud detection and social platform governance. However, existing GAD benchmarks are often restricted to small-scale, curated graphs with relatively balanced anomaly ratios, leaving a substantial gap between academic evaluation and real-world deployment. To bridge this gap, we present a multi-dimensional benchmark that systematically evaluates GAD models under three deployment-relevant challenges: million-scale graphs, extreme anomaly scarcity, and missing node attributes. We derive a family of controlled benchmark variants from five diverse graphs, including two native industrial-scale datasets with over 3.7 million nodes. Our extensive evaluation of nine representative GAD models reveals three major limitations: (1) most GNN-based methods fail to scale to million-node graphs due to prohibitive memory requirements; (2) detection performance drops sharply under realistic anomaly ratios (e.g., 0.1\%), often resulting in zero recall; and (3) reconstruction-based models are highly sensitive to attribute imputation strategies. Our findings suggest that strong performance in laboratory settings does not guarantee robustness in production environments. We release this benchmark and empirical evaluation as a diagnostic testbed to promote the development of robust and scalable GAD systems for large-scale, imperfect graphs encountered in practice. Code is available at https://anonymous.4open.science/r/Benchmark_GAD-E7A3.

Being robust to the presence of outliers is crucial for applying clustering algorithms in practice. In the $\textit{robust $k$-Means}$ problem (i.e., $k$-Means with outliers), the goal is to remove $z$ outliers and minimize the $k$-Means cost on the remaining points. Despite the close connection between robust $k$-Means and outlier detection, both theoretical and empirical understanding of the effectiveness of $\textit{classic outlier detection heuristics}$ for robust $k$-Means remains limited. In this paper, we prove that under a practical assumption on the optimal cluster sizes, simply removing points with large $K$-Nearest-Neighbor distances achieves performance comparable to prior work in terms of approximation guarantees: it yields a constant-factor reduction from robust $k$-Means to standard $k$-Means, without introducing additional centers or discarding extra outliers, as is commonly required by existing approaches. Empirically, experiments on real-world datasets show that our method outperforms or matches several more sophisticated algorithms in terms of clustering cost and runtime. These results demonstrate that simple KNN-based heuristics can be surprisingly effective for robust clustering, highlighting new opportunities to bridge techniques from outlier detection and clustering.

Modern large language models (LLMs) can find a needle in a haystack (locating a single relevant fact buried among hundreds of thousands of irrelevant tokens) with near-saturated accuracy, yet fail to retrieve the last few items in a short list. We call this failure the Position Curse. For instance, even in a two-line code snippet, Claude Opus 4.6 misidentifies the second-to-last line most of the time. To characterize this failure, we evaluated two complementary queries: given a position in a sequence (of letters or words), retrieve the corresponding item; and given an item, return its position. Each position is specified as a forward or backward offset from an anchor, either an endpoint of the list (its start or end) or another item in the list. Across both open-source and frontier closed-source models, backward retrieval substantially lags forward retrieval. To test whether this capability can be rescued by post-training, we constructed PosBench, a position-focused training dataset. LoRA fine-tuning improves both forward and backward retrieval and generalizes to a held-out code-understanding benchmark (PyIndex), yet absolute performance remains far from saturated. As LLM coding agents increasingly operate over large codebases where precise indexing becomes essential for code understanding and editing, position-based retrieval emerges as a key capability for future pretraining objectives and model design.

In-context reinforcement learning (ICRL) refers to the ability of RL agents to adapt to new tasks at inference time without parameter updates by conditioning on additional context. Recent empirical studies further demonstrate that Chain-of-Thought (CoT) generation can amplify this ICRL capability. This paper is the first to provide a theoretical understanding on how CoT interacts with ICRL. We conduct our analysis in a policy evaluation setup with linear Transformer. We prove that with specific Transformer parameters, the CoT generation process is equivalent to repeatedly executing temporal difference learning updates. Additionally, we provide finite sample convergence analysis showing that the policy evaluation error decreases geometrically with CoT length and eventually saturates at a statistical floor determined by the context length. We also prove that the desired Transformer parameters are a global minimizer of the pretraining loss, providing a theoretical understanding on the empirical emergence of those parameters.

Template tasks have emerged as a clean testbed for asking whether transformers reason with abstract symbols rather than concrete token names. We study the fixed-label classification version of this problem, where train and test examples share latent templates but may use disjoint vocabularies. Unlike next-token prediction, the model need not emit unseen symbols; it must learn a decision rule invariant to symbol renaming. We analyze regularized kernel logistic classification in the transformer-kernel regime. Our main result decomposes the learned predictor into an ideal template-level classifier and a finite-sample perturbation caused by accidental token overlaps in the training data. We encode these overlaps by a colored collision graph and prove high-probability margin-transfer guarantees for fresh-symbol classification. This perspective extends template-based analyses to logistic classification and refines scalar diversity conditions: vocabulary size controls the average rate of collisions, but collision geometry controls whether the ideal classification margin is preserved. More broadly, the same perturbation framework applies to abstraction-augmented inputs, yielding a general margin-versus-collision criterion for identifying when prompting strategies improve fresh-symbol generalization. Synthetic template experiments illustrate the predicted roles of regularization, sample size, and transformer-kernel structure.

Physics-informed neural solvers offer a promising route to model-based reinforcement learning in continuous time, where optimal feedback synthesis is governed by Hamilton--Jacobi--Bellman (HJB) equations. Practical implementations often occupy a regime that is neither a classical grid method nor a continuous-PDE PINN: the value function is represented by a neural network, finite-difference HJB policy-evaluation operators are evaluated by network queries at shifted points, and residuals are minimized by random continuous collocation. This regime preserves the stabilized finite-difference policy-evaluation structure while avoiding grid-based value unknowns. We develop an error theory for this hybrid regime. Interpreting finite differences as shift operators acting on neural networks, we prove a population $L^2$ stability estimate for one policy-evaluation step with learned dynamics. The bound separates residual error, initial and exterior-collar mismatch, policy mismatch, and model-identification error, with an explicit gradient amplification factor for learned dynamics, while the underlying linear evaluation stability remains free of hidden inverse-viscosity blow-up. We further give a finite-sample collocation certificate and a conditional multi-step propagation result through greedy policy improvement. Experiments on compact-control LQR upto 64 dimensions, Allen--Cahn control, pendulum, Hopper, and 3D quadrotor benchmarks compare against representative model-based and model-free RL baselines, demonstrating the predicted residual, policy-mismatch, and learned-model error trends.

Stochastic bandit algorithms are usually analyzed under a mean-reward criterion, yet many problems favor arms with strong upper-tail performance, which we study herein. For a fixed miscoverage level \(α\), the natural upper-tail target of arm \(j\) is the upper endpoint \(F_j^{-1}(1-α/2)\) of a central prediction interval. This target can rank arms differently from their means, creating a central mismatch with the classical bandit objective. To this end, we propose ACP-UCB1, a conformal-style policy that combines an adaptive conformal estimate of the upper endpoint with a UCB-type optimism bonus. The technical challenge is that the conformity scores used by ACP-UCB1 are recomputed from evolving empirical quantile estimates and evaluated at an adaptive level. We control this endpoint through reward-quantile concentration, a perturbation argument for recomputed score quantiles, and deterministic localization of the adaptive level. ACP-UCB1 achieves logarithmic upper-quantile regret with per-arm contribution \(O(\nicefrac{\log n}{Δ_j^{\mathrm{ACP}}})\). We also provide metric-specific regret decompositions comparing ACP-UCB1 with UCB1 and use numerical experiments to validate performance and improvement.

Reinforcement learning with verifiable rewards (RLVR) has emerged as a central paradigm for improving the reasoning capabilities of large language models. Group-based policy optimization methods, such as GRPO, typically allocate a fixed number of rollouts to every prompt. This uniform allocation can be inefficient: it over-allocates compute to prompts whose sampled groups are already saturated while under-exploring prompts for which additional samples may reveal useful correct trajectories. To address this limitation, we introduce hit utility, the posterior probability that at least one rollout in a proposed additional allocation for a prompt will be correct. Building on this notion, we propose Hit-Utility Optimal Rollout Allocation (HORA), a learning-free rollout allocation policy that maximizes total posterior hit utility within each allocation batch. HORA adaptively reallocates rollout budgets while leaving the downstream reward evaluation and group-based advantage estimator unchanged. Across four mathematical reasoning benchmarks and three model scales, HORA preserves comparable Pass@1 and improves Pass@K over compute-matched GRPO in ten of twelve model--benchmark configurations, with one tie and one saturated exception. It is also drop-in compatible with other group-based estimators such as RLOO. Ablation studies indicate that the uniform prior used by HORA is competitive with five prompt-conditioned learned-prior alternatives.

Exact solution of hard combinatorial optimization problems often relies on strong convex relaxations, but solving these relaxations repeatedly inside a branch-and-bound algorithm can be prohibitively expensive. Hence, we consider this challenge for Max-Cut, where branch and bound commonly uses semidefinite programming (SDP) relaxations to bound subproblems. We propose a Max-Cut-specific graph neural network that serves as a principled, lightweight neural proxy for these SDP solvers and can be plugged directly into an exact branch-and-bound framework. The proposed architecture has update steps of complexity $\mathcal{O}(n^2 + ne)$, and predicts both primal- and dual-feasible SDP solutions. The primal SDP solutions yield feasible Max-Cut solutions via the Goemans--Williamson algorithm. In addition, it is trained in a self-supervised fashion without requiring solved SDP relaxations as labels. Empirically, we show that our architecture can substantially reduce the cost of bounding in exact Max-Cut solving by up to $10.6 \times$ compared with using the state-of-the-art SDP solver Mosek. Our work highlights the potential of learned, validity-preserving surrogates for accelerating exact optimization over structured convex relaxations.

Agentic AI systems that invoke external tools are powerful but costly, leading developers to default to large models and overspend inference budgets. Model routing can mitigate this, but existing routers are designed for chat completion rather than tool use. We present Switchcraft, the first (to the best of our knowledge) model router optimized for agentic tool calling. Switchcraft operates inline, selecting the lowest-cost model subject to correctness. We construct an evaluation framework on five function-calling benchmarks and train a DistilBERT-based classifier, deployed under a latency budget. Switchcraft achieves 82.9% accuracy -- matching or exceeding the best individual model -- while reducing inference cost by 84%, saving over $3,600 per million queries. We find that larger models do not consistently outperform smaller ones on tool-use tasks, and that nominally cheaper models can incur higher total cost due to token-intensive reasoning. Our work enables cost-aware agentic AI deployment without sacrificing correctness.

Computer-use agents(CUAs)are moving frombounded benchmarks toward real software environments, wherethey operate browsers, desktops, mobile applications, flesystems,terminals, and tool backends. In such settings, reliability isno longer captured by task success alone: perception errors,planning drift, memory use, tool mediation, permission scope,and runtime oversight jointly determine whether agent actionsremain aligned with user intent, Existing surveys organize theCUA landscape by methods, platforms, benchmarks, or securitythreats, but less explicitly connect capability formation, author-ity exposure, failure manifestation, and control placement. Toaddress this gap, the article develops an architecture-lifecycleframework for deployment-grounded reliability in CUAs. Thearchitectural view analyzes Perception, Decision, and Executionas coupled layers that transform software observations intoauthority-bearing actions, The lifecycle view examines Creation.Deployment, Operation, and Maintenance as stages in which priorsare learned, tools and permissions are bound, runtime trajecto.ries are stressed, and assurance must be preserved under drift.Using this lens, the analysis synthesizes representative systems,benchmarks, and security/privacy studies; distinguishes wherefailures become visible from where their enabling conditions areintroduced, and maps recurring intervention surfaces for controloversight, and assurance. OpenClaw is used only as a public moti.vating example of an open deployment pattern, not as a verifedinternal case study. The conclusion highlights open challengesin controllable grounding, long-horizon constraint preservation,safe authority binding, mixed-trust runtime defense, privacy-preserving memory,and continual assurance.

Multimodal Large Language Models (MLLMs) have made remarkable progress on vision-language reasoning, yet most methods still compress visual evidence into discrete textual thoughts, creating an information bottleneck for fine-grained perception. Recent latent visual reasoning methods attempt to reason in continuous hidden states, but we find that they suffer from insufficient manifold compatibility: latent trajectories drift away from pretrained reasoning circuits, collapse into instance-agnostic patterns, and are often bypassed during answer generation. To address these issues, we propose RIS (Retrieve, Integrate, and Synthesize), a spatial-semantic grounded framework that develops latent reasoning as a compatible extension of pretrained MLLM computation. We first construct a step-wise grounded reasoning dataset with bounding boxes and region-specific semantic descriptions. Built on this supervision, RIS anchors latent tokens to both spatial and semantic evidence, enforces their causal role through a progressive attention bottleneck, and introduces short language transition tokens to bridge synthesized latent states back to vocabulary-aligned decoding. Experiments on V*, HRBench4K, HRBench8K, MMVP, and BLINK show consistent improvements over closed/open-source and latent reasoning baselines. Further analyses demonstrate that RIS learns diverse, interpretable, and progressively integrated latent trajectories, offering a practical path toward faithful internal visual reasoning in MLLMs.

Language model (LM) alignment improves model outputs to reflect human preferences while preserving the capabilities of the base model. The most common alignment approaches are (i) reinforcement learning, which maximizes the expected reward under a KL-divergence constraint, and (ii) best-of-$N$ alignment, which selects the highest-reward output among $N$ independent samples. Despite their widespread use, the fundamental limits of reward improvement under a KL budget remain poorly understood. We characterize the information-theoretic limits of KL-regularized alignment by deriving the maximum achievable expected reward gain for a fixed KL-divergence budget. Our first result provides a closed-form expression for the optimal reward improvement, governed by a Jeffreys divergence term rather than the $\sqrt{\texttt{KL}}$ used in prior analyses. We further reformulate this expression as a covariance under the base model, yielding a practical estimator that predicts achievable alignment gains from base model samples alone. We extend our analysis to the proxy reward setting, showing that the gap between ideal and proxy alignment (reward hacking) grows with the magnitude of reward error and when the KL penalty factor decreases. We then prove that reward ensembling mitigates reward hacking, providing a theoretical justification for this technique used in practice. Empirically, we compute the KL-reward Pareto frontier for two tasks for LMs, safety and summarization, and show that best-of-$N$ closely approaches the theoretical limit, while PPO and GRPO remain substantially suboptimal. Our theoretical results shed light on several empirically observed phenomena in the alignment literature and suggest that algorithmic improvements are needed to achieve optimal alignment without high inference costs.

Establishing almost sure convergence rates for stochastic approximation and reinforcement learning under Markovian noise is a fundamental theoretical challenge. We make progress towards this challenge for a class of stochastic approximation algorithms whose expected updates are contractive, a setting that arises in many reinforcement learning algorithms such as $Q$-learning and linear temporal difference learning. Specifically, for a power-law learning rate $O(n^{-η})$ with $η\in (1/2, 1)$, we obtain an almost sure convergence rate arbitrarily close to $o(n^{1 - 2η})$. For a harmonic learning rate $O(n^{-1})$, we obtain an almost sure convergence rate arbitrarily close to $o(n^{-1})$, which we argue is a strong result because it is close to the optimal rate $O(n^{-1}\log\log n)$ given by the law of the iterated logarithm (for a special case of i.i.d. noise). Key to our analysis is a novel Lyapunov drift construction that applies a Poisson-equation based correction for Markovian noise to the well-established Moreau-envelope smoothing for the contractive mapping.

Reaction feasibility prediction, as a fundamental problem in computational chemistry, has benefited from diverse tools enabled by recent advances in artificial intelligence, particularly large language models. However, the performance of individual tools varies substantially across reactions, making it difficult for any single tool to consistently perform well across all cases. This raises a critical challenge: how to effectively leverage multiple tools to obtain more accurate feasibility predictions. To address this, we propose ARMOR, an agentic framework that explicitly models tool-specific utilities, adaptively prioritizes tools, and further resolves the potential tool conflicts to produce the final prediction for each reaction. Unlike existing approaches that rely on simple aggregation or heuristic assignment over various tools, ARMOR organizes tools into a hierarchy that prioritizes top-performing tools and defers others when needed, characterizes their strengths through tool-specific patterns, and resolves conflicts via memoryaugmented reasoning. Extensive experiments on a public dataset demonstrate that ARMOR consistently outperforms strong baselines, including single-tool methods as well as various tool aggregation and tool selection approaches. Further analysis shows that the improvements are particularly significant on reactions with conflicting tool predictions, highlighting the effectiveness of ARMOR in leveraging the complementary strengths of multiple tools. The code is available via https://anonymous.4open.science/r/ARMOR-E13F.

Evaluating literary quality requires assessing interpretive dimensions such as cultural representation, emotional depth, and philosophical sophistication that resist straightforward computational measurement. We introduce SAGE, a hierarchical evaluation framework that decomposes literary quality into ontology-grounded interpretive dimensions assessed through structured large language model evaluation with multi-round iterative reflection and independent validation. We validate the framework on 100 short stories (50 canonical works, 30 pulp fiction, 20 LLM-generated narratives) across three analytical layers (cultural, emotional-psychological, existential-philosophical) using dual-mode assessment. Across 600 evaluations, the framework achieves 98.8% score convergence and greater than 94% inter-rater agreement, with near-perfect mode invariance between content-based and metadata-based evaluation. Statistical analysis reveals a consistent genre hierarchy (Canonical > Pulp > LLM, all p<0.001) with layer-specific discrimination: cultural critique and philosophical depth exhibit very large effect sizes (Cohen's d>2.4), while emotional representation shows smaller gaps (d=1.68), suggesting that affective patterns are more learnable from training data than critical stance or philosophical depth. Cross-layer correlations (r=0.649-0.683) confirm the three dimensions capture empirically distinguishable quality facets. These findings demonstrate that theory-driven LLM evaluation can achieve measurement-grade reliability and support systematic identification of where current generative models fall short of human literary production, with direct implications for scalable automated evaluation of open-ended text generation.

This paper introduces the Active-Importance-Sampling Actor-Critic (AISAC) algorithm, an extension of the Actor-Critic framework for reducing variance in policy gradient estimation. AISAC optimizes the behavior policy to minimize gradient variance while preserving unbiased gradient estimates. Using importance sampling principles, the algorithm adapts the behavior policy toward efficient data collection distributions aligned with target policy gradients. For continuous action spaces, AISAC employs Gaussian behavior policies optimized through cross-entropy minimization. We provide theoretical analysis demonstrating variance reduction and unbiasedness. Experiments on Inverted Pendulum and Half Cheetah tasks show improved learning speed, sample efficiency, and training stability compared to standard Actor-Critic methods. Results indicate that optimizing the behavior policy improves both target policy updates and critic estimation accuracy across different hyperparameter settings. AISAC accelerates convergence and stabilizes reinforcement learning training, making it promising for real-world applications. Future work includes integration with advanced algorithms such as Soft Actor-Critic and TD3 for more complex environments.

The Translation Tax is often treated as a scalar: translated benchmarks are assumed to inflate scores by preserving English-source cues. We audit this claim in an English-to-Chinese setting. Three proxy estimators disagree: back-translation gaps are small and parser-fragile; cue-score calibration does not predict item-level gains; and a six-model native-control comparison shows model-family rather than uniform benchmark effects. We add a same-item LLM-naturalization stress test that holds answer, options, and content fixed while rewriting Chinese surface form. After correcting a prompt-construction bug, this contrast no longer supports a model-family interaction, but it preserves a residue dose-response: high-residue items benefit while low-residue items do not. The result is not a single Translation Tax, but a set of estimator- and item-dependent validity risks. We release per-cell evidence, the naturalization protocol, human QC, and a reporting checklist for translated multilingual benchmark papers.

Channel importance in vision networks is usually summarized by a single score. That summary hides two different questions: how much a channel is related to the task, and whether its function can be supplied by same-layer peers when the channel is removed. We call the second property local replaceability. We introduce a two-axis view that separates these questions. The local axis measures input capture and peer overlap, while the target axis measures task information and target-excess information. Across ResNet-18, VGG-16, and MobileNetV2 trained on CIFAR-100, the two axes are weakly aligned, induce different channel groupings, and separate rapidly during training despite being strongly coupled at random initialization. A Gaussian linear analysis accounts for how this separation can arise through residualized gradient directions, and lesion plus peer-replacement experiments show that peer support refines removability beyond input capture and task relevance alone. Under the fixed FLOPs-matched pruning protocol, local-axis metrics are more reliable predictors of removability than target-axis metrics across the three CIFAR-100 backbones, with the same direction preserved in stress tests on CIFAR-10, Tiny-ImageNet, ImageNet-100, and a ConvNeXt-T/ImageNet-100 pilot. These findings identify an axis-level distinction rather than a universal ranking of pruning scores: local replaceability is a more reliable guide to removability than target relevance, while norm-based baselines remain competitive in architectures such as VGG-16. Relevance-based scores ask what a channel says about the task; pruning asks whether the network still needs that channel when its peers remain available.

Automatic speech recognition (ASR) evaluation compares system output to ground truth transcripts, with Word Error Rate (WER) quantifying the distance between them. But ground truth transcripts are not discovered - they are produced by human annotators following conventions that encode normative assumptions about which speech features matter. Different conventions (verbatim, non-verbatim, legal) produce different transcripts of identical speech and judge the same ASR output differently. This paper argues that reference monism - enforcing a single transcription convention as ground truth - commits epistemic injustice. Speakers with aphasia, whose speech includes clinically meaningful disfluencies, are systematically disadvantaged when evaluated against "clean" references that treat those disfluencies as errors. The harm is not merely differential performance, but that evaluative infrastructure lacks interpretive resources to recognize their contributions as legitimate. We develop a philosophical framework introducing the hermeneutical gap, formalize Epistemic Injustice Distance (EID) to measure reference monism's cost, and demonstrate empirically using AphasiaBank that WER varies depending on which convention defines ground truth. We propose WER-Range: reporting performance across legitimate conventions rather than assuming a single correct answer.

In the design of surgical guides for implant placement, determining the precise implant position is a critical step. However, the implant region itself is often characterized by a lack of distinctive texture in medical images. Consequently, artificial intelligence (AI) models must infer the correct implant position and angulation (slope) primarily by analyzing the texture of the surrounding teeth, which poses a significant challenge. To address this, we propose ImplantMamba, a network architecture designed for long-range sequential modeling to integrate texture information from adjacent teeth. Our approach explicitly couples the regression of the implant position with its slope. The core of ImplantMamba is a hybrid encoder that combines Convolutional Neural Networks (CNNs) with Mamba layers. This design enables the network to hierarchically extract local anatomical features through CNNs while simultaneously modeling global contextual dependencies across the entire scan volume via Mamba's selective scan operations, leading to a more comprehensive understanding of the implant site. Furthermore, we introduce a Slope-Coupled Prediction Branch (SCP). This branch is designed to connect the prediction of implant position with the slope, ensuring internal consistency and anatomical plausibility by thereby enforcing a coherent relationship between the predicted implant location and its angulation. Extensive experiments on a large-scale dental implant dataset demonstrate that the proposed ImplantMamba achieves superior performance compared to existing methods.

Many real-world resource allocation systems, such as humanitarian logistics and vaccine distribution, must preposition limited supply across multiple locations before demand is realized while stockouts incur irreversible service losses. To study this, we introduce the Online Shared Supply Allocation (OSSA) problem, a stateful online model in which a central hub allocates a finite, unknown supply to multiple sites facing sequential demand under fixed-charge transportation costs and lost-sales penalties. Unlike classical make-to-stock or make-to-order inventory models, OSSA precludes backlogging and replenishment only hedges against future demand. To tackle OSSA, we propose a deterministic threshold-proportional policy GPA and prove that it achieves a $4/3$-approximation to the offline optimum up to an additive term independent of the total supply. We complement this with matching lower bounds showing that the $4/3$ ratio is tight and that the additive-error dependence is unavoidable, even for randomized algorithms that know the total supply upfront. Finally, we develop a learning-augmented extension to GPA that principally incorporates imperfect forecasts (e.g., from human experts or ML models) commonly available in practice, enabling us to exploit high-quality advice while being robust against arbitrary bad ones. Synthetic and real-world experiments show that GPA outperforms natural baselines with global supply is scarce.

World models predict future transitions from observations and actions. Existing works predominantly focus on image generation only. Visual feature-based world models, on the other hand, predict future visual features instead of raw video pixels, offering a promising alternative that is more efficient and less prone to hallucination. However, current feature-based approaches rely on direct regression, which leads to blurry or collapsed predictions in complex interactions, while generative modeling in high-dimensional feature spaces still remains challenging. In this work, we discover that a new type of latent action representation, which we refer to as *Residual Latent Action* (RLA), can be easily learned from DINO residuals. We also show that RLA is predictive, generalizable, and encodes temporal progression. Building on RLA, we propose *RLA World Model* (RLA-WM), which predicts RLA values via flow matching. RLA-WM outperforms both state-of-the-art feature-based and video-diffusion world models on simulation and real-world datasets, while being orders of magnitude faster than video diffusion. Furthermore, we develop two robot learning techniques that use RLA-WM to improve policy learning. The first one is a minimalist world action model with RLA that learns from actionless demonstration videos. The second one is the first visual RL framework trained entirely inside a world model learned from offline videos only, using a video-aligned reward and no online interactions or handcrafted rewards. Project page: https://mlzxy.github.io/rla-wm

Compositional generalization requires models to produce novel configurations from familiar parts. In diffusion models, prior compositional generation methods typically assume that the relevant concepts or conditioning signals are already available. We instead ask whether a pretrained diffusion model can discover query-specific concepts from the time-indexed scores it learns for the noisy marginals $p_t(x_t)$ and compose them at test time. Given a single out-of-distribution query, our method performs gradient ascent on $s_θ(x_t,t) \approx \nabla_{x_t}\log p_t(x_t)$ at multiple noising timesteps to recover local density modes, maps these modes into clean-space Gaussians, greedily selects relevant prototypes with a submodular likelihood objective, and combines them into a product-of-experts (PoE) teacher model with an analytic score. This teacher model can be sampled directly through classifier-free guidance or used to generate a sample pool for training a new class embedding and low-rank adapter. On held-out composition benchmarks built from ColorMNIST and CelebA, both the analytic PoE sampler and the low-rank adapted model outperform query-only and nearest trained-class baselines. These results suggest that the time-indexed score geometry of the diffusion model contains reusable density-mode concepts that support test-time compositional generation without a predefined concept library.

The open-source model ecosystem now contains hundreds of thousands of pretrained models, yet picking the best model for a new dataset is increasingly infeasible: new models and unbenchmarked datasets emerge continuously, leaving practitioners with no prior records on either side. Existing approaches handle only fragments of this in-the-wild setting: AutoML and transferability estimation select models from small predefined pools or require expensive per-model forward passes on the target dataset, while model routing presupposes a given candidate pool. We introduce ModelLens, a unified framework for model recommendation in the wild. Our key insight is that public leaderboard interactions, though scattered and noisy, collectively trace out an implicit atlas of model capabilities across heterogeneous evaluation settings, a signal rich enough to learn from directly. By learning a performance-aware latent space over model--dataset--metric tuples, ModelLens ranks unseen models on unseen datasets without running candidates on the target dataset. On a new benchmark of 1.62M evaluation records spanning 47K models and 9.6K datasets, ModelLens surpasses baselines that either rely on metadata alone or require running each candidate on the target dataset. Its recommended Top-K pools further improve multiple representative routing methods by up to 81% across diverse QA benchmarks. Case studies on recently released benchmarks further confirm generalization to both text and vision-language tasks.

Agent systems often decompose a task across multiple roles, but these roles are typically specified by prompts rather than enforced by access controls. Without enforcement, a team pass rate can mask whether agents actually coordinated or whether one role effectively did another role's work. We present TeamBench, a benchmark with 851 task templates and 931 seeded instances for evaluating agent coordination under operating system-enforced role separation. TeamBench separates specification access, workspace editing, and final certification across Planner, Executor, and Verifier roles, so that no role can read the full requirements, modify the workspace, and certify the final answer. Prompt-only and sandbox-enforced teams reach statistically indistinguishable pass rates, but prompt-only runs produce 3.6 times more cases where the verifier attempts to edit the executor's code. Verifiers approve 49% of submissions that fail the deterministic grader, and removing the verifier improves mean partial score in the ablation. Team value is also conditional. Teams benefit when single agents struggle, but hurt when single agents already perform well. A 40-session human study under the same role separation shows that our benchmark exposes interaction patterns that pass rate misses. Solo participants work through the task directly, human participants paired with agents often collapse into quick approval, and human teams spend more effort coordinating missing information across roles.

Poisson subsampling is the default sampling scheme in differentially private machine learning, largely because its unstructured randomness yields tractable privacy amplification analyses. Yet this same randomness introduces substantial participation variance: each sample appears in very different numbers of training iterations. In this work, we show that this variance is not merely a practical artifact to be tolerated, but a fundamental source of suboptimal privacy amplification. We prove that Balanced Iteration Subsampling (BIS), a structured scheme in which each sample participates in exactly a fixed number of iterations, achieves stronger privacy amplification than Poisson subsampling and is optimal at both extremes of the noise spectrum ($σ\to 0$ and $σ\to \infty$). Our analysis reveals that the privacy-noise tradeoff is governed not by maximizing randomness, but by eliminating participation variance while preserving uniform marginal participation across iterations. To translate this asymptotic theory into finite-noise guarantees, we introduce a practical near-exact Monte Carlo accountant for BIS, which removes the analytical slack of existing RDP and composition-based PLD analyses. Evaluations across more than 60 practical DP-SGD configurations show that BIS consistently outperforms Poisson subsampling in the low-noise regimes most relevant for high-utility private training, reducing the required noise multiplier by up to $9.6\%$. These results overturn the common intuition that more sampling randomness necessarily yields stronger privacy amplification: in DP-SGD, structured participation can be both more practical and more private. Our implementation is available at https://github.com/dong-xin-ao-andy/bis-mc-accountant.