arXiv:2605.25143v2 Announce Type: replace
Abstract: Test-time scaling improves language model reasoning by spending additional compute to explore multiple solution trajectories. The key challenge is to maximize accuracy while minimizing the total number of generated tokens during reasoning. Recent PRM-guided methods score intermediate prefixes to steer this search, but most are frontier-only: they keep only the current active prefixes and irreversibly prune or resample away the rest using noisy PRM scores. This can cause premature commitment, diversity collapse, and the loss of prefixes that still admit correct continuations. We introduce stochastic backtracking over a persistent pool of historical prefixes, allowing test-time compute to revisit previously generated states instead of only expanding the current frontier. To make this efficient, we propose two complementary mechanisms. Subpool Selection strengthens greedy PRM-guided search by applying Top-N selection within random subpools, giving historical prefixes a chance to bypass over-scored frontier candidates. Power Backtrack Sequential Monte Carlo extends SMC-style resampling to the persistent pool using powered PRM scores and mixture-corrected weights. Across mathematical reasoning benchmarks and model scales, our methods consistently achieve higher accuracy per token count, and the same level of accuracy using only a fraction of the token count in comparison to strong PRM-guided baselines, demonstrating that persistent-pool stochastic backtracking provides a simple and effective way to improve the accuracy-token trade-off in test-time scaling.
Science Journals
arXiv:2606.01165v1 Announce Type: cross
Abstract: Ramesh's 2023 dissertation introduces the categorical notions of introspective theories and geminal categories, which formalize "self-internalizing" structures sharing the form of L\"ob's theorem ($\Box A \vdash A$ implies $\vdash A$). We reorganize the theory of geminal categories in a self-contained manner by introducing "code structures on fibrations," which serve as a categorical abstraction of G\"odel coding. This framework leads to a significant simplification of the proof of L\"ob's theorem for geminal categories, as well as to a new categorical counterpart of the G\"odel-L\"ob axiom ($\Box(\Box A \to A) \to \Box A$). This formulation offers an accessible framework for Ramesh's approach and suggests connections to modal type theories, where similar meta- and object-level interactions arise.
arXiv:2605.29365v3 Announce Type: replace
Abstract: Formality transfer is commonly framed as a symmetric bidirectional task between informal and formal registers. We argue that this framing conceals a supervision design flaw in existing benchmarks such as GYAFC: binary human rewrites encode relative stylistic shifts rather than absolute human notions of formality. Consequently, models learn to generate pseudo-formal outputs that satisfy benchmark labels while failing to produce genuinely formal language. We quantify this misalignment by re-evaluating benchmark formal labels under a human-aligned definition of formality, revealing substantial discrepancies that propagate to consistent informal-to-formal failures across model families. To address this issue, we reconceptualize formality transfer as a graded dimension rather than a binary attribute. We introduce a three-level spectrum: informal, casual, and formal, where casual serves as an explicit intermediate state that clarifies supervision signals. Based on this framework, we introduce 3LF, a dataset providing parallel supervision across all three levels. Training on 3LF substantially reduces informal-to-formal failures and improves alignment with human perception. For example, GPT-4.1-nano improves from 0.06 to 0.88 F1 in the informal-to-formal direction despite 3LF being significantly smaller than GYAFC. We further demonstrate that these gains cannot be reproduced through in-context learning alone and provide qualitative analyses of ambiguity-driven errors and meaning distortions. Overall, our findings demonstrate how supervision design shapes stylistic alignment and highlight the importance of alignment-aware benchmark construction in controllable text generation.
arXiv:2605.29463v2 Announce Type: replace
Abstract: Reflexion-style agents rely on self-generated reflections as memory, implicitly assuming that agents can accurately diagnose their own failures. We show that this assumption can fail systematically: across ALFWorld and HumanEval, agents store confident but incorrect interpretations of the task and continue acting on them across trials, even though the environment resets to the correct task each time. We call this failure mode memory confabulation and introduce the Reflection Repetition Rate (RRR), a log-based metric that detects repeated reliance on incorrect reflective content. Using RRR, we identify 16 frozen environments in ALFWorld, where 0 of 121 reflections mention the correct target object, and 4 analogous cases in HumanEval. Our mitigation replaces open-ended self-diagnosis with programmatic extraction of trajectory-level failure signals, increasing correct object mention from 0% to 86%, reducing RRR from 0.64 to 0.10, and solving 3 of 16 frozen ALFWorld environments, suggesting that reflective memory can reinforce false beliefs rather than correct them.
arXiv:2605.29477v2 Announce Type: replace
Abstract: Recently, the runtime analysis of multi-valued estimation-of-distribution algorithms in the framework of Ben Jedidia et al. (TCS 2024) has made significant advancements. However, almost all existing analyses are limited to multi-valued objective functions that in each dimension only distinguish between two types, also called categories, of values and hence can be treated with similar methods as pseudo-Boolean problems. Only recently, Adak and Witt (GECCO 2025) have presented a first runtime analysis of a multi-valued compact genetic algorithm (cGA) on the multi-valued OneMax function G-OneMax$\colon \{0,\dots,r-1\}^n \to \mathbf{N}$ defined by G-OneMax$(x_1,\dots,x_n)=\sum_{i=1}^n {x}_i$ and truly depending on all $r$ categories. We improve their runtime result from $\textrm{O}\bigl(n r^3 \log^2( n)\log (r)\bigr)$ to $\textrm{O}\bigl(n r \log^3(n)\log^3(r)\bigr)$, both for an optimal choice of the update strength $K$. Our result matches, up to polylogarithmic factors, the existing bound for the simpler $r$-valued OneMax function depending essentially only on two values and analyzed in several previous works. To show the new bound, we use improved drift theorems for processes with high self-loop probabilities and specifically derived concentration inequalities to analyze how probability mass in the multi-valued cGA moves into successively smaller and smaller intervals of the $r$-valued frequency matrix.
arXiv:2605.29488v2 Announce Type: replace
Abstract: Conditional human motion generation remains a fundamental challenge in computer vision and robotics. Despite significant progress, current methods are often constrained by fixed modality configurations and task-specific architectures, leaving cross-modal interactions and the scaling laws of multimodal-conditioned synthesis largely underexplored. A key bottleneck is the scarcity of large-scale modality-aligned motion data, limiting generalization across diverse control signals. In this work, we introduce OmniHuMo, a large-scale, high-quality dataset comprising over 5,000 hours of motion and 3.2 million sequences with precisely aligned multimodal annotations (e.g., text, speech, music, and trajectory). Leveraging OmniHuMo, we propose AnyMo, a unified multimodal framework combining a Residual FSQ-based motion tokenizer with a scalable masked modeling transformer, enabling high-quality motion synthesis under arbitrary modality combinations. Extensive experiments show that AnyMo achieves high-fidelity synthesis while offering flexible control over both spatial and stylistic attributes.
arXiv:2605.29548v2 Announce Type: replace
Abstract: Larger models learn tasks smaller models do not. What drives this phenomenon? We develop a simple phenomenological argument that power-law scaling already suggests that a larger model will be able to learn a part of the data distribution that a smaller model fails to learn, even with infinite training data. To validate this claim and identify its causes, we study the effects of model scaling on a synthetic setup consisting of a mixture of tasks that show monotonic scaling curves. The results point to a data-induced competition over resources (neurons). Specifically, smaller models allocate their neurons to high frequency or low complexity tasks, and so they learn solutions that perform poorly on rare and complex tasks. Moreover, this happens even when solutions capable of expressing the desired task exist. We then assess how a larger model circumvents this data-centric bottleneck, finding that it traces to a reduced interference mechanism: larger models can allocate enough resources to common tasks that the gradient updates for those tasks become weak, which means that they do not overwrite rare-task features as they slowly accumulate. Finally, to further validate these claims, we pretrain OLMo models (4M to 4B parameters) on novel tasks of varying frequency and complexity. The results mirror those from our synthetic data experiments: only the larger OLMo models learn the infrequent and complex tasks, and these larger models embed more task features in their representations and show less gradient interference between tasks. Overall, we offer a data-centric account of why larger models learn tasks that smaller models fail to. This helps explain why larger models are better in practice, and it can inform practical questions concerning model sizing and training data mixtures.
arXiv:2605.29948v2 Announce Type: replace
Abstract: Unified speech foundation models require a holistic tokenization space that is both learnable by language models and decodable into high-quality waveforms. Existing speech tokenizers, however, often fail to satisfy these requirements simultaneously, leading to increased architectural complexity and more involved training designs. We propose HoliTok, a continuous Holistic speech Tokenization model designed for unified generation-understanding modeling. HoliTok encodes 48~kHz speech into a compact 25~Hz sequence of 128-dimensional latents. It is trained with a progressive strategy that jointly preserves signal-level fidelity, incorporates semantic information, and maintains strong latent learnability. Based on this tokenization, we build a unified AR+DiT model for speech synthesis and recognition, where the same latent sequence supports both generation-specific and unified generation-understanding tasks. Experiments show that HoliTok achieves competitive reconstruction fidelity, improves generative learnability for high-quality and controllable synthesis, and, among the evaluated representations, is the only one that operates robustly in our unified generation-understanding architecture without additional optimization tricks. These results suggest that HoliTok serves as an effective speech tokenizer and a foundational representation interface for unified spoken language modeling. The code is available at: https://github.com/bovod-sjtu/HoliTok.
arXiv:2605.29973v2 Announce Type: replace
Abstract: Robot behavior is often validated through simulation-based testing, yet the replicability of such campaigns depends critically on transparent documentation of how tests are configured, executed, and post-processed. We argue that data provenance, coupled with the FAIR principles (findability, accessibility, interoperability, and reusability), addresses this gap by explicitly tracking links between artifacts and by attaching machine-readable metadata about file origins and key design decisions. Moreover, provenance and metadata cannot be treated as an afterthought confined to final datasets; they must be integrated into the testing processes that generate those datasets so that evidence can be reconstructed end-to-end. We demonstrate this by augmenting an existing simulation-based testing framework with provenance tracking and metadata collection mechanisms, and by using these extensions to enrich a mobile robot navigation dataset with structured provenance and FAIR-aligned metadata. Finally, we discuss obstacles encountered in this integration -- such as vocabulary alignment, attribute selection, and adoption of domain standards -- and provide actionable recommendations for implementing provenance-centric, FAIR metadata in robotics validation workflows.
arXiv:2605.29977v2 Announce Type: replace
Abstract: High-fidelity ECG interpretation is increasingly reliant on massive foundation models, yet their deployment in clinical edge-care remains hindered by extreme computational demands. While knowledge distillation (KD) is a promising solution, traditional methods fail to capture the complex spatio-temporal dependencies of ECG signals when transferring knowledge across heterogeneous architectures. In this paper, we propose EVL-ECG, a framework specifically designed for cross-architecture distillation of cardiac diagnostic logic. EVL-ECG introduces three ECG-aware innovations: (1) Multi-Head Cross-Attention Alignment, which harmonizes architectural discrepancies to preserve fine-grained morphological features; (2) Optimal Transport-based Visual Feature Matching, utilizing optimal transport to maintain global structural relationships across ECG leads despite mismatched token representations; and (3) Geometric Intra-Architecture Relation Matching, which distills the latent diagnostic reasoning of the teacher model. Evaluations across ECG benchmarks demonstrate that EVL-ECG yields improvements of up to 2.4% AUC and 1.1% clinical accuracy over existing baselines. Notably, EVL-ECG establishes an efficient 2B-parameter ECG foundation model, suitable for resource-constrained clinical environments.
arXiv:2605.29987v2 Announce Type: replace
Abstract: Although multi-scales representation learning enables elastic-dimension embeddings, nested subspaces often suffer from dimensional redundancy and spectral collapse. To address this, we introduce MIC, a framework that optimizes the geometric landscape of multi-granular embeddings through isotropic subspace alignment. MIC employs Soft Collapse Regularization (SCR) to mitigate redundancy between prefix and residual subspaces via cross-correlation penalties, alongside Spectral Isotropy Regularization (SIR) to ensure hyper-spherical uniformity in low-dimensional prefixes. By unifying these strategies through a self-distillation objective, MIC generates semantically dense representations that maintain high discriminative power. Our experiments demonstrate that MIC significantly outperforms standard baselines, particularly in high-compression scenarios where maintaining informational capacity is most critical.
arXiv:2605.30000v2 Announce Type: replace
Abstract: Front-end web code has become a core product surface for every frontier LLM release, yet evaluating these interactive applications at development speed remains costly because human-judged leaderboards like Arena do not scale. Existing automated proxies typically lean on reference implementations, test suites, or rigid checklists, and tend to miss the reasoned synthesis a human reviewer performs over a live session. We articulate a new evaluation regime that is simultaneously reference-free, autonomously driven, and holistically reasoned, and instantiate it through two artifacts. \textbf{\dataname} is an 11-domain, 54-leaf, 1,000-query WebDev benchmark spanning both static-presentation and interactive-application tasks, balanced across three difficulty tiers and three target-language groups, with briefs rewritten to resist recall from circulated prompts. \textbf{\framename}, grounded in Flavell's metacognitive monitoring, separates evidence accumulation from judgment across three stages: Static Perception forms a first impression from passive observation; Agent-Driven Interaction explores the application autonomously while capturing continuous screen video, audio, and per-step screenshots; Dynamic Scoring issues holistic functionality and aesthetics verdicts with structured failure attribution only after the evidence chain is complete. On \dataname, \framename aligns closely with expert human ratings while surfacing substantial headroom across 13 frontier LLMs on interactive web generation. \noindenthttps://anonymous.4open.science/r/Cookie-3CE/
arXiv:2605.30122v2 Announce Type: replace
Abstract: Deep-learning precipitation nowcasting models are often optimized using pointwise losses such as mean squared error or mean absolute error, which can lead to overly smooth forecasts and poor representation of heavy rainfall. This study investigates whether the predictive performance of an established deterministic nowcasting architecture can be improved by reformulating training as a multi-quantile regression problem. Using SmaAt-UNet as a core model, we compare MSE, MAE, and multi-quantile pinball-loss training on radar precipitation nowcasting over the Netherlands. The results show that multi-quantile training improves the central deterministic forecast, decreasing test-set MSE by 8.6\% compared to a model trained using MSE, while also producing upper-quantile outputs that are useful for risk-sensitive prediction of heavy precipitation. These findings suggest that quantile regression provides a simple alternative to standard pointwise losses without requiring a new architecture or generative sampling procedure. The implementation of our models and training setup is available on \href{https://github.com/gijsvn/Multi-Quantile-Precipitation-Nowcasting}{GitHub}.
arXiv:2605.26068v3 Announce Type: replace
Abstract: Weakly supervised anomaly detection (WSAD) has developed in three primary directions: incomplete, inexact, and inaccurate supervision. However, these directions remain isolated, lacking a unified framework to assess whether they address unique challenges or share fundamental mechanisms. This paper introduces WSADBench, the first benchmark that unifies evaluation across distinct weakly supervised scenarios, benchmarking diverse approaches from specialized WSAD methods to advanced tabular foundation models. WSADBench establishes standardized protocols to evaluate 36 algorithms across 4 modalities by systematically varying label quantity, granularity, and quality, revealing the performance boundaries of various methods. Based on over 700K experiments, WSADBench reveals four critical insights: (i) Strong intrinsic correlations exist between these weak supervision scenarios, challenging the isolation of current research directions. (ii) Specialized WSAD algorithms excel only in extreme label-scarcity regimes but are quickly dominated by tabular foundation models and general classification methods as supervision increases or in OOD scenarios. (iii) Unlabeled data shows inconsistent utility across settings, with marginal gains compared to label refinement. (iv) Models exhibit asymmetric sensitivity to different types of label noise. We release WSADBench as an open-source benchmark with code and datasets to facilitate future WSAD research: https://github.com/SUFE-AILAB/WSADBench.
arXiv:2512.23351v2 Announce Type: replace
Abstract: The flexibility and accuracy of methods for automatically counting objects in images and videos are limited by the way the object can be specified. While existing methods allow users to describe the target object with text and visual examples, the visual examples must be manually annotated inside the image, and there is no way to specify what not to count. To address these gaps, we introduce novel capabilities that expand how the target object can be specified. Specifically, we extend the prompt to enable what not to count to be described with text and/or visual examples, introduce the concept of `pseudo-exemplars' that automate the annotation of visual examples at inference, and extend counting models to accept visual examples from both natural and synthetic external images. We also use our new counting model, CountGD++, as a vision expert agent for an LLM. Together, these contributions expand the prompt flexibility of multi-modal open-world counting and lead to significant improvements in accuracy, efficiency, and generalization across multiple datasets. Code is available at https://github.com/niki-amini-naieni/CountGDPlusPlus.
arXiv:2605.30188v2 Announce Type: replace
Abstract: Reliable probability estimates are critical in many machine learning applications, yet modern classifiers are often poorly calibrated. Post-hoc calibration provides a simple and widely used solution, but the large number of proposed methods, combined with small-scale and inconsistent evaluations, makes it difficult to determine which approaches are truly effective in practice. We introduce a large-scale, standardized benchmark for post-hoc calibration, covering nearly 2000 experiments across tabular and computer vision tasks, including binary, multiclass, and large-scale classification settings. Our benchmark aggregates predictions from a diverse set of classical models, modern deep learning architectures, and foundation models, and provides unified, reproducible implementations of dozens of calibration methods within a common evaluation framework. We argue that Post-Hoc Improvement (PHI) in proper scoring rules offers a principled alternative to traditional calibration error estimators for comparing post-hoc methods, capturing both calibration quality and potential degradation to the model's predictive performance. Using this framework, we conduct the most comprehensive empirical study of post-hoc calibration to date. Our results reveal consistent patterns across domains: smooth calibration functions outperform binning-based approaches, dedicated multiclass methods are essential in high-dimensional settings, and generic machine learning models are not competitive without calibration-specific design. To facilitate future research, we release all data, code, and evaluation tools, providing a plug-and-play benchmark for developing and comparing calibration methods.
arXiv:2605.30190v2 Announce Type: replace
Abstract: Diffusion-based planning has achieved strong results in single-agent offline reinforcement learning, yet scaling to many-agent systems remains intractable due to the curse of dimensionality in the joint trajectory space. We introduce MF-Diffuser, a framework that lifts trajectory planning to the Wasserstein space of trajectory distributions, where the propagation of chaos ensures a small representative subset of agents captures the full population dynamics. Our approach features a value-weighted chaotic entropy objective that reconciles generative fidelity with return maximization, and a hierarchical coarse-to-fine strategy that progressively grows the agent population during denoising. We establish end-to-end suboptimality bounds with four interpretable terms, revealing that mean-field approximation error scales as $O(H^2/\sqrt{N})$ while offline distribution shift provably does not grow with population size $N$, and prove the generated policy is an approximate mean-field Nash equilibrium with explicit convergence guarantees. Experiments on three mean-field RL benchmarks -- spanning stage games, sequential dynamics, and adversarial team competition -- show MF-Diffuser achieves the best return in the majority of settings, with the largest gains on suboptimal offline data and at extreme scales ($N \geq 10^3$).
arXiv:2605.30290v2 Announce Type: replace
Abstract: Self-improvement at scale has been a longstanding goal for reasoning models, and there are two natural places to do it: at test time, through verification-refinement (V-R) loops; and at training time, through self-training methods. Both are gated by the same bottleneck: the verifier. V-R loops stall when verifier scores inflate while accuracy stagnates, and when feedback is too generic to act on; self-training fails similarly when bad self-generated data are added to training. Better verification would unlock both, but the capability we want to train, i.e., catching self-generated errors, lacks training signal. To address this challenge, we propose self-trained verification (STV). Our key observation is that, while a model cannot catch these errors alone, it can when shown the reference solution. We turn this asymmetry into a supervision target and train the verifier to imitate a more informed version of itself. At test time, STV substantially improves V-R loops on hard problems, while alternatives (e.g., SFT, RL on verifier scores, and even meta-verifiers) do not. STV roughly doubles accuracy on hard math and lifts it 14x on scientific reasoning tasks (1.5% to 21%). At training time, we additionally train the generator using RL with STV verifier's feedback inside the V-R loop - a procedure we call verifier-in-the-loop training (ViL). Starting from an RL-converged generator, ViL yields a further 33% gain in pass@1. More notably, the generator's standalone pass@1, with no verifier at test time, climbs 30% relative past where standard RL had converged. Hence, the next frontier in reasoning on hard problems may lie in how we train for and with verification. Website: https://ar-forum.github.io/stv-webpage
arXiv:2605.30355v2 Announce Type: replace
Abstract: We introduce Collective Switched Motion (Cosm), a dynamical system-based heuristic algorithm. Cosm combines locally interacting continuous circular variables with novel global coordination rules that facilitate collective dynamics. Pairwise interactions occur sequentially over a set of conflict-free edge partitions, resulting in an interaction network that switches periodically. Unlike conventional gradient-based approaches, Cosm employs structured, non-smooth switching dynamics with finite-magnitude interactions that sustain collective fluctuations and promote exploration beyond local minima. A correlated perturbation mechanism further promotes coordinated cluster motion in the circular phase space. On the three largest Ising problems from the Gset suite, which have 10,000-20,000 variables and represent 2D spin glasses, Cosm attains the optimal solutions (verified with an exact solver) heuristically for the first time. On two large bounded-degree non-lattice graph instances, Cosm reduces the state-of-the-art times-to-target from hundreds of hours to 36-303 s. Results on benchmark problems with tuned hardness suggest favorable scaling relative to previously characterized dynamical solvers. These results suggest that Cosm's synthesis of local interactions, structured switching dynamics, and global coordination provides an effective computational framework for sparse optimization.
arXiv:2605.30380v2 Announce Type: replace
Abstract: Deep convolutional and transformer-based detectors achieve strong performance for SAR ship detection but are often computationally prohibitive for real-time or onboard deployment. Lightweight models offer improved efficiency yet struggle to capture the complex structural relationships inherent in SAR backscatter. Most existing SAR knowledge-distillation approaches rely on feature or logit matching, which enforces localized activation similarity while neglecting the geometric relationships among object representations. We propose a Structured Unified Relational knowledGE distillation framework for SAR Ship detection (SURGE) that transfers relational geometry from a powerful teacher detector to a compact student detector using a contrastive InfoNCE objective in a shared projection embedding space. To the best of our knowledge, this work presents the first transformer-based SAR ship detector knowledge distillation framework in SAR domain. The framework is architecture-agnostic in the sense that it provides a common region-level distillation interface for two-stage, one-stage and transformer-based detectors without modifying their deployed architectures. Experiments on the SSDD and HRSID benchmarks demonstrate that the proposed method yields substantial improvements for two-stage detectors, achieving up to 6.2 mAP and 8.0 AP75 gains over baseline student and even surpassing teacher performance
arXiv:2605.30443v2 Announce Type: replace
Abstract: Multilingual large language models can generate figurative language, but whether the internal signals driving this behavior are language-specific or reusable across languages is unclear. Using activation steering as a probe, we estimate a direction for a figurative category from figurative--literal activation differences in one language and apply it during generation. Across five figurative categories, six languages, and four multilingual LLMs, these directions steer reliably within their own language, most robustly for metaphor and simile. More importantly, they transfer across languages: a direction learned in one increases the target behavior when applied to another, with German among the most receptive targets. Going further, directions assembled from other languages can match or even surpass a target language's own native direction, while removing this shared component weakens native steering. Together, these results provide direct evidence of a reusable but target-dependent cross-lingual signal for figurative generation.
arXiv:2606.01839v1 Announce Type: new
Abstract: LLM-based agents resolve a user task through many turns of dependent inference and tool calls, producing a workload whose total cost is unknown when the task arrives. Existing multi-turn systems keep the turn as the scheduling unit and decide, turn by turn, whether to disaggregate prefill from decode. That decision rests on the turn's decode length, tool behavior, and KV growth, quantities that are not observable when the scheduler must act, forcing the system to predict them. We show this dependence on prediction is imposed by the scheduling unit, not the workload. Raising the scheduling unit from the turn to the conversation converts turn-level irregularity into a stable, two-phase structure: 1) a compute-bound turn-1 prefill followed by 2) a long, memory-bound tail. Thus, with the conversation as the scheduling unit, placement reduces to reading the first-turn input length and per-decoder KV occupancy, both directly observable. We instantiate this principle in ConServe, which routes the first-turn prefill to a high-throughput prefiller, transfers the KV cache exactly once, and pins the conversation to a single decoder for its entire tail, with no learned model of decode-side cost. Against a per-turn prediction baseline, ConServe reduces p95 time-to-first-effective-token (the latency of a conversation's first user-visible output) by 51.08% and improves energy efficiency by 7.51% while preserving last-turn TBT and SLOs; mapping the two phases onto heterogeneous GPU tiers adds a further 22.75% in energy efficiency.
arXiv:2606.01845v1 Announce Type: new
Abstract: Although large language models (LLMs) have shown considerable progress in pragmatic language understanding, prior research has focused mainly on their comprehension of verbal behavior. Nonetheless, non-verbal behavior remains a fundamental component of human communication, especially when deliberately utilized in isolation to convey indirect meanings. In this work, we present the first systematic evaluation of LLMs' ability to infer pragmatic meaning in dialogue consisting solely of non-verbal responses. We explore three research questions: (1) Can LLMs recognize indirect intent conveyed through non-verbal responses? (2) When and how do LLMs fail to capture non-verbal intent? (3) How can we improve LLMs' ability to interpret non-verbal intent?. Through the evaluation, we observe that LLMs struggle to infer underlying meaning from non-verbal responses, with accuracy dropping by up to 60% points compared to verbal ones. Further extensive analysis reveals a behavioral pattern in LLMs' interpretations of non-verbal behavior and demonstrates that in-context learning facilitates pragmatic inference.
arXiv:2605.26436v2 Announce Type: replace
Abstract: Discrete masked diffusion language models such as LLaDA generate text through iterative denoising, where mask tokens are progressively replaced with predicted tokens. LLaDA2.1 introduced a Token-to-Token (T2T) editing mechanism that accelerates generation by directly replacing committed tokens suspected of being incorrect. However, we identify fundamental limitations of T2T editing: it couples error detection with replacement, pollutes the generation context with potentially incorrect tokens, and introduces a train-inference noise mismatch where systematic model-generated errors differ from the random perturbations seen during training. We propose Token-to-Mask (T2M) remasking, a training-free, drop-in replacement for T2T editing that resets suspected erroneous tokens back to the mask state, allowing the diffusion process to re-predict them under cleaner context. We design and empirically validate three complementary error detection strategies -- probability-based, trigger-mirrored, and temporal-difference-based -- and provide a unified theoretical analysis showing that T2M remasking purifies the generation context, converts systematic inference errors back to the model's native mask noise type, and enables delayed commitment for joint multi-position optimization. Comprehensive experiments across 12 benchmarks spanning knowledge, reasoning, mathematics, coding, and instruction following show that T2M generally improves performance on tasks requiring precise token-level output, with the largest gain on mathematics (+5.92% on CMATH). Error analysis on CMATH reveals that the dominant failure mode is last-mile token corruption -- where correct reasoning produces a corrupted final answer -- and that T2M repairs 59.4% of such cases.
arXiv:2512.21342v2 Announce Type: replace
Abstract: A previously developed multi-ion, two-stream Flux-Corrected Transport (FCT) hydrodynamic model for plasmasphere refilling has been extended to incorporate self-consistent electron temperature evolution. The past assumption of a constant temperature along the modeled flux tube has been replaced by solving the electron energy equation, permitting spatially and temporally varying temperature. This improvement provides a more physically complete representation of the pressure and ambipolar electric-field gradients that influence ion transport. The extended model allows us to investigate two-stage refilling behavior established by prior observations and simulations. The model continues to reproduce the expected dominance of H+, enhanced early-time O+ contributions, and the coupling between H+ and He+ through the ambipolar electric field during the transition between stages. Sensitivity experiments with modified initial ion concentrations, including cases representing seasonal effects, highlight the distinct roles of each ion species in shaping the refilling trajectory. Comparisons across L-shells 3 and 4 further confirm the robustness of the model framework for future extension to three-dimensional geometries. Overall, by incorporating more realistic temperature variations, this enhanced model strengthens the physical understanding for interpreting complex multi-ion transport processes during plasmasphere recovery following geomagnetic storms.