Papers

Disc winds are critical components of accreting astrophysical systems and their observations can yield a wide variety of spectral line profiles. Yet, inferring disc wind parameters remains challenging: the underlying parameter spaces are high-dimensional and degenerate, while the radiative-transfer simulations are imperfect and computationally expensive. This paper tackles bridging this gap. Here, we present 'Speculate', an open-source framework for rapidly emulating and fitting observations with synthetic accretion disc wind spectra, generated from the Monte Carlo radiative-transfer code Sirocco. Speculate includes the release of our cataclysmic variable (CV) training data, a library of 88,680 sparsely gridded Sirocco spectra spanning the ultraviolet and optical wave bands, a suite of pre-trained emulators, and its interactive Marimo user interface. Speculate compresses spectral grids using principal component analysis and then provides three approaches to interpolate the resulting latent representation: with a regular-grid interpolation, a neural network, or a Gaussian process. All three architectures enable Speculate to generate a Sirocco spectrum between $1-40\,\mathrm{ms}$, thereby enabling efficient inference. The first two support rapid $\chi^2$ optimisation. The Gaussian process utilises a Bayesian inference engine coupled with a covariance-aware, flexible likelihood function, adapted and accelerated from the Starfish framework. This function helps absorb correlated discrepancies between simulations and data. Tests show all architectures reproduce unseen Sirocco spectra with broadly comparable accuracy and can recover their physical parameters. Subsequently, we apply Speculate to 5 representative nova-like CV spectra, reproducing several observed line features while also exposing mismatches where the current model grid omits relevant physics.

We present a methodology for analyzing node-level outcomes while experimenting with edge-level treatments in a population connected by an undirected graph. Under our design, nodes are randomly assigned to test or control, and each edge inherits the treatment of its endpoints, with conflicts resolved by randomization. We use each node's assigned status as an instrument for its treatment exposure. We show that the Wald estimator is consistent for the global average treatment effect (GATE), even when the edge weights used to construct the exposure are misspecified. We formalize the assumptions needed both in terms of the linearity of potential outcomes and the sparsity of the graph, and prove the asymptotic normality of the Wald estimator. The estimator is straightforward to implement, requiring no assignment simulations over the graph. Our Monte Carlo study demonstrates the strong performance of our approach in the context of a social platform.

We evaluate large language models (LLMs) in multi-turn interactions through their time-to-event: the number of interaction steps required to produce an event of interest, such as a successful jailbreak or agentic task completion. Under limited compute, interactions may be terminated before the event occurs, so that event times are only partially observed (censored). Existing allocation methods for calibrating time-to-event bounds satisfy the budget only in expectation and can exceed the available budget on a particular evaluation run. Enforcing a hard constraint is particularly challenging as the cost of a trajectory is initially unknown. We introduce Hard-budget Allocation with Reflow for Predictive calibration (HARP), a budget allocation that satisfies hard resource constraints and adaptively reallocates unused budget. We show how to use HARP to construct lower predictive bounds (LPBs) on the time-to-event and to estimate evaluation metrics such as the jailbreak rate on a fixed benchmark. Although HARP induces dependence in acquisition decisions across different trajectories, we prove that HARP never exceeds the target budget, that its LPBs have finite-sample coverage guarantees, and that its metric estimates are unbiased. Experiments on agentic task success, LLM jailbreaks, toxic content generation, and RAG hallucinations show that HARP achieves coverage close to the nominal level with low variance, while never exceeding the given budget.

Several hundred ring-like structures seen at $8\mu$m in the Spitzer GLIMPSE surveys share a uniform morphology: the shell thickness $\Delta R$ is a nearly constant fraction of the outer radius $R$, $\Delta R/R\sim0.3$, across a factor of 40 in size and all exciting spectral types. This invariance is unexplained. We ask what sets this thickness and whether its near-constancy is physical or a measurement artifact. We derive analytic constraints on the location, temperature, and heating of the emitting grains, and confront the predicted scaling with radius against the GLIMPSE I and II catalogs (591 objects, 91 with kinematic distances). Grains in radiative equilibrium at the observed radii reach only 20-35 K and fall short of the required $8\mu$m emissivity by more than 20 orders of magnitude; the emission must come from stochastically heated PAHs and very small grains, and therefore from the photon-dominated region, where the ring width is the far-ultraviolet penetration depth. This gives $\Delta R/R=A_V^{\rm PDR}/A_V^{\rm cloud}$, independent of radius; a swept-up shell instead predicts $\Delta R/R\propto(R/R_S)^{3/2}$, with $R_S$ the initial Strömgren radius. We find a pronounced angular-resolution systematic that inflates the ratio for the smallest bubbles; once this is accounted for, $\Delta R/R=0.217$, with a residual dependence on physical radius of $0.03\pm0.05$, consistent with zero. A fixed ambient density is excluded at $20\sigma$, the swept-up shell at 9-28$\sigma$. The implied columns are $N_{\rm cloud}\approx8.7\times10^{21}$ cm$^{-2}$ with only 0.112 dex of scatter. The near-universal appearance of mid-infrared bubbles follows from the ratio of two nearly invariant column densities – the dust-opacity-fixed far-ultraviolet shielding column and the Larson-like column of the parent molecular cloud – and requires no reference to stellar winds.

The goal of the partial information decomposition (PID) is to quantify the redundant and synergistic information that multiple sources provide about a target. PID has many applications in machine learning, neuroscience, and other fields, but defining and computing it for high-dimensional continuous systems remains challenging. Here, we define a PID for multivariate Gaussian systems based on the Blackwell order, which formalizes when one channel is more informative than another. We prove that Gaussian channels are optimal for extracting both redundant and union information, yielding an intuitive geometric interpretation and an efficient numerical algorithm for the PID. Our union information and synergy coincide with the well-known BROJA measures, and we derive closed-form expressions for both in the case of two sources. We also argue that Blackwell redundancy (which differs from BROJA) is the only existing redundancy measure that satisfies a set of natural desiderata. We demonstrate the scalability of our method on systems with up to a thousand dimensions or sources. Our approach is illustrated on an optimal control problem, where it identifies redundant and synergistic interactions between sensor and memory.

Runtime gates for agent tool calls are stacked on the assumption that their errors multiply. We test it on 1,119 labelled agent actions from three corpora, without an adaptive adversary. The stack has one deterministic rule layer and four LLM judges, three of them re-collected with the served model recorded on every call. We read each stack as a number of multiplication-equivalent layers, n_mult, with its floor under perfect coupling. Under the STRICT miss definition (escalation to a human scored as not stopped), any two judges compose to about 1.2 to 1.4 layers ({\phi} median +0.430, 6 of 6 pairs significant, floors 1.02 to 1.17). The rule layer plus one judge composes to 1.86 to 2.09 layers ({\phi} median +0.014, 0 of 4 significant, floors 1.01 to 1.09). Under PRIMARY (escalation scored as caught) the bands are 1.21 to 1.57 and 1.80 to 2.13. Intervals separate on the pooled data, point estimates split on each corpus, and a third-vendor judge lands in the judge band. Solo accuracy does not predict what a layer adds: a cloud rule pack lowers the rule layer's solo miss rate by 20% and adds no new joint coverage. The difficulty share of judge coupling is not identifiable: 31.8% to 61.8% depending on the probe and the miss definition. One judge tier was served by an unrequested model version in 50 of 112 batches, concentrated on the external corpus. That event overturned a pre-declared analysis rule, and the scoring of review verdicts reversed five conclusions. We report both.

Post-wildfire debris flows (PFDFs) are destructive sediment-laden hazards triggered when intense rainfall strikes recently burned terrain, destabilizing hillslopes and threatening infrastructure, local economies, and community safety. Data-driven methods have been proposed to learn predictive patterns directly from historical PFDF observations. However, the current research landscape of data-driven PFDF prediction remains highly fragmented across feature spaces, model architectures, and evaluation protocols, making rigorous comparison and the derivation of scientific insights difficult. Moreover, existing studies lack a systematic investigation into the relative importance of heterogeneous factors (e.g., meteorological conditions, terrain characteristics, soil properties, and burn severity) in triggering PFDF. To address these limitations, we present a unified benchmark for data-driven PFDF prediction, enabling fair and comprehensive evaluation across diverse models and feature configurations. Furthermore, to better understand the underlying drivers of PFDF formation, we propose a reinforcement learning-based feature selection framework that identifies factors whose perturbations render positive and negative events indistinguishable, thereby discovering the regional underlying mechanisms of PFDF occurrence across regions. Our code and benchmark are publicly available at https://github.com/KINDLab-Fly/PFDF-Benchmark.

Bandpass calibration corrects instrumental responses across observing frequencies in interferometric radio observations, so the measured calibration solutions inherently contain spectral variation that varies with instrumental configurations and noise conditions. However, some signals contain anomalies, which if not caught can lead to poorly constructed science products. Observatory pipelines often use scalar metrics with individual, linear decision boundaries. Constructing a simple parametric model can be effective for observations with a relatively limited parameter space, but it becomes progressively less accurate as the parameter space expands. To address this challenge, we design a machine learning framework that treats anomaly identification as a supervised classification task and learns complex, non-linear decision boundaries across multiple features optimized for predictive performance. In particular, we use an XGBoost model that sits on top of several features and specialized scan statistic models to classify anomalies in amplitude solutions from ALMA bandpass calibrations. Trained on an extensive, quality-controlled ALMA dataset, this classifier framework leverages features derived from kernel-based scan statistics to characterize deviating spectral intervals, while remaining extensible to additional feature sets. This method substantially reduces misclassifications in bandpass anomaly detection compared to the current pipeline heuristics. By improving flagging decisions and calibration quality, this classifier framework provides a scalable supplement for observatory operations and is generalizable to next-generation radio facilities.

Systems engineers use several diagrams to describe the structure and behavior of systems. Engineers create these diagrams together to make sure that they use the same elements and remain consistent with one another. Large language models can generate diagrams as text or code, which makes it possible to create system diagrams automatically. However, their ability to generate coherent sets of diagrams is not well understood, and existing datasets and benchmarks do not directly measure this ability at scale. We introduce SEMAADB (Systems Engineering Modeling Assistant with AI Dataset and Benchmark), a dataset of 3,000 engineering contexts and 15,000 diagrams. Each context contains five connected SysML views: Requirement, Block Definition, Activity, State Machine, and Sequence. Here, a view is a diagram that presents one aspect of a system. We checked the diagram sets for consistency and valid rendering. A set of 100 contexts is also human-verified and forms the benchmark test set. We evaluate three language models on two tasks. In diagram repair, the strongest model repairs 64.3% of semantic errors . In cross-diagram update the best propagation F1 is 80.7% when a model applies one change across related diagrams. The results show that syntax repair is nearly solved, but semantic repair and consistency across diagrams are still challenging tasks for models. SEMAADB therefore provides both a large diagram resource and a set of benchmarks for measuring coherent multi-diagram SysML generation.

Video world models track objects they can see; we ask what they keep of objects they cannot. We hide an object from a frozen V-JEPA 2 predictor and compare its prediction for the hidden region with the encoder's representation of two worlds that differ only inside that region. The predictor's decision keeps a stationary object in part and one carried inside a container not at all, and loses a moving one within 0.3 s (0.5 s under V-JEPA's own tube mask; ViT-H keeps it to 1.1 s at pretraining's 90% masking ratio); in projection a trace remains, below the midpoint, at 14-60% of what a baseline copying the last view retains. The information is there: the encoder reads the object's presence at 1.00 and keeps a closed container's contents decodable for 3.5 s, while the predictor's output, read with the encoder's own probe, contains the ball in 2% of scenes once the box has been closed for half a second. On rendered scenes, permanence is missing on the predictor's side, and training installs it cheaply as a prior: three thousand predictor-only steps on synthetic containers take this belief from 0.05 to 1.00 against two matched controls. They also raise IntPhys-2019 from 84.2% to 93.3%, but so does a curriculum without containers, and which training habit the benchmark credits changes with its scoring rule. Continued training with tube masks produces 1.1-1.6 s of moving-object carry-over on manipulation and internet-style video, so the deficit is not intrinsic to latent prediction. VideoMAE keeps almost nothing, and Cosmos's next-token prediction keeps a stationary hidden object but not one carried inside a moving container.

Deciding when to escalate a query from a small language model to a larger one requires a cheap signal that predicts, before the large model is called, whether escalating would help. Semantic entropy, originally developed to detect hallucinations, is a natural candidate: it measures how much a model's sampled answers disagree in meaning, and high disagreement often signals an unreliable answer. We test it across three benchmarks and two model families. On GSM8K, with a small/large pair about twelve times apart in size, semantic entropy reliably distinguishes the small model's mistakes (AUROC 0.871) and improves routed accuracy over random escalation by up to nine points at matched cost. An earlier strong-looking result on a synthetic benchmark proved misleading: a simple rule based only on question difficulty, with no model involved, matched semantic entropy almost exactly. This paper's main contribution is a set of checks that catch this before it is reported as real. We show that scoring a cheap, question-only difficulty estimate alongside any signal reveals whether the signal adds real information or just tracks how hard a question looks; that two reasonable definitions of "escalation worked" can produce very different results on the same data; that a benchmark can leave almost no room for any signal to beat simply always using the large model; and that the true cost of live sampling can make routing more expensive than calling the large model directly. For a cheaper alternative that reuses cached past outcomes, we show how to predict whether it will work on a new dataset – confirmed by correctly forecasting a collapse from AUROC 0.908 to chance level (0.518) ahead of time. We offer these as a general checklist for evaluating escalation signals.

We study the magnetic-field-driven critical end point (CEP) of the QCD thermal transition at zero baryon chemical potential using a five-dimensional Einstein–Maxwell–dilaton (EMD) model. We calibrate the model to lattice entropy at zero and finite magnetic fields using a neural-network formulation of holographic radial evolution. As the magnetic field increases, the thermal transition changes from a crossover to a first-order transition. The merger of temperature turning points, a divergent specific heat, and a vanishing squared sound speed identify $(eB_c,T_c)\simeq(7.22\,\mathrm{GeV}^{2},95.9\,\mathrm{MeV})$. The transition-region inflection of $s/T^3$ defines the crossover temperature, while an equal-pressure Maxwell construction determines the first-order coexistence line. The two lines meet at the CEP. The change of transition order agrees qualitatively with lattice results. The phase boundary is a prediction of the calibrated model.

In recent years, Krylov space approaches have emerged as a useful framework for characterizing dynamical chaos in a wide variety of physical systems. However, the connection between Krylov dynamics and the established structures of classical chaos remains largely unexplored. Classical chaos is characterized by exponential stretching and mixing. For quantum maps with a classical limit, we show that these two signatures are encoded in different structures within Krylov space: stretching is encoded in the geometry of the Krylov operators as determined by the Lyapunov exponent, while mixing manifests itself as an exponential decay of the Arnoldi sequences in agreement with the Ruelle-Pollicott resonances.

Extremely Lossy Functions (ELFs) are a standard model primitive that captures many useful properties of random oracles (Zhandry, Crypto 2016). While there are many variations of ELFs with additional properties, every construction (excluding obfuscation) has followed essentially the same template of bootstrapping from a sequence of ELFs secure only against fixed-size adversaries, and every construction was based on only the exponential hardness of DDH (or $k$-Lin), an assumption that is only reasonable over elliptic curves. We introduce and construct Extremely Lossy Trapdoor Hashing (ELTDH), a stronger notion that implies all known variations of ELFs. Our construction achieves ELTDH in one go, without bootstrapping from schemes secure for only fixed-size adversaries, which makes it simpler and more efficient than existing ELFs. We assume exponential security of the small-exponent discrete logarithm, together with polynomial security of decisional composite residuosity (DCR). Exponential security is only required in the size of the secret exponent, not the size of the group, so the assumption plausibly holds for multiplication modulo $N^2$ (and for many other cryptographic groups), despite the subexponential-time discrete logarithm attacks from index calculus. Our results diversify the assumptions underlying ELFs, while also giving a simpler construction.

Long chain-of-thought reasoning increases sequential decoding cost while creating a growing history of potentially reusable continuations. We investigate when this history supplies useful drafts and complements an existing drafter. Controlled source comparisons reveal trajectory-specific reuse, motivating our method Trajectory-Local Adaptive Retrieval (TLAR). TLAR retrieves approximately matched continuations from the current trajectory and uses recent verification outcomes to adapt retrieval activation and candidate width. TLAR combines retrieved continuations with model-generated drafts in a shared candidate tree, preserving the target model's output distribution through exact verification. Across code debugging, mathematics, and open-ended writing, our evaluation connects source reuse, incremental acceptance, and execution cost. Combining TLAR with strong retrieval baselines improves token acceptance under matched verification budgets and increases end-to-end throughput over the draft-model baseline. These findings support generated trajectories as runtime memory for adaptive inference.

Group Relative Policy Optimization (GRPO) avoids a separate critic by estimating advantages from rollout groups. For multi-turn agents, however, trajectory-level supervision provides coarse, noisy credit: terminal rewards do not locate errors and can penalize useful actions alongside mistakes. Group-in-Group Policy Optimization (GiGPO) and subsequent methods refine supervision through state-conditioned comparisons, but their credit estimates remain sensitive to downstream decisions and outcomes. We introduce RELACE, Retrospective Likelihood-based Action, a critic-free framework that integrates retrospective action assessment with state-conditioned advantage estimation. RELACE evaluates executed actions through teacher-forced likelihood scoring under both their original contexts and outcome-augmented contexts. Comparing these likelihoods yields a trajectory-normalized retrospective factor that captures outcome-dependent changes in action plausibility, rather than hindsight plausibility alone. We use this factor to reweight discounted task returns and construct local advantages by comparing weighted returns among actions from equivalent states within a task. This couples retrospective relevance with observed reward, producing fine-grained credit that complements trajectory-level GRPO supervision. Temporal smoothing and success-protecting masking further stabilize the local signal. RELACE requires neither auxiliary value nor reward models nor additional autoregressive rollouts for credit estimation. Experiments on ALFWorld and WebShop with Qwen2.5-1.5B-Instruct and Qwen2.5-7B-Instruct demonstrate substantial improvements over GRPO, GiGPO, and HCAPO. With the 1.5B model, RELACE achieves $96.35\%$ success on ALFWorld and $79.43\%$ on WebShop, surpassing GiGPO by $5.47$ and $5.60$ percentage points, respectively.

Training large language models (LLMs) is resource-intensive, and adapting them for diverse deployment scenarios with varying computational constraints remains challenging. While elastic architectures enable flexible model deployment and sparsely activated models allow input-adaptive computation, existing approaches treat these dimensions independently. Moreover, models catered towards on-device edge inference need to conform to the memory and compute limitations of the serving devices. In this paper, we introduce a unified framework that combines elastic structures with sparsely gated architectures to create models that adapt simultaneously to both deployment constraints and task requirements. Our approach employs a model backbone that conditions on both the context and target efficiency specifications, enabling fine-grained control over the accuracy-efficiency trade-off at inference time. The model learns to activate task-relevant parameters within elastically-nested sub-networks, allowing a single model to span multiple capacity points while maintaining input-adaptive routing. Through experiments we demonstrate that we can create a model that allows the flexibility to use 1,2,3,4 billion parameters while being more accurate than their dense counter-parts (2-5% on knowledge-intensive benchmarks) and at par with their static versions while delivering similar latency metrics as dense models. Overall, we save on device disk space by sharing the model parameters, allow flexibility of serving based on DRAM and compute available while delivering more accurate results.

We prove that the Boca–Cobeli–Zaharescu (BCZ) map is strongly mixing with respect to normalized Lebesgue measure on the BCZ triangle.

Repeated evaluation and differentiation of Signal Temporal Logic (STL) robustness can become a computational bottleneck in robot planning and control. Sequential temporal recurrences limit parallelism, while dense masking increases memory requirements. We propose ScanSTL, which combines associative temporal aggregation with parallel scans and ordered block reductions. Eventually and Always use range extrema, while inclusive strong Until composes compact segment representations. A common range engine handles bounded and shifted intervals, including the guards required before delayed Until witnesses. Each exact temporal operator computes complete robustness traces with linear work and storage and logarithmic parallel depth on uniformly sampled finite signals. An open source JAX implementation supports automatic differentiation, batching, and compilation. We compare ScanSTL with STLCG and STLCG++ using CPU and GPU operator benchmarks and nine composed specifications. Across these nine specifications at 512 samples, ScanSTL achieves geometric mean speedups of 243 times for forward evaluation and 104 times for gradient computation over STLCG++ in JAX on the CPU. On an RTX~5090 GPU, ScanSTL evaluates unbounded Until over more than two million samples with median times below 0.1 ms for forward evaluation and 0.25 ms for gradient computation. Simulated escort and patrol experiments with a robot dog further demonstrate faster repair of violating plans and greater solver capacity in model predictive control.

IAM policy analysis typically emphasizes the authorization capabilities encoded in a policy, but security analyst review priority may also depend on the behavioral and environmental context surrounding a policy event. This paper evaluates whether contextual information provides measurable incremental value for interpretable IAM policy risk prioritization beyond policy and effective-authorization information. AWS is used as the experimental cloud provider because its IAM and audit-telemetry ecosystem enables controlled evaluation using AWS IAM Context Bench, a benchmark containing 534 real AWS experimental observations across policy, environment, and behavioral scenarios, including matched cases where policy and environment remain fixed while behavioral context changes. Three Explainable Boosting Machine models are evaluated under the same leakage-controlled grouped cross-validation protocol: a policy-centric baseline, a policy-plus-environment model, and a full-context model incorporating CloudTrail telemetry. The full-context model substantially reduces analyst-priority prediction error relative to the policy-centric baseline and closely tracks the reference priority ordering. In matched same-policy context pairs, the policy-centric model remains invariant, whereas the full-context model separates benign and suspicious behavioral conditions with high directional accuracy. The results also show improved concentration of high-priority cases at the top of simulated analyst review queues. These findings indicate that behavioral and environmental context can provide useful incremental information for analyst-oriented IAM risk prioritization while preserving an interpretable additive model structure. The formulation is applicable beyond AWS conceptually, although cross-provider validation remains future work.

Estimating extreme conditional quantiles faces two major challenges: understanding complex nonlinear relationships between variables and accurately extrapolating into the tails of the distribution, where data are sparse. To address both issues, we introduce a novel approach that combines the theoretical framework of block maxima with the predictive power of generalized random forests. The conditional distribution of maxima is modeled by a generalized extreme value (GEV) distribution whose parameters depend on the covariates and are estimated via a weighted maximum likelihood procedure, with weights derived from generalized random forests to capture complex high-dimensional structures. The conditional quantile estimator follows from the inversion of the estimated GEV distribution function. We establish theoretical results ensuring the existence and consistency of the proposed weighted maximum likelihood estimator. The method provides a robust and flexible framework for extreme quantile regression. The performance of the proposed approach is illustrated using simulated data, as well as an application to financial portfolio losses from stocks listed on the NYSE, AMEX, and NASDAQ.

Identity documents (IDs) are some of the most consequential sites through which transgender people encounter state infrastructures. This article examines document transition, the work of aligning names, gender markers, and related information across official records in India based on focus groups with sixteen trans participants, most of whom were transmasculine and gender diverse, in mid-2025. Participants described encountering an absence of clear protocols, changing demands for proof, and objectionable conduct from officials, arising often from limited understandings of transness in systems and policies. As a result, participants faced harms related to livelihood, housing, voting, travel, redress from violence and discrimination, and education. We situate these findings within information studies and trans studies scholarship on classification and state recognition. Participants with mismatching IDs faced a form of torque or administrative violence that we call `trans tax' including disproportionate tax deductions, reluctant employers, and resultant insecure jobs. Further, gender-concordant IDs could redistribute harm: participants who became administratively legible as male lost jobs, benefits, and rights despite remaining vulnerable to the discrimination the programs sought to address. The disruption of people's lives by updated gender-concordant data calls for a re-examination of `haunting' at the intersection of transness, classification, and data practices. The article concludes by calling for sensitization efforts directed at street-, system-, and policy-level decision-makers about trans identities and state provisions together with established and widely disseminated protocols for document transition in the short term; as well as a broad and coordinated upheaval of systems and policies to undo cisgender-heteronormative assumptions.

On-policy reinforcement learning has become a central paradigm for improving the reasoning abilities of large language models. However, its effectiveness is often limited by reward sparsity: when a model fails to discover correct trajectories for difficult problems, the optimization process receives little useful signal and may stagnate. Existing approaches mitigate this issue by incorporating off-policy demonstrations, expert traces, or model-generated solutions, but they typically require the auxiliary data to match the format of the reinforcement-learning task, often relying on rejection sampling from stronger models to obtain suitable training trajectories. We introduce Rationale-Guided Policy Optimization (RGPO), a framework that adaptively leverages ground-truth rationale information according to the model's current capability while preserving its freedom to explore. Rather than treating reference solutions as fixed imitation targets, RGPO uses them as temporary scaffolds: rationales help the model generate improved responses, after which only higher-reward, model-generated solutions are transferred back to the original unguided setting. This design allows training to exploit available ground-truth information without requiring off-policy data to follow the same format as the RL task. Across both language-only and vision-language reasoning settings, RGPO consistently improves performance over RLVR baselines, and ablation studies show that adaptive rationale guidance is a key contributor to these gains. These results suggest that RGPO offers a practical and general approach for reducing reward sparsity, stabilizing reinforcement learning, and improving reasoning performance in both text-only and multimodal models.

A single spatial snapshot of a system with multiple interacting populations captures positions but not the underlying dynamics that have led to the observation. This work asks what the stationary second-order spatial statistics of a snapshot can reveal about the structure of the linear reaction-diffusion dynamical model, and in particular about the pairwise directed entries of its interaction operator. The observation is the matrix-valued structure factor $S(k)$ as a one-parameter family in the diffusion-inherited dependence on the wavenumber $\kappa=|k|^2$. The main result is that the directed interaction operator $A$ and the diffusivity matrix $D$ are generically identifiable for any number of interacting populations from $\{S(\kappa)\}$ up to a single global time rescaling. The findings do not contradict known bounds and the recent result that spatial observational data only determine the Markov equivalence class of the underlying graph. In this work, we identify physical assumptions that lift this barrier. The absolute time scale is not identifiable from a single snapshot of any sample size and requires an external clock. For tumor-immune tissue, the results imply that directed cell-cell interactions are in principle recoverable from one section, that the direction cannot be inferred from the mean densities or a covariance matrix, but from the wavenumber dependence of the cross and autospectra, and that a baseline section determines the interaction structure but not the absolute rates.

Protein-molecule virtual screening is increasingly cast as a problem of representation learning in a shared embedding space. Existing methods rely on dense holistic alignment, entangling invariant binding determinants with nuisance correlations and limiting transfer to new targets. It has been noted that binding in protein-molecule systems involves sparse cross-modality interactions: binding is governed by a small contact interface and a few decisive local interactions (e.g., hydrogen bonds, hydrophobic contacts, and salt bridges) rather than the global structures of the protein and molecule. We hypothesize that uncovering and leveraging sparse interaction patterns is critical for generalization beyond the training data, as these patterns are reusable and expected to improve performance across different scenarios. In this paper, we aim to identify and leverage sparse interaction patterns, and verify our hypothesis. Since the training data contain only observed binding pairs, we formalize this prior via a V-structure causal model under Heckman-style selection, and establish three theoretical results: (i) the latent concepts of interacting proteins and molecules are not identifiable without appropriate sparsity constraints; (ii) these concepts and their sparse interactions are component-wise identifiable under structural sparsity conditions; and (iii) a low-rank relaxation of these conditions yields subspace identifiability of the concepts and interactions. Inspired by these principles, we propose CausalBind with three implementation variants. Extensive experiments on DUD-E and LIT-PCBA benchmarks show that all variants consistently outperform strong retrieval baselines, with the largest gains on LIT-PCBA early enrichment, and further generalize to target- and scaffold-level out-of-distribution splits. Code is available at https://github.com/lokali/CausalBind.

Tactical Combat Casualty Care (TC3) requires responders to connect visual observations of injuries and interventions with established clinical guidance. Developing vision-language models to support this process requires supervision that links visible evidence to traceable doctrine. We present TC3-VQA, a dataset constructed from public instructional and field TC3 videos and authoritative TC3 documents. It contains 581 items spanning 11 concepts, with 1,860 questions covering intervention recognition, doctrine, clinical reasoning, procedural guidance, and refusal when visual information is insufficient. Doctrine-based answers preserve verbatim source passages and character offsets. Construction combines visual annotation, passage retrieval, entailment checks, and verification across model families. Equipment boxes, anatomical labels, temporal segments, and source metadata accompany the question-answer pairs. Automated audits and ratings by two physicians and two medical students characterize annotation quality, with human ratings available for 88 retained items. The dataset provides a resource for adapting vision-language models to TC3, studying the connection between visual evidence and clinical knowledge, and evaluating recognition, doctrine recall, and abstention.

Audio benchmarks are built around short, pre-segmented clips, limiting model design to brief inputs or fixed vocabularies. To close this gap, we introduce Logbook, a benchmark for hour-scale audio understanding, with recordings ranging from ten minutes to six days. Given a continuous audio recording and an event label vocabulary, a system must predict a gap-free segmentation with an event label and a description per segment. We compare 52 systems, end-to-end and cascaded, and ablate fine-tuning, context length, and reasoning budget. We find the task tractable, though the best systems remain below the human reference. Also, over-segmentation is pervasive, and fine-tuning partially mitigates it. Finally, end-to-end are often better than cascaded systems, but degrades with longer context.

Removing copyrighted, unsafe, or user-specified concepts from a deployed text-to-image diffusion model is now a practical requirement. Weight-editing methods can suppress fixed targets, but they require per-target retraining and modify the model checkpoint. Training-free methods, on the other hand, are deployment-friendly, but they suffer from text-side routing failures on compositional prompts. In such prompts, the erase target may be invoked through a related class rather than its lexical name, and its modifiers may migrate onto preserved objects. This paper proposes Hierarchically Grounded Semantic Surgery (HGSS), a training-free framework for compositional concept erasure. The framework lifts both the routing signal and the edit operator used by text-side erasure. First, hierarchical span grounding resolves erase-target spans through lexical, taxonomic, and semantic evidence, while guarding against broad-hypernym and compound-head false positives. Second, dynamic attribute binding refines the text conditioning during early denoising via a counterfactual reference and a preserve-aware cross-attention objective, keeping surviving attribute-noun bindings intact. HGSS selectively removes the erase target without updating model weights or adding learned parameters. On SEE, HGSS cuts hierarchical evasion from 29.54 to 10.02 and roughly halves pairwise attribute leakage, achieving the best Neighbor E and AttrP scores among the reported erasure methods. On UnlearnCanvas, HGSS slightly improves the six-metric average over the matched Semantic Surgery baseline, reaching state-of-the-art.

The sequence in which a person encounters spatial evidence can change a later choice, yet an order effect alone does not identify a quantum-like cognitive structure. We introduce Quantum-Like Spatial Decision Dynamics (QSDD), a falsifiable state-space account of embodied decision making in which cue exposures are completely positive trace-preserving maps, intermediate judgments are quantum instruments, and route commitment is an operationally defined measurement. We distinguish this formal claim from any assertion that cognition is microscopically quantum. We then specify a minimal binary model with a fixed decision basis, noncommuting cue rotations, a fixed symmetry-breaking initial azimuth, dephasing, and lapse; its five fitted parameters must generalize across environments rather than being refit to individual conditions. A reproducible simulation study evaluates recovery and model discrimination under both QSDD and seven-parameter classical logistic ground truths. Across 40 replicates per design cell, QSDD was preferred on held-out environments in 85% of QSDD-generated datasets at 240 independent observations per environment-order cell and 95% at 480, but in 0% of classically generated datasets at every tested sample size. The simulation also exposes weak identification of the lapse parameter and is explicitly a design analysis, not human evidence. We provide a preregistrable virtual-reality experiment, telemetry schema, classical comparison set, and failure criteria. Finally, we show how the same process-measurement logic can be transferred to human inspection and debugging of quantum circuits. QSDD is therefore offered as a constrained model to be defeated or supported by data, and as a methodologically continuous route from immersive interaction research to human-quantum computer interaction.

Improving the reliability of large language model (LLM) agents in long-horizon decision-making remains a key challenge. When deployed as autonomous agents interacting with complex environments, early mistakes can propagate through trajectories and cause cascading failures. Recent approaches improve reliability by incorporating external critique or deliberation, but invoking these mechanisms at every step substantially increases token consumption and latency, limiting practical deployment. We propose SAG (Self-improving Agent with Gated critique), a cost-aware framework that formulates critique invocation as a step-wise decision problem during long-horizon interaction. SAG introduces a lightweight, training-free gating mechanism that estimates the utility of critique using action-level ambiguity signals–global entropy and local top-2 margin–computed over admissible actions. From a decision-theoretic perspective, this mechanism approximates the Value of Information (VoI) of critique, enabling the agent to selectively allocate expensive feedback only when its expected benefit justifies the cost. SAG further incorporates online bootstrapped self-improvement, allowing the actor to internalize critic-assisted behaviors and progressively reduce reliance on critique. Across three long-horizon interactive benchmarks and multiple backbone models, SAG substantially improves the performance-cost trade-off compared with both no-critique and always-on critique agents. On ALFWorld, SAG increases task success from 24.6% to 78.4% while maintaining a token budget comparable to ReAct, yielding a $3.1\times$ improvement in normalized token efficiency. Moreover, a 7B actor with a lightweight 3B critic achieves performance comparable to a 14B actor without critique, showing that selective critique can recover most of the reliability benefits of deliberation while dramatically reducing inference cost.