Discovering effective reward functions remains a fundamental challenge in motor control of high-dimensional musculoskeletal systems. While humans can describe movement goals explicitly such as "walking forward with an upright posture," the underlying control strategies that realize these goals are largely implicit, making it difficult to directly design rewards from high-level goals and natural language descriptions. We introduce Motion from Vision-Language Representation (MoVLR), a framework that leverages vision-language models (VLMs) to bridge the gap between goal specification and movement control. Rather than relying on handcrafted rewards, MoVLR iteratively explores the reward space through iterative interaction between control optimization and VLM feedback, aligning control policies with physically coordinated behaviors. Our approach transforms language and vision-based assessments into structured guidance for embodied learning, enabling the discovery and refinement of reward functions for high-dimensional musculoskeletal locomotion and manipulation. These results suggest that VLMs can effectively ground abstract motion descriptions in the implicit principles governing physiological motor control.
Context: Generative AI (GenAI) and Large Language Models (LLMs) are increasingly used for academic tasks in software engineering and beyond, including systematic literature reviews (SLRs). However, whilst their capabilities are rapidly developing, there is no guarantee they can meet the rigour, reliability, and transparency that SLRs require. Objectives: To support researchers intending to conduct SLRs using GenAI, or those conducting empirical studies to evaluate how well GenAI supports SLR tasks. Methods: First, we conducted a rapid review to identify studies that propose guidelines for evaluating and using GenAI and LLMs to support SLRs. Second, we drew on thought experiments, relevant guidance from the literature, and our own experience from conducting SLRs and evaluating tools, to develop recommendations for how to use and assess GenAI in the context of SLRs. Results: We classify methodological issues for the roles of evaluation and reviewing into four categories spanning SLR practice adjustments, general secondary-study standards, GenAI evaluation practices and core empirical rigour. Building on this, we introduce GUEST (GenAI Use and Evaluation in SLR Tasks) which is a comprehensive framework of 51 recommendations designed to guide researchers through planning, executing and transparently reporting GenAI-assisted literature reviews and validation studies. Conclusion: Current empirical evidence indicates that GenAI cannot be relied upon to perform unsupervised systematic reviews so requires human oversight. Nevertheless, it can offer cost-effective assistance for repetitive screening tasks and secondary validation. By standardizing how these tools are deployed and evaluated, the GUEST framework provides a necessary methodological baseline for trustworthy GenAI-assisted research in software engineering.
How many latent variables are needed to fit the cosmic microwave background (CMB) temperature and polarisation power spectra when a neural network is given no cosmological model, and to what extent do they map onto the standard cosmological parameters? We use multi-task learning with variational autoencoders trained on CLASS-generated spectra (TT, TE, EE) over $2\le\ell\le2500$. No cosmological model is built into the network, and the number of latent variables, $L$, is varied over $L=5$, $6$ and $7$. Their fits are compared to the Planck 2018 data using $\hatχ^2$, AIC, BIC and Bayesian evidence. Their correspondence with cosmological parameters is measured with normalised mutual information post-training. The loss weights are set within a single training run by uncertainty weighting, with the weight of the disentanglement term either fixed (semi-automated, SA) or learned (completely automated, CA). The reconstruction loss is split over multipole ranges in three configurations that separate the roles of temperature and polarisation and of the low- and high-$\ell$ ranges. In four of the six combinations of configuration and strategy, the fit improves from $L=5$ to $L=6$ and not beyond, and $L=6$ is the best, or close to the best, in five of the six cases. So, six latent variables are needed, as many as the free parameters of the standard, model-dependent approach. SA gives clearer disentanglement, while CA needs no a priori choice of the disentanglement weight and has the higher evidence at $L=5$ and $6$: the two strategies are points on a trade-off between interpretability and fit to the data. Notably, the latent variables for $τ$ and $A_s$ share almost no information in any model, consistent with large-scale polarisation separating the two, as in standard analyses. A deep-learning approach can thus give a compact, data-driven parametrisation of the CMB spectra.
Deep audio classifiers can achieve strong predictive performance, but their hidden representations remain difficult to interpret. Existing audio explainability methods often focus on input-level attribution, such as waveform segments, spectrogram regions, or acoustic features, and provide limited insight into what hidden neurons encode. Interpreting hidden neurons at the concept level can provide a more meaningful view of what an audio model learns internally. This paper investigates whether hidden neurons in audio classifiers can be associated with human-understandable symbolic acoustic concepts. A model-agnostic neurosymbolic framework is presented for analyzing the final hidden layer of audio classifiers, containing 32 neurons, through symbolic concept induction. The framework is evaluated on two audio classification settings: music genre classification using GTZAN dataset and musical instrument classification using IRMAS dataset. Results show that several hidden neurons can be associated with meaningful induced concepts, and most confirmed neuron-concept pairs are statistically supported using the Mann-Whitney U test. These findings suggest that symbolic concept induction is a promising direction for explaining latent audio representations beyond input-level attribution.
Large language models (LLMs) are increasingly being integrated into web browsers to create agentic browsing systems that execute actions on behalf of the user. Prior work considering the security of agentic browsers focuses on indirect prompt-injection attacks. However, by failing to consider traditional web attacks, previous agentic browser threat models have a blind spot to web social engineering attacks originally designed to trick humans. In this paper, we propose the first web-focused threat model for agentic browsers and use it to derive a taxonomy of 20 attacks across both the web and LLM space, and implement 18 of the attacks. Our threat model extends the original See$\rightarrow$Act browser agent model to account for all components of a browser, and frames the agent as a confused deputy unable to distinguish task steps from traditional web attacks. We show that 10 web threats can reemerge often in amplified forms once an agent can be influenced by untrusted page content. We further conduct a generalizability study on 14 of the 20 attacks, showing that our attacks reproduce across 4 major LLM models spanning multiple vendors. We show that agentic browsers exhibit five major failure modes when facing traditional and LLM web threats, demonstrating the need to rearchitect agentic browsers before they are ready for the current web.
Complex questions in retrieval-augmented generation require evidence spread over several documents, and an answer is reliable only when the retriever finds all of it. Chain embedding retrievers collect such evidence step by step, re-encoding the question together with the passages retrieved so far, and we find that they are already strong: trained only on gold histories, a small chain embedding retriever that calls no language model outperforms graph-based systems and is competitive with a search agent. This training, however, leaves a gap. At test time the retriever conditions on its own retrievals, which soon contain mistakes: by the third step, about half or more of the states it visits hold a wrong passage, and these passages lower its chance of retrieving a missing gold passage well beyond what the difficulty of the states explains. We propose on-policy state training, which trains the encoder on the states it reaches itself. At every epoch the current encoder retrieves chains for the training questions; each state it visits, including those built on its own mistakes, is labeled with the earliest gold passage still missing while the other missing gold passages are masked, so that no correct continuation is penalized; and the encoder is updated contrastively on these states mixed with teacher-forced ones, so that its training distribution follows it as it learns. The resulting retriever retrieves complete evidence more often than teacher-forced training and the systems we compare, and it transfers without further training to harder long-chain variants of HotpotQA and 2WikiMultiHopQA that we construct. On the same states, it recovers from its own mistakes far more often, and controlled comparisons attribute the gain to training on the states the encoder currently reaches.
Cooperative aerial manipulation is difficult under wind, payload motion, and noisy goal broadcasts because multiple Unmanned Aerial Vehicles (UAVs) are coupled through a shared payload. We present CALM (Cooperative Aerial Learned Manipulation), a decentralized controller for aerial manipulators carrying a basket while tracking a moving ground robot. CALM combines a parameter-shared Multi-Agent Reinforcement Learning (MARL) policy, trained using Multi-Agent Proximal Policy Optimization (MAPPO), with goal estimation, per-UAV Extended State Observers (ESOs), basket-mass estimation, and anticipatory formation-yaw feedforward. The anticipatory formation-yaw feedforward uses estimated goal motion to rotate the formation before position error develops, reducing basket-yaw lag without adding net commanded force. A disturbance-conditioned critic supports centralized training while execution remains decentralized. Across held-out trajectories, CALM achieves lower mean position and orientation errors than all evaluated baselines, including the matched arm baseline (7.68 versus 10.95 cm and 3.3 versus 4.7$^\circ$), the cable-suspended realization, and two model-based controllers. Under unseen full-information-path degradation, CALM remains statistically comparable or superior to the arm baseline. Videos are available at https://petreric.github.io/CALM/.
Science often advances when researchers pivot from one discipline to another and import new ideas. However, for established researchers, such pivots are associated with fewer publications and citations. Here, we investigate pivoting at the beginning of a research career--doctoral admissions--when the benefits of importing ideas might outweigh the switching costs. Analyzing 171,347 PhD applications at a large research university from 2013 to 2023, we measured the semantic distance between each applicant's prior degree and target program. Applicants who pivot further had lower undergraduate grades and, controlling for academic and demographic predictors, lower admission chances: the average pivot corresponds to a 5.3% lower admission probability (0.80 percentage points). Among enrollees, the average pivoter takes longer to graduate than non-pivoters, with a 10.7% lower graduation hazard, and similar publication and citation counts on average and at the tails. Our results suggest that pivoting at the outset of research careers, even if it benefits research communities, comes with significant costs to individuals.
Reinforcement learning with verifiable rewards (RLVR) improves language-model reasoning by scoring sampled responses with a rule-based verifier and reinforcing those that pass. Where a classifier's training couples predicted probability to labels, nothing ties an RLVR policy's confidence in a response to its reward: aligned training would drive responses with the same advantage toward equal confidence, yet the advantage is blind to confidence. Under this sequence-confidence misalignment, a success the policy barely found earns the same update as one it produces with near certainty, and a confident error is corrected no harder than an exploratory one. We propose SCARL (Sequence Confidence-Aligned Reinforcement Learning), which reshapes, but never re-signs, each advantage with the alignment gap between the response's group-relative confidence and the outcome its advantage certifies, normalized into shares of each class's advantage budget. Misaligned samples absorb the force, responses with the same advantage are driven toward equal confidence, and repairing the misalignment doubles as a remedy for the entropy collapse that erodes exploration. SCARL provably reduces to its base estimator when a class's confidences agree, adds no extra passes and one hyperparameter. On Qwen3-4B-Base it lifts average Pass@32 by 2.4 points over DAPO and is best among nine compute-matched methods on both coverage and per-draw accuracy; applied unchanged across four estimators and four backbones, it improves 14 of 16 settings while preserving policy entropy and widening the confidence-correctness separation. Our code is available at https://github.com/kezhh163/scarl.
Automated mechanism design has enabled the learning of complex auction mechanisms using neural networks. However, designing non-truthful auctions remains challenging because bidders' equilibrium strategies depend on the auction mechanism and thus evolve as the mechanism is optimized. In this work, we formulate auction design as a Stackelberg game with multiple strategic followers and develop a fully first-order algorithm for revenue optimization in non-truthful auctions. We establish non-asymptotic guarantees for hypergradient approximation and show that, under standard smoothness and strong-monotonicity assumptions, our method reaches an $O(ε)$-stationary point using $\tilde{\mathcal{O}}(ε^{-2})$ first-order oracle calls. Experiments on symmetric and asymmetric single-item auctions, as well as multi-item auctions, show that our method converges faster than second- and zeroth-order baselines and scales better as the number of bidders increases. Our method also discovers auctions that achieve revenue close to truthful baselines. These results provide a scalable framework for automated mechanism design beyond truthful auctions.
We introduce Continuous-Action World Models (CoAWM), a framework for learning cross-embodiment spatiotemporal dynamics priors that support robot action generation across embodiments and execution frequencies. At its core is a shared latent function space that captures the spatial and temporal structure of robot actions. We learn this space through two stages. First, function pretraining brings trajectories from different embodiments and sampling frequencies into a common representation, learning shared dynamics priors through action reconstruction and consistency across temporal sampling views. By accommodating heterogeneous robot trajectories, this stage broadens the range of data sources available for pretraining and facilitates scaling to larger and more diverse datasets. Second, we connect the pretrained function space to a world-model backbone, enabling observations, language instructions, and the current robot state to condition the prediction of a latent action function. A single decoder takes any query time and returns the corresponding end-effector action. By separating shared dynamics priors from task-conditioned action selection and temporal sampling, our approach provides a unified foundation for transferring robot experience across embodiments and control frequencies.
This research proposes a two-level keystroke dynamics authentication system that provides defense-in-depth security through continuous behavioral drift monitoring. Level 1 employs fixed-text password authentication using Temporal Convolutional Networks (TCN) with contrastive learning and population-based training (PBT), achieving 3.77% EER for high-security login verification. Upon successful authentication, Level 2 continuously monitors user typing during natural free-text sessions using a 4-Model Deep Learning Ensemble. If suspicious behavioral drift is detected during monitoring, the system re-authenticates the user by returning to Level 1, providing a feedback loop against session hijacking and behavioral anomalies. The system explicitly models long-term behavioral drift through the fixed-text component's PBT mechanism, which adapts to temporal variations in password typing patterns. This research leverages free-text (Buffalo dataset) and fixed-text (GREYC dataset) keystroke dynamics datasets, extracting temporal and spatial features including hold times, up-down latencies, down-down intervals, and trigram patterns. The final system achieves 96%+ accuracy for fixed-text authentication (3.77% EER) and 83.53% True Acceptance Rate (TAR) at 5% False Acceptance Rate (FAR) for free-text continuous authentication (10.05% EER), demonstrating superior performance and strong resistance to impostor attacks through continuous behavioral monitoring.
We prove that, with high probability, refuting a random $k$-CNF formula on $n$ variables with ${M = n2^{k}}$ clauses for $ω_n(1) \leq k \leq O(\log{n})$ requires polynomial calculus degree at least ${n/2-O(n/\sqrt{k})}$ over every field. Through Pang's recent tradeoff between $\mathsf{AC}^0[p]$-Frege size and polynomial calculus degree [Pang 2026, arXiv:2610.00837v2], this implies a size lower bound of $\exp(Ω_p(\sqrt{k} \log{n}))$ for depth-$2$ $\mathsf{AC}^0[p]$-Frege. In this range of $k$, the bound is superpolynomial in $M$ while $M \leq \mathrm{poly}(n)$. We also prove a nearly matching degree upper bound of $n/2-n\ln(k)/{2k} +O(n/k)$, which shows that other formula families are needed to obtain lower bounds for full $\mathsf{AC}^0[p]$-Frege through Pang's tradeoff.
Model combination offers an alternative to selecting a single forecasting model, particularly when individual models are sensitive to random initialization and hyperparameter choices. We study online aggregation through Follow-the-Leader (FTL) and Hedge-type algorithms, with an application to ensembles of Multi-Frequency Echo State Networks (MFESNs) for macroeconomic forecasting. On the theoretical side, we derive Hoeffding- and Bernstein-type expected-regret bounds for FTL and decreasing Hedge under independent and $\varphi$-mixing expert losses. Under a positive gap between the best expert's expected loss and those of its competitors, the bounds are uniform in the number of forecasting rounds and exhibit logarithmic dependence on the number of experts. Monte Carlo experiments investigate the effects of temporal dependence and the separation between experts' expected losses. Empirically, we construct MFESN ensembles for forecasting U.S. quarterly GDP growth using mixed-frequency macroeconomic and financial indicators. The results show that adaptive aggregation, particularly FTL and AdaHedge, achieves substantial reductions in out-of-sample mean squared forecasting error relative to individual MFESN models and conventional forecast combinations.
Understanding and Mitigating Role Spoofing in LLMs
Ezra Edelman, Benjamin Edelman, Surbhi Goel, Eran Malach
AI 速读 基于摘要
论文研究LLM如何判断对话中“谁说了什么”,揭示角色欺骗攻击的机制,并提出基于线性探针的检测方法。
cs.CR · cs.AI
查看原文摘要
Despite growing capabilities, LLMs remain susceptible to prompt injections: malicious inputs that can manipulate or hijack models. In particular, it has recently been observed that LLMs can exhibit role confusion (Ye et al., 2026), where models attribute the source of text based on its style instead of its labeled role. Malicious text in tool outputs (for instance, a malicious webpage accessed by an agent) can leverage role confusion to impersonate the user, an attack we call role spoofing}. In this work, we seek to understand and mitigate role spoofing. We study how models track "who said what" in a conversation, showing that they use cues from both content (e.g., semantics and writing style) and structure (e.g., control tokens and formatting). We show that LLMs are highly susceptible to role spoofing attacks that combine both fake structure and content. Additionally, in a toy setting, we show that adversarial training which teaches the model to ignore content-based injections does mitigate content-based attacks but leaves the model susceptible to structure-based attacks. This can explain why frontier models, with increased resistance to many injection attacks, can still be susceptible to role spoofing. Finally, we propose a new method for detecting role spoofing using linear probes trained to predict roles in a conversation, and show that it is effective across different settings.
Mathematical supervision for language models is typically flattened into isolated problem-solution pairs, discarding the curricular organization and evidential structure of the original sources. We introduce TOME-Math (Textbook-Oriented Mathematical Evidence), a large-scale corpus that converts 2,118 textbooks spanning the core areas of university-level mathematics subjects into approximately 380K structured proof-supervision records while preserving chapter order, proposition-proof links, dependency references, and source provenance. Leveraging these structural signals, we explore curriculum learning across multiple mathematical subjects and characterize scaling patterns with supervision volume and source breadth. At full scale, fine-tuning on TOME-Math raises the unweighted mean score across our held-out evaluations and established public benchmarks by 16.7 percentage points over the base model, while exceeding all three public instruction-data baselines evaluated in aggregate. Our findings highlight structured textbooks as both a scalable source of mathematical supervision and an empirical framework for studying the impact of data organization on mathematical reasoning. We will release the corpus to facilitate further research on structured mathematical supervision.
We consider the problem of finite-sample identification of an unknown LTI system from a finite hypothesis class. In contrast to existing results that assume the hypothesis class contains the true system, we consider the misspecified setting, where the true system does not belong to the hypothesis class. We first characterize the resulting estimation target and its dependence on the data-generation scheme. We then establish a finite-sample lower bound on the failure probability that any identification algorithm must incur. Finally, we analyze the robustness of an existing exponential weighting-based identification algorithm to the misspecification and derive a finite-sample upper bound on its failure probability. Our results illustrate when using a misspecified finite hypothesis class can provide an advantage over identification without prior knowledge of the candidate systems.
This paper sets out what a sustained federal research and investment agenda would need to include to close that gap--not within a single budget cycle, but over the decade such an undertaking will realistically require. It identifies where the gap is most pronounced today: in the methods needed to verify and certify components of uncertain origin, in the foundational hardware research that has long been underfunded relative to software, in the physical infrastructure researchers need to test systems at scale, and in the workforce required to build and sustain all of it. Robotics should not be treated merely as a downstream beneficiary of a broader artificial intelligence investment strategy. It constitutes its own supply chain, with its own points of failure, and merits its own sustained federal commitment.
Machine learning continues to emerge as an important tool to be utilised within structural engineering and structural health monitoring, due to its ability to accurately and quickly perform both regression and classification tasks. However, a purely data driven approach has its limitations, particularly where we lack data from relevant environmental and operational conditions, a situation that has led to the development of physics-informed machine learners for structural health monitoring. These "grey-box" models take into account the physical insight that an engineer would have about the structure they are modelling and have shown promising results in the structural engineering field among many others. This work compares black and grey-box models through a "green" lens, comparing them in terms of their environmental impact, and investigating how the high extrapolative performance of grey-box models can reduce their runtimes and therefore carbon emissions. The authors aim to develop physics-informed models with reduced computational costs, while maintaining high performance, illustrated through a structural health monitoring case study.
With the emergence of massively parallel processing units, parallelization has become a desirable property for new sequence models. The ability to parallelize the processing of sequences with respect to the sequence length during training is one of the main factors behind the uprising of the Transformer architecture. However, Transformers lack efficiency at sequence generation, as they need to reprocess all past timesteps at every generation step. Recently, state-space models (SSMs) emerged as a more efficient alternative. These new kinds of recurrent neural networks (RNNs) keep the efficient update of the RNNs while gaining parallelization by getting rid of nonlinear dynamics (or recurrence). SSMs can reach state-of-the-art performance through the efficient training of potentially very large networks, but still suffer from limited representation capabilities. In particular, SSMs cannot exhibit persistent memory, or the capacity of retaining information for an infinite duration, because of their monostability. In this paper, we introduce a new family of RNNs, the memory recurrent units (MRUs), that combine the persistent memory capabilities of nonlinear RNNs with the parallelizable computations of SSMs. These units leverage multistability as a source of persistent memory, while getting rid of transient dynamics for efficient computations. We then derive a specific implementation as proof-of-concept: the bistable memory recurrent unit (BMRU). This new RNN is compatible with the parallel scan algorithm. We show that BMRU achieves good results in tasks with long-term dependencies, and can be combined with SSMs to create parallelizable hybrid networks that have both transient dynamics and persistent memory.
Diffusion models have become the leading paradigm in generative modeling, yet many applications require sampling from distributions that are not those of any single pretrained model, but are instead defined through combinations of several models, possibly tilted toward high values of a downstream reward. Naively combining the scores of the models yields incorrect samples, while existing corrections are restricted to specific families of targets. In this work, we introduce the generalized Feynman-Kac correctors (GFKC), a principled framework for sampling from an arbitrary target function of several pretrained diffusion models. We show that the geometric interpolation between an arbitrary initial function and the target function of the models satisfies a Feynman--Kac partial differential equation, and derive its potential in closed form. To evaluate this potential during generation, we further generalize the log-probability ODE to track the density of each model along trajectories driven by a different velocity field. The resulting algorithm operates entirely in the log domain, incorporates reward-based steering at no additional conceptual cost, and corrects the generation on the fly, without retraining or fine-tuning the models. We validate the method on analytic Gaussian models, where it closely matches the ground-truth targets, and on diffusion models trained on MNIST, where it samples from non-trivial combinations of the models and enables effective steering.
The last year has seen a shift in AI-assisted mathematics, with frontier language models producing proofs of and counterexamples to long-standing open conjectures. However, studying only the current set of major breakthroughs provides an incomplete picture of LLM capabilities and the computational resources needed to achieve reported results. To address this, we introduce a new dataset, OpenConjecture, which collects conjectures from mathematics papers published on the arXiv since February 2026. We use these conjectures as a source of open mathematical problems, and apply frontier GPT models to a random sample of 200 of these conjectures over three passes using a reproducible protocol without problem-specific mathematical guidance. Rather than attempt expert verification of these candidate solutions, we analyze the types of problems for which LLMs produce solutions and the characteristics of these solutions. The models we test produce candidate solutions for roughly 60% of the conjectures, with counterexamples about 1.3 times more common than proofs. We estimate solution novelty, assess how solutions cite and relate to existing literature, and analyze whether the proposed resolution relies on either an error in or underspecification of the conjecture. We find that models sometimes propose resolutions that are outside the authors' intent (e.g., proposing the trivial group as a counterexample) and that a number of proofs follow closely from existing results, but that some resolutions exhibit at least moderate novelty. Our findings underscore the need for evaluations that go beyond correctness to assess fidelity to the intended statement, attribution, and the nature of the mathematical contribution.
Large language model (LLM) agents are proposed for clinical prediction but are untested in pediatrics, where cohorts are small and outcomes are uncommon. Using records from 2,489 youth with type 2 diabetes (T2D) and 4,018 with type 1 diabetes (T1D), spanning structured data, note-derived social determinants and, in T1D, continuous glucose monitoring data, we predicted seven T2D outcomes, four also in T1D, at 24 months from 18 months of follow-up. CEDAR (Clinical Evidence-Driven Agent for Risk-prediction), an evidence-citing, self-auditing LLM pipeline, outperformed trained machine learning in both cohorts (pooled area under the receiver operating characteristic curve 0.815 versus 0.656 to 0.700 in T2D and 0.756 versus 0.603 to 0.631 in T1D), most on the sparsely screened outcomes. Only 4 of 70,246 values it cited did not match the chart. Reading only the evidence and reasoning cited by CEDAR, two pediatric endocrinologists reached the same prediction as CEDAR in about nine of ten cases, agreeing with each other on 90.5%.
Recently, Guilmant, Lefaucheux, Ouaknine, and Worrell [2] proved decidability of the termination problem for linear constraint loops over two real-valued variables, thus settling a wellknown open problem. Their algorithm is based on a theoretical result that such a loop does not terminate if and only if it has a witness of an infinite run of a specific shape. By analysing the structure of the witness carefully, we show that termination such loops can be decided in polynomial time, improving on the polynomial space bound argued by Guilmant et al. We also present an efficient implementation of the polynomial-time algorithm based on the satisfiability modulo theories (SMT) solver Z3.
Logic-based workflow patterns capture recurring behavioural structures together with reusable temporal specifications. Their composition raises distinct problems as workflow structures become richer, including instantiation, nesting, conditional execution, repeated occurrences, and traceability of generated formulae. This paper proposes a hierarchy of Pattern Composition ($ΠC$) procedures, from $ΠC_0$ to $ΠC_4$, that organizes these requirements as progressively richer compositional capabilities. The hierarchy separates compositional complexity from the expressive power of the temporal-logic fragment, represented by $TL_j$, yielding a framework $ΠC_i[TL_j]$ for developing logical workflow specifications. We develop activation-aware Pattern Composition ($ΠC_2$), which distinguishes the structural presence of a nested pattern from its execution in a particular workflow context. We formalise its recursive procedure, analyse its properties, and define an evaluation protocol for conditional activation, conservativity under mandatory composition, and specification growth. The hierarchy also characterizes selective occurrence qualification for iterative patterns ($ΠC_3$) and provenance for tracing generated formulae to their sources ($ΠC_4$). Finally, we outline $ΠC_{\mathrm{Assist}}$, an orthogonal mode in which AI proposes and refines pattern expressions subject to the hierarchy's compositional checks.
Vision-Language-Action (VLA) models have demonstrated promising capabilities in robotic manipulation by integrating visual perception, language understanding, and action generation. Recent advances have further incorporated future prediction to enhance the understanding of environmental dynamics and task progression. However, existing methods often model short-term dynamics and long-term task objectives separately, while directly incorporating predicted futures into action policies without sufficiently considering their relevance to current action requirements. To address these limitations, we propose HiFOR, a hierarchical foresight framework with action-guided world retrieval. Specifically, Hierarchical Foresight Prediction jointly predicts short-term state evolution, semantic milestones, and final goal states through a Hierarchical Dependency Mask (HDM), establishing dependencies across temporal horizons. Building upon these predictions, Structure-Aware Latent World Modeling organizes future latent representations along the horizon, view, and spatial dimensions to facilitate fine-grained retrieval. Furthermore, Action-Guided Factorized World Retrieval adaptively retrieves relevant future information from the structured latent world based on evolving action representations during denoising, providing targeted guidance for action generation. Extensive experiments on two simulation benchmarks and six real-world robotic manipulation tasks demonstrate the effectiveness of our approach.
Zero-shot text-to-speech (TTS) models clone a target voice from a short reference clip, commonly referred to as the prompt. In practice, the prompt has to be cut from a longer recording of the speaker, so which part to use must be chosen. To understand what makes a good prompt, we analyze hundreds of thousands of outputs synthesized by four TTS models from candidate prompts of YouTube recordings. The choice of prompt substantially changes intelligibility, naturalness and speaker similarity. Standard predicted-quality scores capture mainly the naturalness part but miss most of the intelligibility gain. We find that the models largely agree on which prompt yields the highest measured speaker similarity, but much less on which prompt yields intelligible speech. We propose a small network that learns from the outputs a TTS model produced to rank a recording's candidate prompts from their audio alone. Without synthesizing anything at selection time, it picks prompts that lower the word error rate on held-out recordings by a third (Qwen3-TTS) to two fifths (F5-TTS) relative to a randomly chosen prompt, more than predicted quality achieves and nearly as much as synthesizing a dozen test sentences from every candidate first.
Large language model (LLM) agents for personal health read wearable data and recommend, but do not learn about the individual in a way that can be audited or falsified. We present EpiCoach, an agentic system that treats one athlete, a cross-country mountain biker, as the subject of accountable N-of-1 science. Claims about the athlete live in an evidence-gated hypothesis registry: each must issue a probabilistic prediction about a future session before it occurs, and its belief moves only through computed evidence from a logged statistical fit. Predictions come from a mechanistic digital twin inferred from heart rate, GPS and barometric altitude: a lagged heart-rate response model, a fitness-fatigue model with learned time constants, and a latent technical-descending skill state, measured on discovered trail segments. The agent prescribes the next training week as a constrained experiment maximizing information gain subject to load and safety constraints, around two instrument rides. We evaluate the system two ways over eighteen months of data. A rolling-origin replay benchmark shows the twin beats standard forecasters on heart rate but not segment time, and the agentic loop matches a static model. A live prospective phase runs the weekly cycle: in its first three weeks the ledger sealed 20 predictions, scored 14, the descent forecast calibrated after one correction, the climb forecast pessimistic as the athlete improved, and the safety officer vetoed the agent twice. Every correction is preserved in the registry, a template for auditable, falsifiable personalization in digital health. We release the registry, benchmark and protocol.
Graphical User Interfaces (GUIs) serve as the dominant medium for human-computer interaction, yet building GUI agents that generalize across the vast diversity of real-world interface environments, with the same flexibility and robustness that humans naturally exhibit, remains unsolved. Notably, GUI data are inherently non-stationary: the continual emergence of previously unseen interface instances (e.g., novel domains and resolutions) induces persistent distribution shifts, significantly impeding the continual learning of existing GUI agents. Reinforcement fine-tuning (RFT) has attracted considerable attention as a promising approach. Nevertheless, RFT exhibits pronounced instability in its grounding capability, where rollout outcomes are highly imbalanced and fixed clipping can suppress valuable exploration. More importantly, policy probability alone does not reveal the semantic role of a grounding prediction. To address these challenges, we propose GUI-AC, a continual RFT framework that jointly characterizes grounding rollouts through advantage and policy probability. This yields a four-region view of grounding behavior and enables targeted optimization of Exploration and Overconfidence. GUI-AC further introduces grounding certainty to support two core mechanisms: (i) Adaptive Advantage, which strengthens the negative correction for overconfident grounding failures; and (ii) Dynamic Clipping, which relaxes the upper clipping bound for exploratory grounding rollouts. Extensive experiments show that GUI-AC consistently outperforms state-of-the-art baselines. Code is available at https://anonymous.4open.science/r/GUI-AC-Code.
For many optimization problems on random graphs, most prominently the largest independent set problem, natural algorithms reach only half the optimum, and such factor-two gaps are known to be sharp within broad algorithmic classes. We prove such a gap for finding large induced matchings in Erdős-Rényi random graphs. In the worst case, requiring a matching to be induced turns a polynomial-time solvable problem into an NP-hard one; our results reveal an average-case counterpart. In $G(n,p)$ with $C/n\le p\le 1-κ$, the largest induced matching has size $(1+o(1))L$ with high probability, where $L=\log_{1/(1-p)}(np)$. A simple online greedy algorithm finds an induced matching of size $(1/2-ε)L$, while no online algorithm, even with adaptive inspection order and unbounded computation, finds one of size $(1/2+ε)L$ except with probability $\exp(-Ω(L\log(np)))$. In the sparse regime $p=d/n$, low-degree polynomial algorithms exhibit the same sharp threshold. Our guarantees rest on a maximality principle: with high probability, every maximal induced matching is half-optimal. Greedy outputs a maximal induced matching and our low-degree algorithm a near-maximal one. We bypass the delicate coupling arguments classically used to analyze greedy since Karp. For low-degree hardness, we adapt Wein's overlap gap argument, with forbidden structures that track overlaps between endpoint sets rather than edge sets, since a new edge may join two previously used endpoints and thus encounter no fresh randomness. For online hardness, we give a new and significantly simpler argument that avoids the correlated instances of prior OGP-based proofs: at a suitable stopping time, all pairs involving uninspected vertices remain unrevealed, so a direct first-moment bound suffices.