Papers

Generative and embodied AI services exchange token streams within continuing inference and control loops. Their communication requirements depend on each token set's purpose, useful timing, and execution context. This article organizes these properties into service, temporal, and stateful semantics and proposes a token service interface (TSI) between applications, radio access networks (RANs), and edge runtimes. TSI binds delivery requirements, readiness forecasts, and execution-state references to each schedulable token set, specifying field ownership, versioned updates, and admission feedback. A drone inspection case study over a decoupled RAN illustrates importance-aware radio allocation, advance preparation for timely delivery, and selective state migration that balances interruption against forwarding delay.

Evaluating WebUI code generation at scale is difficult: outputs may compile and look plausible yet fail under user interaction, and prior benchmarks largely rely on free-form prompts with static checks (build success, screenshots) that miss functional correctness. We introduce WebUIProof, an execution-oriented benchmark that provides structured specifications and dense, executable interaction tests for WebUI generation across two task families: general WebUIs (e.g., dashboards, game, interactive tools) and 3D interactive simulation (e.g., particle/galaxy systems, physics dynamics). WebUIProof includes a UI-agent harness that runs executable interaction tests in a headless browser using an iterative plan–act–observe loop: it locates DOM elements, performs actions, observes resulting UI/DOM changes, and checks the specified assertions. We evaluate across eight commercial LLMs and observe frequent failures on interaction-based requirements even when pages render successfully, especially on 3D simulation interfaces. Finally, we show the UI-agent harness can provide outcome-level training signals. Training compact models (e.g., Qwen2.5 14B and MIMO 7B) with RL rewards derived from executable interaction tests improves functional completion while reducing build failures.

Harness evolution improves an LLM agent's prompts, tools, and workflow, while the optimizer's own tools and procedures often remain fixed. We study whether an optimizer can improve another agent more effectively by also improving how it diagnoses failures, develops edits, and tests their effects. Two observations guide our design. In a controlled study, optimizer self-evolution fails to improve performance without execution-based verification, but achieves the best result of that study when verification is available. Across five executors, self-evolving optimizers build their own tools for failure analysis, verification, training audits, and workflow control. Motivated by these findings, we introduce VERSE, a Verified Self-Evolving optimizer for agent harnesses. VERSE lets the optimizer test draft edits, replay failures, and perturb suspected steps before submission, while tracking fixes and regressions across rounds. Using this feedback, the optimizer revises both the executor harness and its own prompts, skills, tools, hooks, and notes, while the weights of the optimizer and executor models stay fixed. Under a shared protocol with disjoint training, validation, and test tasks, VERSE improves all four evaluated harness optimizers on held-out SWE-rebench tasks and newer out-of-distribution tasks in five languages. Its best validation-selected harness reaches 42.3% and 37.7% accuracy, respectively, against 39.2% and 29.3% for the strongest baselines. Code is available at https://github.com/wzekai/VERSE.

Learning research, as one of its central goals, tries to measure, model, and understand how learning-problem properties affect average-case learning performance. For example, we would like to quantify the value of constructive induction, noise filtering, and background knowledge. This paper describes the effective dimension, a new learning measure that helps link problem properties to learning performance. Like the Vapnik-Chervonenkis (VC) dimension, the effective dimension is often in a simple linear relation with problem properties. Unlike the VC dimension, the effective dimension can be estimated empirically and makes average-case predictions. It is therefore more widely applicable to machine and human learning research. The measure is demonstrated on several learning systems including Backpropagation. Finally, the measure is used to precisely predict the benefit of using FRINGE, a feature construction system. The benefit is found to decrease as the complexity of the target concept increases.

AI models serving a heterogeneous population must act on the principles appropriate to each user and context. While post-training has been shown to narrow the views large language models express, prior work has focused on default behavior rather than the ability to adapt to in-context information. We show that post-training also degrades a model's ability to be steered in-context toward perspectives it was not trained to favor. In controlled experiments, we fine-tune models toward one side of cultural-value disagreements and evaluate checkpoints throughout training. The trained side becomes increasingly dominant in ordinary use, while the ability to recognize and faithfully enact the opposing view declines. These findings point to a tension between prioritizing a single set of values and preserving the technical capacity needed to serve diverse stakeholders. Finally, we propose and analyze an alternative objective that maximizes reward subject to a prescribed distribution over expressed perspectives, and present stance-distribution matching as a practical implementation.

Simultaneous speech translation must emit useful target text before the source is complete while preserving every committed token. We adapt a full-utterance speech language model using prefix supervision derived from its own complete- and partial-waveform translations, requiring neither transcripts nor human translations. We compare single-turn forced-prefix and multi-turn append-only decoding, use a confidence threshold to control the inference-time quality–latency trade-off, and vary the density of training prefixes with a separate synthesis margin. On FLEURS and CoVoST2 in three language directions, prefix training improves quality–latency frontiers over the unadapted model, and confidence provides the broadest consistently competitive operating range. Multi-turn decoding is generally stronger at low latency; under multi-turn training, commit-calibration error falls by 63–68% overall and 68–80% at early prefixes, whereas single-turn training provides only modest overall calibration gains and no early-prefix improvement. A small synthesis margin sometimes extends the frontier to lower latency, particularly on shorter utterances, while a larger margin degrades translation quality and calibration. Prefix adaptation therefore improves simultaneous speech translation, especially under multi-turn append-only decoding, while synthesis density introduces a non-monotonic quality–latency trade-off.

Fine-resolution precipitation estimates support flood risk assessment and water management, but coarse satellite products cannot resolve rainfall within each grid cell. Generative models address this ambiguity by producing ensembles of plausible high-resolution rainfall fields. Among these models, rectified flows generate samples by iteratively transforming random noise into rainfall fields. Reducing the number of sampling steps accelerates generation but can make ensemble members too similar, understating uncertainty. We propose a scale-recursive rectified flow that generates broad patterns before local details and guides sampling-step allocation by comparing ensemble variability with prediction error across spatial scales. Validation scores and rainfall power spectra constrain the allocation to avoid excessive amplification. In satellite-to-radar downscaling over the contiguous United States, our analysis identified broad rainfall patterns as the main source of insufficient ensemble variability under reduced sampling budgets. Allocating more steps to the coarse flow improved probabilistic accuracy and rain detection across training seeds at fixed architecture and computational cost. The proposed model also achieved better probabilistic accuracy with shorter sampling time than a nonrecursive flow using more steps.

The research presented here focuses on modeling machine-learning performance. The thesis introduces Seer, a system that generates empirical observations of classification-learning performance and then uses those observations to create statistical models. The models can be used to predict the number of training examples needed to achieve a desired level and the maximum accuracy possible given an unlimited number of training examples. Seer advances the state of the art with 1) models that embody the best constraints for classification learning and most useful parameters, 2) algorithms that efficiently find maximum-likelihood models, and 3) a demonstration on real-world data from three domains of a practicable application of such modeling. The first part of the thesis gives an overview of the requirements for a good maximum-likelihood model of classification-learning performance. Next, reasonable design choices for such models are explored. Selection among such models is a task of nonlinear programming, but by exploiting appropriate problem constraints, the task is reduced to a nonlinear regression task that can be solved with an efficient iterative algorithm. The latter part of the thesis describes almost 100 experiments in the domains of soybean disease, heart disease, and audiological problems. The tests show that Seer is excellent at characterizing learning-performance and that it seems to be as good as possible at predicting learning performance. Finally, recommendations for choosing a regression model for a particular situation are made and directions for further research are identified.

Time series forecasting (TSF) increasingly drives decisions in transportation, energy, finance, healthcare, and infrastructure, yet current evaluation remains overly narrow: standard benchmarks reward low held-out error, while robustness studies typically reduce failure to Gaussian noise, random masking, or bounded adversarial perturbations. This obscures the real failure modes of deployed forecasting systems. Input-side anomalies are not merely noisier inputs: they often reflect structured events that alter temporal dynamics, break cross-variable dependencies, induce regime shifts, or propagate from faulty sensors to downstream decisions. These semantic, causal, and system-level failures cannot be faithfully captured by i.i.d. perturbations alone. The rise of TSF foundation models makes this evaluation gap more urgent, as unauditable pretraining corpora make held-out generalization increasingly unreliable. We therefore advocate scenario-grounded stress testing. Each test instance should include historical inputs and future targets, together with a semantic scenario, an explicit failure operator, and a measurable difficulty level. This shift makes evaluation interpretable, attributable, and deployment-relevant and friendly, enabling the community to ask not only which model is accurate, but under what conditions it fails and why.

Modern query optimizers use analytical cost models to estimate the cost of a given query plan. Such cost models are typically functions of a set of "cost units" that specify unit CPU cost when processing a row or unit IO cost when accessing a disk page. These cost units are traditionally viewed as platform-dependent constants, that is, they require a one-shot calibration when a database is deployed on a hardware/software platform, but are fixed afterward regardless of the query being optimized for. Some very recent work has taken a different perspective by viewing these cost units as tunable parameters that we call "cost model parameters (CMPs)" in this paper. However, so far there is no approach that offers any optimality guarantee for the tuning results. We present a new approach to systematically explore the query plan space spanned by the CMPs and find the best plan in terms of execution time. Compared to alternative exploration approaches that use random search (RS) or Bayesian optimization (BO), our new approach is deterministic and, therefore, avoids the undesirable instability that is inevitable when applying RS or BO. Moreover, it is guaranteed to find all candidate plans in the query plan space without suffering from the overhead of an exhaustive enumeration. We also present a set of optimization techniques to reduce the overall evaluation time spent on executing the candidate plans found, a factor that is often overlooked by previous work but is critical from a practical point of view. Experimental evaluation on top of PostgreSQL and Microsoft SQL Server demonstrates the efficacy of tuning the CMPs, which can find query plans that are orders of magnitude faster in execution time than the ones found by RS or BO.

Large language models are increasingly used as math tutors, raising two questions: can a single LLM detect a student's error, diagnose the underlying misconception, and remediate it through multi-turn dialogue - and can we trust the numbers that report how well it does? We address both with MISCONCEPTDIALOGUE, a fully automated framework that runs the full pipeline on five open-weight LLM tutors (7B-14B) interacting with LLM-simulated students over 5,000+ dialogues - a deliberately simulated protocol (paired conditions, a wrong-diagnosis control unsafe with real learners, scale) whose results do not establish real-student learning. On measurement, our headline finding is uncomfortable for the field: benchmark-style exact-match scoring common in standard math NLP - a lower-bound stress test - reports 42.1% measured Remediation Success Rate on identical dialogues where a human-calibrated LLM-as-judge reports 81.3%, a 39.2 percentage-point spread larger than the combined effects of tutor model, teaching strategy, and supplied diagnostic information. This is not a single-judge artifact: four independent LLM judges and 200 human annotations converge to within a few points (kappa up to 0.91). On capability, the verdict is equally striking: the score barely changes whether we tell the tutor the student's actual misconception or a deliberately swapped one - under the tested diagnosis-presentation format, these tutors show little evidence of reliably conditioning on the diagnostic information they are handed. For LLM tutoring evaluation to mature, multi-criterion scoring with multi-judge convergence and human anchoring should become standard practice - and the measurement question deserves as much scrutiny as the capability question it claims to answer.

Many LLM studies compare how models respond to different follow-up prompts. For example, one prompt may neutrally ask a model to reconsider, while another insists that an invented film really exists. The difference is used to measure the effect of conversational pressure. An automatic scorer then labels whether each response makes an unsupported factual claim. Researchers commonly validate the scorer with one overall accuracy or $F_1$ score. However, high overall performance can hide prompt-specific mistakes: if the scorer makes more errors after one prompt, the measured difference can be too large, too small, or point in the wrong direction. An audit containing mostly difficult responses can also misrepresent performance in the full study. We derive how such errors alter a prompt comparison and examine the problem using six models, 180 invented-entity questions, and six prompt conditions. In a balanced audit of 288 responses, overall $F_1$ was $.916$. Yet, depending on the prompt, the scorer flagged 2.6%-33.3% of responses that the primary LLM rater judged not to contain an unsupported claim; these prompt-specific estimates were uncertain. In a second audit, the scorer and two separate LLM raters evaluated the same 300 model-question cases under each prompt. An apparent 9.3-point increase from asserting certainty rather than neutrally asking the model to reconsider fell to 2.3 points under both raters. The primary-rater interval included zero, so it did not establish an increase. Two other prompt comparisons kept their direction under both raters, while an attempted prompt-specific correction remained too uncertain for a firm conclusion. Researchers should therefore validate the exact difference they plan to report, not just the scorer as a whole: audit both prompts on matching cases and include scorer and rater uncertainty in the final result.

Radio-frequency (RF) and wave-based imaging systems play a critical role in sensing under adverse conditions, but are fundamentally constrained by a trade-off between collection speed and spatial resolution. Reducing measurements to enable rapid collection of fragment geometric information across sensing dimensions, leading to highly non-injective inverse problems that destabilize conventional image-domain reconstruction and super-resolution methods. Synthetic aperture radar (SAR) exemplifies this challenge in sparse-aperture sampling. We introduce SPIRIT, a physics-informed inverse rendering framework for sparse-aperture SAR. Rather than enhancing formed images, SPIRIT reconstructs an explicit three-dimensional surface. By adopting an optimization strategy and explicitly modeling visibility, phase coherence, and wave propagation, SPIRIT aggregates distributed geometric features across sparse aperture views into a coherent reconstruction. Extensive experiments on synthetic and real-world data demonstrate that SPIRIT recovers accurate geometry and signal under extreme sparsity, reducing data collection by over 90% while achieving low geometric and signal error and consistently outperforming baselines.

Calibrating expensive engineering simulators requires tuning model parameters to match reference data, a process heavily constrained by computational costs. This study investigates whether Large Language Models (LLMs), conditioned on physical diagnostic descriptions of simulation errors, can propose informative parameter updates more efficiently than conventional derivative-free optimizers. We test this on a cylinder-wake Large-Eddy Simulation and a damped-oscillator inverse problem. Physics-labeled prompts are evaluated against four text ablations (removed, anonymized, padded, misdirected) and standard optimizers (Nelder-Mead, Bayesian optimization, CMA-ES, and random search). Results demonstrate a strong early semantic advantage: the first physics-informed LLM proposal improves the initial design in 23 of 25 trials. After five evaluations, the median calibration error is 21.9 times lower than physics-free controls. However, this advantage is transient - disappearing by fifteen evaluations - and inherent LLM decoder variability yields a 9.8-fold performance spread. To address this, we introduce a hybrid workflow that uses the LLM as a warm-start, passing accumulated evaluations to a conventional optimizer. This two-stage numerical continuation policy reaches target accuracy in just 9 simulator runs, outperforming the LLM alone (12 runs), Nelder-Mead (17 runs), and Bayesian optimization (>60 runs). Supported by sensitivity analysis validating the test cases, this hybrid physics-labeled LLM scheme proves highly effective as a calibration agent under strict evaluation budgets.

Sampling from continuous flow-based generators such as flow matching models requires numerical ODE or SDE integration, making inference cost depend on the number of function evaluations (NFEs). In practice, the sampling budget is inherited from prior work or selected through expensive FID-versus-NFE sweeps, using FID to assess image-generation quality, requiring sample generation and evaluation at multiple budgets. This makes sampling-budget selection dependent on repeated empirical evaluation. We therefore ask a complementary question to solver design and timestep allocation: for a fixed pretrained model and solver, what numerical resolution is sufficient? We introduce FLOWBUDGET, a training-free estimator of solver-relative sampling complexity. Using simulation-free interpolation probes of the learned dynamics, FLOWBUDGET normalizes the leading one-step discretization error by the motion, estimates its coefficient from interpolation probes, and maps a numerical tolerance to a step budget using the solver order. We derive estimators for numerical solvers in flow~matching sampling, namely Euler and Heun for ODE sampling and additive-noise Euler–Maruyama for SDE sampling, and connect the ODE coefficients to terminal integration error. Experiments on SiT, FlowDCN and Rectified Flow models, spanning flow-model architectures and formulations, test whether FLOWBUDGET can estimate the NFE budget without generating samples at multiple budgets or performing an FID-versus-NFE sweep. The code is available at https://github.com/Yasaman-dt/FlowBudget

Permutation generation with language models requires both text-conditioned ranking and a global one-to-one assignment between candidates and output positions. We study this problem through three diffusion-LLM parameterizations: hard-assignment prediction over shuffled candidate-index sequences, soft-assignment prediction over doubly stochastic matrices, and direct ordering-score prediction followed by sorting. The three families place permutation constraints at different stages of prediction and decoding and therefore support different training objectives and sampling procedures. We evaluate these designs on recommendation reranking across three Amazon domains with candidate sets of size 10 and 20. At one function evaluation, the hard- and soft-assignment variants achieve Recall@1 comparable to a constrained masked-diffusion baseline, whereas direct ordering-score prediction guarantees valid permutations without Hungarian assignment or Sinkhorn normalization. Additional function evaluations do not consistently improve ranking quality, and their effect depends strongly on the state representation and training objective. These results show that corruption, prediction parameterization, objective, and decoder should be designed jointly for permutation diffusion with language-model denoisers.

Token-selection methods for on-policy distillation (OPD) prioritize student-visited positions using diverse signals, including uncertainty, teacher-student disagreement, and mismatch direction. We instead derive token importance from the distillation objective itself: how much can each position's loss decrease under an equal local KL budget on the student's predictive distribution? The answer is the utility $s_t=g_t^\top F_t^\dagger g_t$, where $g_t$ is the logit gradient and $F_t$ is the categorical Fisher matrix. For reverse-KL it equals the student-weighted variance of the student-teacher log-probability ratio, and for forward-KL it equals the Pearson $\chi^2$ divergence, so the appropriate utility follows from the choice of objective. On shared training checkpoints, several independently motivated criteria, including student entropy, KL disagreement, and TIP, concentrate their importance on nearly the same positions as the derived utility at a 50% selection fraction, while diverging at stricter fractions and for direction-specific criteria. We instantiate the utility as Fisher Utility Weighting (FUW), a continuous, mean-normalized token weighting. Across seven mathematical reasoning benchmarks and three Qwen3 student-teacher pairs, FUW improves over matched uniform OPD under both reverse-KL (1.35-4.0 macro points) and forward-KL (1.08-2.71 macro points), and is competitive with existing selective methods using a single closed-form score derived from the training objective.

Voice agents built on large language models (LLMs) now handle customer calls, but in regulated applications completing the task is not enough: the agent must follow a conversational flow set by law. In debt collection, it must confirm it is speaking to the debtor before mentioning the debt, identify itself and the creditor, state the amount, and avoid prohibited language. Every creditor turns these rules into its own procedure and revises it. We ask whether a small, self-hosted policy can be trained once and then follow regulated procedures it has never seen. One machine-readable specification per deployment, FlowSpec, generates the policy instruction, gates the tools, supplies the supervision and defines the tests; the agent speaks only approved catalogue sentences. We fine-tune a 9-billion-parameter model on 2,506 conversations from 297 randomized procedures and evaluate a Thai outbound-call agent on four unseen deployment procedures (three debt collection, one non-debt), one from a wholly withheld lineage. Documented checks, fixed by the procedure, rise from 67–71 percents to 93–100 percents, while grounded checks, fixed by the conversation, already pass at 97–100 percents before training; scenario pass rate rises from 15/43 to 40/43. An audit of the shared specification finds 11 defects, 3 of which its own metric did not expose.

High-throughput perturbation screens increasingly use machine-learning models to predict the effects of gene combinations that were never measured. A difficult case arises when both genes in a test pair were seen individually during training, but their combination was not. In this setting, standard OOD methods often fail because neither gene appears unfamiliar on its own. We therefore compare each model prediction with a simple additive prediction obtained by summing the two observed single-gene effects. The difference between them, which we call additive disagreement (AD), measures how much extra interaction effect the model introduces for an unseen combination. Across eight prediction methods and three Perturb-seq datasets, specialised neural predictors do not consistently outperform this simple additive baseline. More importantly, predictions with larger deviations from the additive baseline tend to have larger errors. Conventional output-based and embedding-based confidence signals are inconsistent across datasets and models, whereas AD remains informative across all evaluated compositional settings. We further study how this signal can guide decisions in an agentic perturbation-screening workflow. The code is available at https://github.com/zygao930/OOD.

Protein function annotation operates over an expanding label space, making class-incremental learning (CIL) a natural framework for incorporating newly discovered functions without full retraining. Yet existing CIL methods have been developed largely in vision and have not been systematically evaluated over frozen protein representations. We conduct a controlled benchmark of nine representative CIL methods across five sequence-, structure-, and sequence–structure-based protein encoders and two enzyme function datasets at different scales, while keeping the encoders frozen and evaluation settings consistent across representations. We find that no method transfers reliably across encoders, and that these failures persist across alternative phase constructions. Diagnostic analyses reveal that this instability arises from weak gradient-based stability signals, substantial heterogeneity in embedding geometry, and covariance assumptions whose appropriate capacity changes with both dataset scale and encoder. These findings suggest that the representation itself should be treated as a central component of continual-learning design rather than a passive feature extractor. More generally, reliable protein CIL will require encoder-aware methods that adapt their statistical assumptions and capacity to the geometry of the representation on which they operate. Code is available at https://github.com/zygao930/ProteinCIL.

Post-training reasoning LLMs often requires making a tradeoff between costly supervision and compute inefficiency: supervised finetuning typically requires reasoning traces, which is data inefficient since in many applications one only has prompt-answer tuples and lacks the groundtruth chain-of-thought (CoT) sequence. On the other hand, reinforcement learning (with verifiable rewards, for example) can operate without CoT data but it is much more compute-inefficient since it requires multiple autoregressive completions per optimization step, none of which are guaranteed to give a non-zero policy gradient. In this work we explore a middleground, a self-distillation technique that uses answer-only data, treating CoT as latent variables. This relies on training a partner “inverse reasoning LLM" that generates the original's reasoning traces conditioned on the answer. The inverse model is trained once and reused across incoming batches of answer-only data, with all generation happening offline and up front, making the approach particularly suited to continual post-training. Using Qwen3-0.6B on the Countdown and GSM8K tasks, we show that after GRPO training has saturated, a single offline distillation round from the inverse model roughly doubles pass@1 on hard prompts and, on Countdown, outperforms rejection-sampling self-distillation on every hard-tail metric at every mix ratio (with smaller gains on GSM8K); rejection sampling by construction has no signal on problems the model never solves on its own. Continued GRPO on the fresh data retains an edge at hard pass@1 but trails distillation at pass@8 and pass@32 while regenerating rollouts at every step. As an added bonus we see that the inverse model has a second application as an accurate verifier of correctness. When used for best-of-$n$ selection at test time, it raises selection accuracy on Countdown from 0.753 (majority voting) to 0.969.

Time-series language models typically align numerical signals with language using labels, captions, or question-answer pairs. Although such supervision encourages linguistic representations, it does not explicitly require them to preserve information about how the underlying time-series signal evolves. We introduce forecast grounding, a pretraining principle that requires language-aligned time-series representations to remain predictive of the future evolution before they are adapted to downstream tasks. We instantiate this principle in LOTUS, which divides each time-series channel into non-overlapping patches and maps them into the LLM embedding space. It then predicts the next patch from the LLM hidden states of the time-series tokens, encouraging the learned representation to retain temporal information. To capture both temporal changes and periodic patterns, we also propose a temporal-spectral patch representation that combines each raw time-series patch with its spectral representation before mapping it into the LLM embedding space. Across three language model backbones and six downstream evaluation datasets, LOTUS achieves the best average performance with minimal trainable parameters, reaching 59.91% accuracy and outperforming the strongest matched baseline by 3.36%. Our representation analyses further show that LOTUS uses the numerical input and preserves temporal structure in its learned patch representations. These results establish forecast grounding as a promising pretraining principle, enabling compact time-series language models to outperform models 25 to 50 times their size on TSRBench.

Professionals are increasingly turning to artificial intelligence (AI) for career advice. We tested whether large language models (LLMs) provide different advice to female and male job candidates in the context of career negotiations. Across two large-scale audit experiments, including four leading models, the LLMs not only recommended lower salary targets for female than male personas, but also constructed distinct negotiation agendas, prioritizing flexibility, caregiving, and well-being more highly for female candidates, and financial rewards, mobility, and advancement more highly for male candidates. These differences closely tracked independent human ratings of employment terms associated with primary caregiver and breadwinner roles, respectively. Our findings identify a subtle pathway through which personalized AI advice may reinforce gender inequality at work: steering otherwise equivalent users toward traditional gender roles.

Contact-rich manipulation requires a robot to be precise in free space to establish the intended contacts, and compliant once in contact to avoid excessive forces. Imitation learning has proven effective for simpler manipulation tasks, but the compliance underlying a demonstration is not directly observable and is therefore difficult to infer from demonstration data alone. We present a framework that makes compliance explicit at every stage. During data collection, the operator teleoperates the real robot while a synchronized digital twin visualizes the error between the commanded reference and the achieved pose, making the current compliance visible, and selects between two discrete compliance modes. To generalize across object poses, we combine Laplacian editing, which retargets trajectories efficiently but is not physics-consistent during contact, with simulation-based trajectory optimization, which enforces physical plausibility but offers no manual guidance. Finally, a residual reinforcement learning policy refines the interaction forces by adjusting the reference pose of the underlying admittance controller. We evaluate on three flipping and pushing tasks on a Franka FR3 based on a total of 7 demonstrations across tasks. Our method is successful on all 40 out-of-distribution trials, compared to 33/40 and 24/40 for two baselines, with our base policy alone reaching 35/40.

Foundation models are expanding robot capabilities, enabling robots to handle more diverse objects, follow sophisticated instructions, and operate in complex environments. These advances bring us to a future in which robots become part of everyday household life. However, cluttered, dynamic household environments also expose robots to unfamiliar objects, tasks, and interactions, making reliable zero-shot task completion an ongoing challenge for robotics research. From a security perspective, we argue for open-world safety: household robots should preserve specified safety requirements in unseen situations, including when they cannot complete the requested task. Safety protections must account for context and interactions that persist across actions and tasks. In this position paper, we examine why existing safety methods can fail to provide this protection. Protection can fail through misinterpreted context, overlooked past interactions, or stopping and recovery strategies that become unsafe under unfamiliar conditions. We therefore propose a research agenda to specify, learn, and ground safety requirements, enforce them with reliable recovery, and evaluate the generalization of such protections. We call for household robots whose safety protections remain reliable beyond the situations, context, and history used to train and develop them, to achieve open-world safety.

Recent progress in learned whole-body controllers for humanoid robots has enabled robust free-space motion by tracking kinematic targets such as joint or end-effector trajectories. This interface provides a useful abstraction for upstream systems, including motion capture, teleoperation, or generative models, which can specify desired motion without reasoning about forces or balance. However, this abstraction has largely been limited to robot motion, requiring task-specific controllers or retraining for object interaction. We introduce OIL, an Object Interaction Layer that extends the kinematic command interface of whole-body control to include desired object motion and contact. OIL augments a standard controller with object-centric commands and translates them into coordinated whole-body behavior while preserving locomotion and balance. This design enables a single learned policy to execute non-prehensile interactions, including pushing, pulling, two-arm pick-and-place, and locomotion-coupled transport, directly from object-level kinematic targets without task-specific retraining. In simulation, we demonstrate that object-level commands improve performance over robot-centric controllers while matching task-specific policies across variations in object properties such as mass and size. We further demonstrate zero-shot generalization to novel behavior combinations and sim-to-real transfer on the Digit humanoid robot.

Multimodal large language models (MLLMs) often answer visual reasoning questions by relying on linguistic priors rather than task-relevant visual evidence. Textual chain-of-thought reasoning can partially mitigate this issue by encouraging models to decompose visual questions into intermediate evidence-seeking steps, but generating these steps autoregressively increases inference cost. Latent reasoning avoids explicit rationale generation, but existing approaches provide limited control over what intermediate states encode, making it difficult to impose separate supervision for planning, grounding, and evidence selection. We propose Structured Latent Visual Reasoning (SLVR), a training framework that bridges explicit chain-of-thought and latent reasoning by organizing multimodal reasoning into typed latent stages for planning, grounding, evidence selection, and reasoning integration. SLVR first trains the model to rely on the image by masking answer-revealing text and contrasting the correct answer with visually plausible distractors. It then organizes reasoning into latent stages for planning, grounding, evidence selection, and integration, supervising each stage with the corresponding signal: plans, boxes, visual evidence, and final rationales. This gives latent reasoning an explicit functional structure while avoiding generated textual chains at inference time. Built on Qwen2.5-VL-7B, SLVR improves consistently across multimodal reasoning benchmarks, with absolute gains of +9.4 on MMVP and +14.2 on BLINK Relation, as well as improvements on V*, MathVista, and ChartQA. These results suggest that structured latent supervision can improve fine-grained visual reasoning without the decoding overhead of textual CoT. Project page is available \href{https://bogao-code.github.io/SLVR/}{here}.

AI foundation models pretrained on weather and climate data are increasingly fine-tuned to Earth science tasks well beyond weather forecasting. Their development and adoption are outpacing the scientific community's ability to evaluate them. These models are judged almost entirely by benchmark skill metrics, which measure how closely a forecast reproduces a reference product but not whether a model represents the physical processes governing the system it predicts. Forecast skill and physical reliability are therefore distinct properties. The distinction is most consequential under the nonstationary conditions of a changing climate for which these models were never trained. We identify five priorities for the physical evaluation of AI models in Earth science from task-specific emulators to foundation models, spanning training data, fine-tuning, behavioral testing, mechanistic interpretability, and output validation. We recommend three activities for the coming decade: 1) open AI-ready evaluation datasets, 2) a shared reporting standard for physics-based evaluation, and 3) a dedicated research program on the safety of these models.

Spoofing-aware speaker verification (SASV) must confirm who is speaking and that the speech is genuine, but current systems gain one at the expense of the other. Modular cascades are accurate yet run two large self-supervised encoders, whereas end-to-end models are efficient but lose accuracy when a single embedding has to serve two conflicting objectives. We show that this trade-off between efficiency and specialization is avoidable. A single frozen W2V-BERT~2.0 backbone is adapted per task by deep hybrid Wavelet Prompt Tuning (WPT), so each branch obtains its own view of the shared encoder through dedicated prompts and a task head, and the scores are combined only at inference. Training only about 7M parameters, a small fraction of those a strong two-encoder cascade requires, the system surpasses it on SpoofCeleb evaluation with 0.03% CM-EER and 0.038 min a-DCF against 0.16% and 0.047. Because the backbone is shared and frozen, new branches attach without retraining the heads and prompts already in place. On ASVspoof5, with adversarial attacks and codec distortions, it reaches 0.092 min a-DCF and 3.60% CM-EER using only the provided data, indicating that the design holds under realistic in-the-wild threats.

Detecting synthetic and voice-converted speech remains difficult for low-resource languages with dialectal diversity, where systems must generalize across regional dialects and unseen acoustic channels. We present the UZH-CL submission to the ArA-DF 2026 Shared Task on Arabic speech deepfake detection, covering Track~1 (dialect generalization) and Track~2 (acoustic robustness). We freeze a W2V-BERT-2.0 backbone and adapt it with Wavelet Prompt Tuning, updating under 1% of parameters, and aggregate multi-layer representations with cross-layer attention and a general attentive-statistics pooling head rather than a specialized graph backend. Complementary detectors are obtained by varying adaptation strategy, training data, augmentation, and encoder family. We find that the two shift types require different fusion regimes: a broad multi-window ensemble for dialect generalization, and a compact, channel-matched, center-crop ensemble for acoustic robustness. Official evaluation yields 1.96% EER on Track~1 (6th place) and 1.04% EER on Track~2 (3rd place), corresponding to 87% and 96% relative reductions over the XLS-R+AASIST baselines.