The manifold hypothesis suggests that data lies on manifolds with smaller intrinsic dimension (ID) than their ambient dimension. However there is no empirical agreement on the estimates for ID from different estimators for realistic datasets. Thus it is important to test ID estimators (IDEs) with targeted stressors. In this work, we consider the role of anisotropy. To this end, we propose HomID, a collection of homogeneous spaces with anisotropic embedding, for benchmarking ID estimators. We observe that methods that perform well on standard benchmarks systematically degrade on HomID under identical resource allocation. We further observe that anisotropic distortion of such benchmarks also results in performance degradation. Finally, we demonstrate that controlled anisotropic distortions induce systematic shifts in the distributions on which these methods rely, providing a concrete mechanism for the resulting estimation errors in two particular IDEs. The code for the project is available at https://github.com/a-writes-code/homid
Diffusion-based generative models usually perturb data through Markov processes driven by Brownian motion. Volterra processes allow more flexible temporal dependence, but their non-Markovian structure complicates score learning and reverse-time simulation. We develop Volterra generative models and provide Markovian approximations. Gaussian conditioning then reduces the denoising problem to the original data dimension. We present the connections with variance-exploding and variance-preserving representations. Our framework also permits reuse of pretrained denoisers such as EDM2 networks. With one auxiliary factor, our models trained from scratch achieve FIDs of $0.547$ on MNIST and $1.860$ on conditional CIFAR-10. With autoguidance, ImageNet64 FID reaches $1.912$ after fine-tuning, while ImageNet512 FID reaches $1.903$ using frozen pretrained networks.
This paper presents a layered control framework that recasts high-dimensional system dynamics as a composition of dynamically interconnected subsystems, with dynamic dual-arm manipulation on the humanoid serving as a representative application. The humanoid is decomposed into a single rigid body (SRB) representing the torso and floating-base dynamic models representing the articulated arms and lower-body. The rigid connections between arm and torso and the closed kinematic chain formed between the end-effectors during cooperative manipulation are represented as holonomic constraints. The corresponding Lagrange multipliers parametrize the interaction wrenches transmitted at the shoulders and between the end-effectors. The high-level nonlinear model predictive control (NMPC) jointly optimizes the subsystem state trajectories, arm torques, ground reaction forces (GRFs), and interaction wrenches, enabling online wrench design and optimization while preserving the dynamic coupling among the independently modeled subsystems. The optimized arm torques are applied directly for cooperative dual-arm manipulation, whereas the prescribed GRFs are mapped to lower-body joint torque for whole-body coordination. We validate the proposed framework through numerical simulations and hardware experiments on a G1 humanoid robot. The results demonstrate optimal interaction wrenches consistent with simulated contact measurement, cooperative dual-arm trajectory tracking, and stable whole-body coordination while executing the NMPC in online manner.
Gautam Chandrasekaran, Jesse Goodman, David Zuckerman
A quadratic source is a random variable $X\sim\mathbb{F}_2^n$ generated by applying $n$ quadratics to a hidden uniform string. Previously, the best-known explicit extractor for this model required min-entropy roughly $k=n-\sqrt{\log n}$. In this paper, we give an explicit construction for min-entropy $k=\mathrm{polylog}(n)$. Even for the special case of $2$-local sources, where each bit of $X$ depends on two hidden uniform bits, achieving an entropy requirement below $k=\sqrt{n}$ was an important open problem. This is because such sources provably cannot be reduced (in a black-box manner) to our most powerful general-purpose seedless extractor, known as a sumset extractor. Our new extractor bypasses this barrier, providing the first explicit construction to go beyond sumset extraction.
EHR foundation models read medical events within a fixed context window, and long context has so far been measured only in tokens. How far back in time a model must look to capture the earliest predictive signal is unknown. We introduce time-context, the calendar lookback window before the prediction point, and provide, to our knowledge, the first systematic evaluation of decade-scale time-context for an EHR foundation model. We pretrained a decoder-only transformer (34.6M parameters) on 1.4B coded events from 3.6M primary-care patients in North West London and fine-tuned it to predict incident colorectal cancer under a 12-month pre-diagnostic buffer. A frozen cohort of 1,008 cases and 9,582 matched controls, all registered for at least 30 years, was evaluated at 15 nested lookback windows from 2 to 30 years, so that every truncation is compared within the same patients. Discrimination rose from an AUROC of 0.551 at 2 years to 0.590 at 18 years, then plateaued, with about 80% of the gain reached within 10 years; an XGBoost baseline converged to the same level (peak AUROC 0.589). A simulation further suggests that rotary position embeddings (RoPE) over elapsed time alias beyond about 20 years at the default base frequency, motivating a retuned $θ$. Time-context is a measurable design axis for clinical prediction models, distinct from token-context, and should be reported.
Image quality assessment predicts a mean opinion score (MOS), but for image aesthetics the mean hides the individual, so personalized assessment must start from the few images a new user rates. Fitting each person from scratch fails with ten ratings; networks fine-tuned per person carry over poorly to new kinds of images. We present Undertone-P, which learns from other raters a Gaussian prior over per-person regression weights on read-outs of frozen vision-language model (VLM) judges: the probability-weighted rating and the hidden state at the score token. Since people differ from the crowd in ways other raters reveal, a few ratings need only choose among those ways: a new user costs one closed-form solve, and no network is trained. On Ryu and Yanaka's PARA benchmark, the prior alone reaches a Spearman correlation of .640, above the best published 10- and 100-rating results (.581 and .620); 10 and 100 of a person's ratings raise it to .648 and .676. These ratings add most where tastes depart from the crowd: with 100 ratings, +.067 for the raters who agree least with it and +.206 on LAPIS artworks, against +.037 overall. Without labels, the judges' expected scores predict PARA's MOS at .835 (best published model we ran: .694). PARA-trained networks still lead under PARA's original protocol, likely through a stronger end-to-end MOS model.
Clariele Almeida, Renan Beserra, Rafael Moura, Danilo Araújo, Bruno Nogueira, Sérgio Chevtchenko, Ermeson Andrade
The rapid expansion of the Internet of Things (IoT) has increased the volume of data generated at the network edge, where many devices operate under severe memory, processing, and energy constraints. In this context, Tiny Machine Learning (TinyML) enables machine learning (ML) tasks to be executed on microcontrollers, smart sensors, and other resource-constrained platforms. However, the literature remains heterogeneous regarding the algorithms evaluated, energy-measurement methods, reported metrics, influencing factors, and optimization strategies. This study presents a systematic mapping of energy-aware TinyML research. It identifies the main models and learning approaches, classifies energy-evaluation methodologies and metrics, examines the technical factors affecting energy consumption, and organizes the reported optimization techniques into a structured taxonomy. The findings highlight the frequent use of convolutional and lightweight architectures, direct measurements and analytical estimations, and factors related to model architecture, hardware, communication, numerical precision, and input characteristics. The optimization strategies were grouped into model compression, hardware-aware design, system and communication optimization, and data and processing optimization. The results also reveal limited standardization in experimental platforms, measurement procedures, and reporting practices, limiting reproducibility and direct comparison across studies. This mapping provides an updated overview of the field and guidance for developing and evaluating more energy-efficient TinyML solutions.
Parallel block drafting accelerates speculative decoding by predicting multiple future positions in a single backbone pass, but on-device deployment constrains both drafter capacity and context storage. Under tight memory budgets, drafters must model long-range dependencies and fine-grained draft-position interactions without the growing KV-cache overhead of full-context attention. We propose HyFlash, a compact parallel drafter that introduces a prefix-memory prior to separate long-range context modeling from fine-grained draft-position interactions with limited capacity. Local attention captures recent context and draft-position interactions, while recurrent memory compresses historical target features into fixed-size states. This inductive bias enables effective local and long-range modeling with bounded draft-side memory. Experiments show that HyFlash achieves higher acceptance than full-context attention with substantially less KV storage and with larger gains for shallow drafters and long contexts.
Jiajun He, Denis Blessing, Mouyang Cheng, Yuanqi Du, Carles Domingo-Enrich
Sampling from discrete, unnormalized distributions without access to data is a challenging problem. Neural samplers offer a promising approach by training generative models from density evaluations directly. However, existing discrete neural samplers are prone to mode collapse, come without convergence guarantees when trained via fixed-point iterations, and are often tied to a specific reference process such as masked or uniform diffusion. In this work, we introduce Discrete Gibbs Iterative Neural Sampler, a fixed-point neural sampler that addresses these limitations, enabling efficient, scalable learning, substantially reducing mode collapse in practice. Besides masked diffusion, the framework also extends to transport between arbitrary distributions. DIG-IN mitigates mode collapse on Ising and Potts models and scales to amortized sampling with up to 2,500 sites, recovers alloy phase diagrams, and learns to sample biomolecular Boltzmann distributions from energy alone. Its transport formulation accurately estimates solvation and conformational free energies. As a maximum-entropy RL policy, DIG-IN outperforms the Gaussian, diffusion, and flow policies we compare against across 20 continuous-control tasks.
Vinay Rodrigues, Sara Zeynalpour, Gokhan Solak, Sebastian Hjorth, Arash Ajoudani
Deformable linear object (DLO) manipulation is challenging because high-dimensional configurations, nonlinear contact dynamics, and diverse material and geometric properties produce a broad range of object responses that are both difficult to model and data-intensive to learn. This work studies reinforcement learning for planar manipulation of a one-end-grasped DLO, focusing on training efficiency, observation design, robustness, and zero-shot sim-to-real transfer. We compare three Proximal Policy Optimization training stacks and find that the GPU-parallel MuJoCo Playground stack enables faster convergence and more consistent performance than CPU-based alternatives under the same wall-clock budget. Relative keypoint observations and an asymmetric actor-critic formulation further improve learning efficiency. Policies trained with observation noise, delay, domain randomization, and disturbances achieve simulation success rates of 96.4% for single-point tracking, 98.0% for disturbance recovery, and 88.8% for two-point tracking. Using markerless RGB-D keypoint estimation, the policies transfer without real-world fine-tuning, succeeding in 9/10, 10/10, and 8/10 physical trials, respectively.
Magdalena Proszewska, Tomasz Danel, Antoni Antoszek, Dawid Rymarczyk
Understanding the reasoning behind deep learning model predictions is crucial in cheminformatics and drug discovery, where molecular design determines their properties. However, current evaluation frameworks for Explainable AI (XAI) in this domain often rely on artificial datasets or simplified tasks, employing data-derived metrics that fail to capture the complexity of real-world scenarios and lack a direct link to explanation faithfulness. To address this, we introduce B-XAIC, a novel benchmark constructed from real-world molecular data and diverse tasks with known ground-truth rationales for assigned labels. Through a comprehensive evaluation using B-XAIC, we reveal limitations of existing XAI methods for Graph Neural Networks (GNNs) in the molecular domain. This benchmark provides a valuable resource for gaining deeper insights into the faithfulness of XAI, facilitating the development of more reliable and interpretable models.
Multilingual large language models often produce inconsistent reasoning and answers for semantically equivalent prompts in different languages. Prior work suggests that intermediate representations can be relatively language-agnostic, but generation becomes increasingly language-specific as models commit to discrete output tokens. This is problematic because language-specific lexical choices can cause semantically equivalent reasoning paths to diverge across languages. These divergences motivate searching for a cross-lingual alignment signal that is less tied to any single vocabulary item or script. We propose SOLAR, an auxiliary objective for supervised fine-tuning that aligns soft-token representations across languages, using English as a pivot. Soft tokens are probability-weighted mixtures over the vocabulary embeddings, yielding continuous representations that can aggregate information from semantically related tokens across languages. We then align each non-English soft-token summary to its English counterpart in the shared embedding space. Across four multilingual reasoning benchmarks, SOLAR improves accuracy by up to +17.7 points over the base model and +3.8 over standard supervised fine-tuning, with the largest gains on low-resource languages. SOLAR also strengthens final-layer cross-lingual similarity and substantially reduces language-cluster separability, suggesting that aligning soft-token representations helps preserve shared semantic structure during multilingual reasoning.
Hongyu Hè, Alexander Krentsel, Sylvia Ratnasamy, Maria Apostolaki
Invariants, the relations expected to hold among measured signals of a network, underpin applications from verification to traffic generation, telemetry imputation, and input validation. Yet, writing them by hand demands rare expertise in both formal logic and networking. Testing all possible combinations of telemetry variables statistically is intractable at network scale, while existing automatic tools still require a human-authored grammar of admissible invariants and learn only exact, "hard" rules that break under real-world noise. Our key idea is to use an LLM to inspect variable metadata, classify counter data types, and predict candidate syntactic relationships that form a compact grammar, followed by a deterministic statistical search within that grammar. If the initial grammar fails to recover trusted known invariants on held-out telemetry, Autogram instructs the LLM to iteratively widen the grammar until recovery succeeds. By confining the LLM to grammar induction and re-induction while delegating candidate verification to logical and statistical checks, Autogram turns non-deterministic AI proposals into auditable invariants with formal guarantees within the induced grammar. We design and prototype Autogram and evaluate it on both public and production WAN telemetry, recovering expert-derived invariants with full recall and low false positives. We close with a discussion of open problems on the path toward fully open-ended invariant discovery.
Amalia Ionescu, Jose G. Hernandez, Emily F. Wong, Anil K. Saini, Jason H. Moore, Tiffani J. Bright
Reweighting is a widely used pre-processing method for promoting discrimination-free model behavior by assigning different weights to data points during model training. While it is often credited with improving fairness, particularly by mitigating systematic underestimation for underrepresented groups, it can also introduce unintended harms under certain data conditions. Using controlled simulations that vary group size and baseline event prevalence, we show how reweighting can yield misleading improvements in fairness by reducing underestimation for one group while introducing it for another. Across 3,000 simulated datasets with event prevalence of 7-14%, reweighting improved parity in positive classification and false negative rates (FNR) between groups, but did so partly by decreasing positive classifications and increasing FNR for the group with higher baseline prevalence. These effects were most pronounced when that group was a minority: its FNR rose by 6 percentage points, approximately 82 additional missed cases per 10,000 group members, and roughly three-quarters of the improvement in FNR parity came from this increase rather than from gains for the other group. The pattern held across six classifiers, from logistic regression to neural networks, while overall accuracy was essentially unchanged, leaving the shift invisible in aggregate performance metrics. These findings mirror patterns reported in real-world applications and underscore the critical role of data characteristics in shaping the impact of this fairness intervention. While reweighting may create desirable outcomes in some applications, practitioners should critically evaluate it relative to their problem area and goals. Specifically, when the goal is to reduce underestimation in one group without introducing underestimation in another, reweighting may not be the appropriate approach.
Sparsity and quantization reduce the cost of attention by skipping interactions and lowering the precision of those retained. However, useful interactions differ in their precision requirements: some tolerate low-precision scores while remaining sensitive to removal. We therefore introduce TriAttention, which assigns query-key (QK) tiles to skipping, 4-bit, or 8-bit computation, using 4-bit to retain quantization-tolerant interactions at lower cost. To represent compactly, we encode each head's computation regions with two diagonal-distance boundaries, keeping policy storage independent of context length. Offline calibration then evaluates candidate boundary pairs using normalized projected-output errors and complete-tile execution costs. Using these measurements, a model-wide allocator selects each head's boundaries to minimize calibration error under a shared computation budget. During prefill, the selected boundaries generate precision-grouped traversals for a fused kernel that skips omitted tiles and combines contributions from both QK precisions into one normalized attention output. The quantized keys and values prepared during prefill are then reused in a compressed key-value (KV) cache for decoding, reducing repeated reads of prompt data. Experiments show that TriAttention achieves end-to-end generation speedups of up to 7.88x over the Torch attention and 2.65x over the BF16 baseline. Meanwhile, TriAttention maintains quality comparable to the original models on general-purpose and long-context tasks. These results show how a compact retention-and-precision policy translates local sensitivities into attention computation and repeated KV-cache access. The implementation is https://github.com/HanzhiZhang-Ulrica/TriAttention.
The number of neighbours k is the central hyperparameter of the k-nearest neighbour (k-NN) classifier and is usually selected globally through cross-validation or evolutionary search. However, the bias-variance trade-off is inherently local: dense and noise-free regions may favour small neighbourhoods, whereas sparse or noisy regions may require larger ones. Since global-k optimization is one-dimensional, replacing exhaustive search with a genetic algorithm offers little computational advantage. We therefore optimize an instance-dependent neighbourhood policy instead. We propose EANK (Evolutionary Adaptive Neighborhood K), which uses a low-dimensional policy to map local neighbourhood descriptors, including density, label entropy, vote margin, and concentration, to an adaptive number of neighbours k(x). A real-coded elitist genetic algorithm optimizes this policy using leave-one-out accuracy with a neighbourhood-smoothness penalty. Precomputed neighbour votes reduce fitness evaluation costs. We investigate theoretical properties of the method, including dominance over global-k solutions on the training criterion, evolutionary convergence, policy regularity, and generalization guarantees. Experiments on synthetic data with spatially heterogeneous structure and twelve public datasets evaluate predictive performance, robustness, and computational cost. Under nested cross-validation, EANK achieves competitive performance relative to tuned global-k baselines and improves upon the default k=5. We also examine the gap between training and test performance to assess the practical value of adaptive neighbourhood selection.
Marco Schouten, Ioannis Siglidis, Serge Belongie, Dim P. Papadopoulos
We learn explicit, class-conditioned spatial priors for object placement in natural scenes by distilling placement knowledge from text-conditioned diffusion models. Prior methods rely on sparse manual annotations or on inpainting object-removal pipelines, whose artifacts promote shortcut learning. To address these limitations, we propose a scalable pipeline that evaluates dense candidate placements by inserting objects into clean backgrounds. This produces HiddenObjects, a dataset of 27M bounding-box annotations (2M positives) across 27K diverse scenes and object categories. Human evaluation prefers our image editing composites over those guided by sparse annotations in 65.6% of comparisons. We distill the priors into a lightweight placement model that achieves the best accuracy among dedicated baselines and zero-shot vision-language models.
Shuwei Wang, Vadim Bulitko, Matthew R. G. Brown, Michael Youngblood, Ramon Lawrence, Roscoe Smith, Shinichi Nakagawa, William Yeoh
Neural networks controlling behavior of A-life agents can evolve over multiple generations. While effective in making the agents survive, such networks are generally opaque to a human observer. Furthermore, the neural weights cannot be transferred between agents with different neural architectures. Distilling such neural networks into compact symbolic policies can help humans understand agent behavior in A-life. Additionally, it allows the agents to share their knowledge in a compact fashion, improving population-level learning. Distilling a neural policy into a symbolic policy is typically performed by searching the space of symbolic policies to find one that approximates the neural policy. However, even with a domain-specific language the space of policies can be large, necessitating incomplete stochastic search such as genetic algorithms. Yet the distillation remains computationally expensive. We propose to instead use a transformer pre-trained on readily available random percept-action data. Doing so speeds up distillation substantially. We derive the transformer architecture via a neural architecture search. Empirical results compare different distillation methods and the influence of the symbolic space on A-life population dynamics through self-reflection and teaching and learning.
RL-based controllers achieve strong average-case performance in networking tasks such as congestion control and adaptive bitrate streaming. Yet their performance can degrade severely under network conditions where strong performance is still achievable. Identifying such conditions and quantifying the resulting performance gap is intractable by enumeration, while the sequential and closed-loop nature of RL controllers makes formal verification methods impractical. We present ReGuard, a framework that discovers worst-case scenarios for a given RL controller and protects it against them at inference time without retraining. ReGuard formulates scenario discovery as a bilevel regret-maximization problem, certifying a lower bound on the controller's worst-case performance gap. ReGuard then analyzes the discovered counterfactual trajectories and compiles them into lightweight logic rules that intervene only when a risky state is detected, leaving the controller's behavior unchanged otherwise. We evaluate ReGuard across three RL-based network controllers: Pensieve, Sage, and Park. ReGuard discovers scenarios in which each controller suffers a 43-64% performance loss relative to what is achievable under the same conditions. ReGuard exposes performance gaps 57% to 6 times larger than those found by the strongest baselines and shrinks those gaps by 79-85% in its first iteration via lightweight rule-based protection, iteratively driving the remaining gap toward nominal levels. ReGuard's protection extends beyond the scenarios it discovers, improving performance across a wide range of network conditions.
Wearable devices are increasingly used as tools for biomedical research, as the continuous stream of behavioral and physiological data they collect can provide insights about our health in everyday contexts. Long-term tracking, defined in the timescale of months of year, can provide insights of patterns and changes as indicators of health changes. These insights can make medicine and healthcare more predictive, preventive, personalized, and participative (The 4P's). However, the challenges in modeling, understanding and processing longitudinal data are a significant barrier to their adoption in research studies and clinical settings. In this paper, we review and discuss three models used to make sense of longitudinal data: routines, rhythms and stability metrics. We present the challenges associated with the processing and analysis of longitudinal wearable sensor data, with a special focus on how to handle the different temporal dynamics at various granularities. We then discuss current limitations and identify directions for future work. This review is essential to the advancement of computational modeling and analysis of longitudinal sensor data for pervasive healthcare.
The design of sixth generation (6G) mobile systems requires measurements of artificial intelligence (AI) service traffic under controlled network conditions. The 3rd Generation Partnership Project (3GPP) Service and System Aspects Working Group 4 (SA4) documents such measurements in draft Technical Report 26.870. In this paper, we describe the open testbed used for the study and analyze the results recorded in version 0.6.2 and its companion workbook. The testbed combines network emulation, packet capture, application measurements, and scenario execution against hosted and self-hosted models. The reported uplink to downlink volume ratios span more than six orders of magnitude. A personal assistant agent has a ratio of 0.007 at its local control interface and 0.50 on its wide-area egress. Deterministic agent signaling follows the configured round-trip delay. In a tokenized video experiment with 125 questions per profile, the number of correct answers changes by at most three through 10% independent uplink loss. Larger losses reduce accuracy, but the available repetitions do not establish a tolerance threshold. We also analyze emulator behavior that affects the interpretation of the study. Per-packet jitter introduces reordering, correlated random-loss settings suppress rare loss events, and a Gilbert-Elliott parameter mapping changes the stationary loss rate. The results support characterization by scenario, measurement layer, and versioned network configuration.
We study how well approximation guarantees for clustering under one norm translate into guarantees under another. Specifically, we compare $k$-clustering in $\mathbb{R}^d$ under the $L_1$ and $L_2$ objectives, defined as the sums of Manhattan and Euclidean distances, respectively. In the worst case, the elementary relationship between these norms yields a $\sqrt{d}$ approximation bound, which is already tight for $k=2$. We therefore investigate whether stronger guarantees are possible in a natural average-case model, where the dataset is arbitrary but the coordinate system is chosen uniformly at random among all orthonormal bases. Our main result shows that, when $d=Ω(\log k)$, with high probability over the random rotation, any $β$-approximate $L_1$ clustering, with optimally chosen centers for its clusters, is also an $O(β)$-approximation under the $L_2$ objective. We also establish an expected approximation ratio of $O(β\min\{\sqrt{d},k^{2/d}\})$, yielding a dimension-independent bound of $O(β\sqrt{\log k/\log\log k})$. We obtain analogous guarantees in the reverse direction, from $L_2$ to $L_1$. We complement these results with lower bounds of $Ω(\min\{\sqrt{d},k^{1/d}\})$ in both directions, showing that our dimension-independent guarantees have asymptotically optimal dependence on $k$ for constant $β$. Our proofs overcome the dependence of the algorithm's output on the random rotation using geometric projection arguments and small Euclidean coresets. In particular, our results provide guarantees for existing clustering algorithms without requiring their redesign.
Power matters for AI safety. Optimal agents tend to acquire power across a wide range of objectives (instrumental convergence). As we embed AI agents into socio-technical systems, humans risk gradual disempowerment. Both concerns call for a formal, quantitative measure of power. Two such measures in the context of Markov games are information-theoretic empowerment, the capacity of the channel from an agent's actions to future states, and capability-based individual power, a measure of goal attainment built from the probabilities that a human would attain each goal in a family of candidates were they pursuing it. We show where the two meet. From individual power, Fano's inequality yields a floor $B(s)$. This floor is a lower bound on what we term goal revelation, the mutual information $\mathbb{I}(G;ξ\mid s)$ between the human's latent goal and the realised trajectory. Empowerment caps revelation from above, so that $B(s) \le \mathbb{I}(G;ξ\mid s) \le \mathbb{I}(\mathbf{A}_h \to ξ\mid s) \le \mathfrak{E}_h(s)$. Each quantity can be large while the one below it stays at zero, highlighting the following failure modes. An assistant maximising empowerment can prefer an interface offering broad influence over goal-irrelevant degrees of freedom to one that lets the human attain every goal they might hold. An assistant maximising revelation cannot tell a working interface from one rewired so that every goal yields the wrong outcome, since relabelling goals leaves mutual information unchanged while zeroing every attainment probability. Positive guarantees therefore come only from the attainment end of the chain. A satisfaction floor, computable from two statistics of interaction logs, offers a way to monitor the power users retain in a deployed system, and simulations in random environments show the three quantities are correlated but far from interchangeable.
Chang Won Kim, Jaewook Lee, Haeun Min, JunNyeong Jeong, Minwoo Sim, Barom Kang, Kyohoon Jin
In embedding visualizations, nearby instances are interpreted as semantically related, although neighborhoods derived from input embeddings may not reflect content relationships. Neighborhood-based dimensionality reduction (DR) methods can propagate this limitation by optimizing layouts from embedding-based neighborhood graphs. We propose Knowledge-Guided Neighborhood Graph Construction (KGN), which represents relational knowledge derived from each instance as a set of knowledge graph (KG) triples. It uses the degree of triple sharing as pairwise semantic evidence to adjust embedding distances and reconstruct the top-k neighborhood graph; neither KG extraction nor neighborhood construction uses class labels. We also introduce IDProfile (IDP), which interprets residual mismatches by independently assessing whether the source content supports the assigned class label and the neighborhood-majority class. Across eight text and image dataset-encoder conditions, we characterize KGN over the full guidance-strength sweep and four KG similarity measures. Across this sweep, the maximum NMI and LT_2D each exceed the corresponding kNN baseline in all eight conditions, with the largest gains reaching 0.68 and 1.35, respectively. The gains and the extent of neighborhood reconfiguration vary across conditions and are larger under encoder conditions with lower kNN baseline scores. The IDP analysis shows that the largest reductions are concentrated in Placement Issue candidates, whereas Label Issue and Unresolved Case instances account for a larger share of the remaining review set. These results indicate that KGN can narrow the evaluated gap between embedding-based neighborhoods and content-supported semantic relations, while IDP shows that a residual mismatch should not automatically be interpreted as a projection failure.
We study whether \emph{long-horizon model-based value expansion} can serve as an effective framework in pixel-based offline reinforcement learning (RL). Long-horizon model-based value expansion from pixels is substantially harder than state-based environments, as the agent must learn both the representation and its dynamics, and errors from either can accumulate recursively over imagined trajectories. We introduce MAC-v2, which enables long-horizon model-based expansion from pixels by training a reconstruction-free latent world model with action chunking. By learning representations from prediction and control objectives, MAC-v2 does not need to encode control-irrelevant visual details that can hinder the dynamics prediction. Action chunking further mitigates the prediction error by reducing the number of recursive model predictions required for value expansion. Across diverse visual benchmarks including visual OGBench, MMBench2, Robomimic, and LIBERO-10, MAC-v2 achieves strong performance against model-free and model-based offline RL baselines. Our analyses show that long-horizon model-based value expansion is critical for the performance of MAC-v2, and that both action chunking and latent learning objectives are essential for reliable long-horizon latent imagination.
We study deterministic strategyproof mechanisms for facility location in finite-dimensional Euclidean spaces. For every population of at least three agents and every finite dimension, the coordinate-wise median minimizes the worst-case approximation ratio among anonymous mechanisms. Agent identities can improve the ratio at a fixed population size, and we give an explicit four-agent separation. Nevertheless, for every fixed dimension, the optimal ratio over all deterministic strategyproof mechanisms converges, as the population grows, to the supremum of the median's ratios over finite populations. Combined with the sharp bound for the median, this gives the optimal guarantee $\sqrt{6\sqrt3-8}$ uniform over populations and dimensions. The main structural ingredient is a rigidity theorem for strategyproof aggregation. We characterize unanimous, anonymous, translation-equivariant strategyproof mechanisms with bounded individual influence as fixed coordinate-wise order-statistic rules in a single orthonormal frame. We also prove that, for strategyproof mechanisms, continuity is equivalent to convexity of the full range in every finite dimension. An approximation-preserving normalization and a reflection argument yield the fixed-population anonymous comparison. To remove anonymity in the large-population limit, we develop an extraction theorem that converts a common guarantee on arbitrarily large populations into an anonymous deterministic mechanism on any prescribed finite population.
On-policy distillation (OPD) trains a student model to match a teacher's next-token distributions on student-generated prefixes. Extrapolative OPD further extends the teacher's log-probability shift relative to a reference policy to construct stronger learning targets. However, existing methods couple extrapolation direction and distance, allowing the magnitude of the teacher's shift to dictate how far the target moves. We propose $D^2$-OPD, decoupling these two factors: a per-prefix hard KL constraint bounds the target's distance from the teacher, while an adaptive combination of the teacher's post-training shift and the initial student--teacher gap determines the extrapolation direction. We optimize the student toward this target using forward KL, providing learning signals even for low-probability tokens. We evaluate three settings: where standard OPD performs well, where it struggles with base-model initialization, and where no teacher reference is available. On mathematical reasoning benchmarks, $D^2$-OPD attains the highest score in all three settings, outperforming the strongest baseline by up to 9.45 points. Ablations validate the contributions of each component.
Establishing almost sure convergence rates for stochastic approximation and reinforcement learning under Markovian noise is a fundamental theoretical challenge. We make progress towards this challenge for a class of stochastic approximation algorithms whose expected updates are contractive, a setting that arises in many reinforcement learning algorithms such as $Q$-learning and linear temporal difference learning. Specifically, for a power-law learning rate $O(n^{-η})$ with $η\in (1/2, 1)$, we obtain an almost sure convergence rate arbitrarily close to $o(n^{1 - 2η})$. For the harmonic learning rate $α_n=α/(n+1)$, the almost sure squared-error rate is $o(n^{-ζ})$ for every $ζ<\min\{1,2(1-κ)α\}$, where $κ$ is the contraction factor. Thus exponents arbitrarily close to one are available when $2(1-κ)α\geq1$. This approaches the $n^{-1}$ polynomial exponent of classical law-of-the-iterated-logarithm benchmarks. Key to our analysis is a novel Lyapunov drift construction that applies a Poisson-equation based correction for Markovian noise to the well-established Moreau-envelope smoothing for the contractive mapping.
Huimin Zeng, Yuxuan Lu, Jesse Cai, Eric Hanchen Jiang, Di Wu, Po-Nien Kung, Wenbo Hu, Ziyi Wang 等 11 位作者
AI agents may consume massive tokens across a wide range of tasks, but how many of those tokens are really necessary? Although prior work has shown that agents can incur unexpectedly high token costs, we still lack a systematic way to identify and measure token waste. In this work, we introduce a unified framework for measuring token waste across heterogeneous agent systems. Our framework records and normalizes complete agent trajectories, identifies removable spans using an LLM-as-a-Judge (LLMaJ), and maps the identified text spans back to the original trajectory for token-level measurement. We validate the identified waste through direct interventions by removing marked spans, resuming agent execution, and measuring changes in task outcomes and subsequent actions. In online pruning experiments, we remove 41-46% of the model's reasoning while largely preserving task performance across two distinct agent benchmarks. Using this framework, we analyze agents across multiple settings. Our analysis shows that token waste has a system-level structure that aggregate token usage alone cannot reveal. First, waste varies more across tasks and agent settings than across models. Second, reasoning and visible execution can exhibit different waste patterns. Third, different harnesses can introduce distinct sources of inefficiency even under the same model and task. Overall, our framework transforms token waste into an actionable optimization signal, enabling researchers and practitioners to train more efficient models, design better reasoning strategies, and optimize agent harnesses toward reducing unnecessary computation rather than merely reducing token usage.
Jeffrey Lee, Geoffrey Hanks, Evangeline Bouganim, Yue Hu
Robotic automation of laboratory tasks has become a growing interest to remove researchers from hazardous environments and repetitive procedures. However, even equipment as simple as squeeze bottles poses a challenging manipulation problem due to its deformability. This paper presents an adaptive hybrid vision-force controller for robotic grippers to effectively manipulate laboratory squeeze bottles to dispense liquid. This approach integrates force and vision feedback according to the interaction state, enabling effective squeezing while accounting for bottle deformation and recovery. The proposed controller is validated by comparing its performance with a pure force controller on 12 combinations of conditions consisting of 2 target dispense volumes, 2 plastic squeeze bottles, and 1 to 3 squeezes used to dispense. Human demonstrations of these operating conditions were collected using electromyography (EMG) sensing mapped to a desired grip force, which was then used as the force reference for the controllers to squeeze the bottles. Results demonstrated that the proposed adaptive hybrid vision-force controller achieved more accurate liquid dispensing with a mean absolute error (MAE) of 6.54 mL compared to a MAE of 15.92 mL from the pure force controller, while also replicating the human action recorded from the EMG signal more closely.