Machine Learning
Invariant Atoms: Sparse Coordinates of Local Semantic Geometry in Language Model Representations
Muhammad Ahtesham and Xin Zhong
Large
language models often preserve meaning despite substantial changes in wording, style, and syntax, while small semantic edits can systematically alter their hidden representations. This suggests that semantic variation may be organized along recurring local directions. We propose the Invariant Atom Hypothesis: local semantic motion admits preferred sparse coordinates along directions that remain stable under meaning-preserving transformations. We learn a shared semantic frame and sparse coordinates that reconstruct semantic displacements while suppressing nuisance variation, with anchor-dependent diagonal modulation adjusting atom strengths without sample-specific rotations. Empirically, the atoms exhibit strong semantic–nuisance separation, sparse reconstruction, reproducible directions, and causal effects on model predictions. The learned geometry generalizes to unseen semantic neighborhoods and nuisance families, while local reweighting improves semantic selectivity and preserves a consistent global-to-local structure. Atom signatures also remain stable under model modification. These findings support reusable invariant directions as a sparse coordinate system for local semantic geometry in
language models.
Read the full abstract ▾
Machine Learning
Theory on Attention Dynamics for Out-of-Distribution In-Context Learning
Junze Deng, Daouda Sow, Sen Lin et al. (+1 more)
Transformers have demonstrated remarkable in-context learning (ICL) capabilities, enabling them to perform new tasks without additional fine-tuning. However, their performance often deteriorates when encountering out-of-distribution (OOD) inputs that deviate from the training distribution, and the underlying theory remains poorly understood. To fill this gap, we characterize the OOD error under the input distribution shift through the interplay between the dynamics of the so-called $\alpha$-type and $\beta$-type
attention weights, which represent the
transformer's confidence in identifying the correct and incorrect features, respectively. Our results indicate that the OOD error for each feature depends on all pairwise interactions between the training features and OOD features, and under certain cases the
transformer performs no better than random guessing. To improve the OOD generalization performance, we next investigate the impact of model fine-tuning with the OOD data, and particularly, characterize the model forgetting performance on the source domain. Interestingly, the performance on the source domain may not always degrade after fine-tuning, which highly depends on the nature of the feature shift: fine-tuning on OOD domain keeps enhancing the confidence of identifying correct features from the original distribution, while the interference from other incorrect features may either increase or decrease. Extensive experiments on both synthetic and real data are conducted to corroborate the theoretical insights.
Read the full abstract ▾
All authors Junze Deng, Daouda Sow, Sen Lin, Yingbin Liang
Computer Vision
Online Versatile Incremental Learning: Towards Class and Domain-Agnostic Adaptation at Any Time
Jae-Ho Lee, Min-Yeong Park, Jun-Yeong Moon et al. (+2 more)
Continual learning enables
vision systems to adapt to ever-changing data distributions. Despite significant advances, existing approaches fail to capture continuous and concurrent shifts in classes and domains, a critical capability for real-world deployment. This work introduces Online VIL (Online Versatile Incremental Learning), a novel scenario where class concepts and visual domains evolve simultaneously online without explicit boundaries. To better adapt to the challenges of such dynamic environments that more closely resemble real-world conditions, we propose a novel framework TopFlow,
Topology preservation with Flow matching representation that contains two complementary mechanisms: Domain-agnostic Flow Matching (DFM) and Global
Topology Preservation (GTP). DFM guides the model to have domain-agnostic representations by integrating the geodesic flow kernel into contrastive learning. In contrast, GTP maintains the global structure of the feature space without explicitly storing past examples. Our extensive experiments demonstrate that TopFlow effectively addresses the limitations of existing methods within the Online VIL scenario, achieving state-of-the-art performance in challenging Online VIL. The proposed methods suggest potential directions for building continual learning systems in realistic dynamic environments. Our implementation code is available at https://github.com/KU-VGI/Online-VIL.
Read the full abstract ▾
All authors Jae-Ho Lee, Min-Yeong Park, Jun-Yeong Moon, Jung Uk Kim, Gyeong-Moon Park
Computer Vision
DARE to Mitigate Hallucination: Dual-path Auto-Regressive-aware Editing
Jae-Ho Lee, Jeong-Eun Lee and Gyeong-Moon Park
Large
vision-
language models (LVLMs) have recently achieved remarkable progress across
multimodal tasks, yet object
hallucination remains a persistent challenge where models generate descriptions inconsistent with the visual input. Recent work mitigates
hallucinations through training-free representation editing, typically by constructing
hallucination-related directions from teacher-forcing (TF) contrasts between hallucinated and truthful responses. However, LVLMs operate through autoregressive (AR) decoding during generation, raising the question of whether TF-based analysis fully reflects the generation dynamics that lead to hallucinated outputs. In this paper, we analyze the relationship between TF-based editing and AR generation behavior and find that TF-based editing alone may be insufficient to capture both decoding dynamics and
multimodal interactions associated with
hallucinations. To address this limitation, we propose DARE (Dual-path Auto-Regressive-aware Editing), a hybrid
hallucination editing framework that integrates two complementary contrast pathways: textual contrasts and image contrasts, together with autoregressive-aware representation signals. Specifically, DARE constructs
hallucination editing directions from (1) TF-based textual contrasts, (2) AR-aware representation transitions during decoding, and (3) controlled visual differences between paired images. Extensive experiments on multiple LVLM
hallucination benchmarks demonstrate that DARE consistently reduces object
hallucinations while preserving
multimodal perception capability and inference efficiency. Our implementation code is available at https://github.com/KU-VGI/DARE.
Read the full abstract ▾
Robotics
World4Scorer: Outcome-Grounded World Modeling for Autonomous Driving
Jieyuan Pei, Meiyi Lu, Sining Ang et al. (+12 more)
Autonomous driving requires choosing a safe and efficient plan as surrounding traffic evolves. Generate-and-select planners propose multiple trajectories and score them for execution, and they have outperformed representative direct-prediction baselines on NAVSIM. Their scorer must compare plans that were never executed. Driving logs record the future of only the executed trajectory, so matching the logged future can leave predictions for the alternatives unconstrained; a simulator, in contrast, can label the outcome of every candidate. We introduce World4Scorer, which builds the scorer as a trajectory-conditioned JEPA-style predictor: it predicts a state for each candidate and reads the candidate's scores from that state. Simulator outcome labels supervise the states of all candidates, and the observed future of the executed trajectory anchors the predictor to real scene
evolution. Because one predictor produces every candidate's state, the anchor can constrain shared parameters used to score unexecuted plans, while the future itself is needed only during training. Generated candidates mostly score well, so a scene-matched bank adds low-scoring plans to the outcome supervision; framewise choices can conflict, so inertial re-ranking keeps consecutive selections consistent. World4Scorer achieves state-of-the-art NAVSIM-v2 performance and a strong adapted-system result on closed-loop Bench2Drive. With the LeWM world model and planning budget fixed, outcome-based scoring also improves manipulation planning on the OGBench-Cube
benchmark.
Read the full abstract ▾
All authors Jieyuan Pei, Meiyi Lu, Sining Ang, Yubo Zhao, Zhangyi Hu, Mingwei Xu, Haokai Ding, Wei Li, Zihan You, Jianwei Zheng, Li Yu, Yifeng Pan, Ji Tao, Rongjunchen Zhang, Yan Wang
Artificial Intelligence
Bits Under ZK-LLM: Evaluating Zero-Knowledge-Friendly Quantization for Verifiable Private LLM Inference
Taeung Yoon, Yupeng Zhang and Xiaojing Liao
Zero-knowledge proofs are emerging as a promising approach for enabling private, verifiable LLM governance and
auditing, where regulators, users, and auditors need to verify claims about training-data usage or LLM inference-time behavior, while model providers must protect proprietary model parameters. However, despite the growing interest in ZK-LLMs, the understanding of ZK-friendly quantization remains limited. This gap matters because in the ZK setting, quantization directly shapes the arithmetic structure, constraint complexity, and proving cost of ZK inference. ZK protocols operate over finite fields and incur costs that depend heavily on the number and type of arithmetic operations, nonlinearities, and lookup constraints. Understanding ZK-friendly quantization is therefore essential for making ZK-LLMs practical. In this work, we present the first systematic study of ZK-friendly quantization for LLMs. We first formalize the definition of ZK-friendly quantization, capturing the properties required for ZK proof generation. We then evaluate nine
language models, including Qwen2.5-14B and the mixture-of-experts model Qwen3-30B-A3B, across a broad design space of weight, activation, and nonlinear lookup table precision. Our results show that activation precision is substantially more sensitive than weight precision, while nonlinear lookup approximations can become the dominant source of utility degradation. Also, we identify RMSNorm inverse-square-root lookups as a recurring bottleneck in several large models and recover near-baseline utility by selectively increasing precision only at the bottleneck. Finally, we show that reducing bit-width or lookup-table size does not necessarily yield proportional end-to-end proving savings, showing that conventional low-bit quantization heuristics do not directly translate to ZK proving efficiency and motivating operator-aware precision selection.
Read the full abstract ▾
Computer Vision
Merlin Plus: A Large-Scale, Multi-Cancer, Image-Mask-Report Dataset
Pedro R. A. S. Bassi, Wenxuan Li, Szymon Plotka et al. (+10 more)
Multi-cancer
segmentation in computed
tomography (CT) is fundamentally limited by the scarcity of tumor masks across different organs. We present Merlin Plus, the first large-scale CT
dataset with radiologist-created tumor masks across 9 organs. Merlin Plus extends the Merlin
dataset by adding 1,153 per-voxel tumor masks and longitudinal metadata. To create these tumor masks, we developed a report-based active-learning framework in which radiology reports identify tumor cases for annotation and support training of a tumor
segmentation model. The model generates initial masks, which radiologists review and correct to produce the final masks, reducing annotation burden while maintaining high-quality annotations. Besides tumor masks, the longitudinal metadata in Merlin Plus enables temporal modeling of cancer progression. By directly addressing the major bottleneck of limited multi-cancer
segmentation masks, Merlin Plus supports scalable multi-organ cancer detection,
segmentation, and longitudinal analysis in CT.
Dataset is available at: https://github.com/MrGiovanni/MerlinPlus
Read the full abstract ▾
All authors Pedro R. A. S. Bassi, Wenxuan Li, Szymon Plotka, Ruby Honjol, Jakub Przado, Xinze Zhou, Kang Wang, Yang Yang, Malte Jensen, Akshay S. Chaudhari, Curtis P. Langlotz, Alan L. Yuille, Zongwei Zhou
Computation and Language
MemFold: Learning Compact Soft Memory for Long-Context Personalization via On-Policy Optimization
Jingxuan Wu, Yuzhe Yang, Yiqiao Huang et al. (+6 more)
An assistant that serves the same user over a long horizon has to answer from what that user has revealed: which preferences still hold, which were revised, and which constraints apply now. Retaining that information is not the same as acting on it, and the two are usually optimized as if they were. Keeping the information as text makes the reader's input grow with the retained history, while compressing it into a fixed number of latent vectors bounds the interface but is typically trained to reconstruct text or imitate reference answers, both of which are scored on sequences the reader never produced. We present MemFold, which optimizes a fixed-budget soft
memory by the behavior it supports. A query-conditioned textual
memory is compressed into K continuous vectors that form the reader's
memory interface, and the reader is then trained on its own rollouts under two complementary signals: group-relative rewards for task outcomes, and confidence-gated on-policy
distillation in which a frozen textual-
memory teacher re-scores the student's sampled
tokens under the textual
memory. The teacher is never sampled from, so supervision stays on the student's current distribution and adds no autoregressive decoding; at inference it is removed entirely. Across three Qwen backbones, MemFold attains the highest accuracy we measure on PersonaMem-32K and PersonaMem-128K, with margins that widen at the longer history length, and transfers to PrefEval and LongMemEval without target-domain training. Ablations attribute most of the task gain to the reward term and a smaller additional gain to the teacher signal, and
memory interventions show that the reader depends on the instance-specific content of its soft
memory.
Read the full abstract ▾
All authors Jingxuan Wu, Yuzhe Yang, Yiqiao Huang, Chengzhi Liu, Qingni Wang, Chengxuan Qian, Shutong Wu, Jiawei Zhang, Xin Eric Wang
Artificial Intelligence
The Safety Operator: Modulating the Expression of Safety Instructions via Spectral Optimization
Benoit Dherin, Michael Munn, Xavier Gonzalvo et al. (+14 more)
Context
tokens in a
transformer-based
language model can be absorbed into the model's weights as a multiplicative operator. We study this operator in the setting of safety instructions and show that influencing its dominant
eigenvalue modulates how strongly the instruction shapes generation. We derive a Contrastive Safety Loss with a suppression weight that controls the tradeoff between emphasizing the safety instruction on harmful queries while suppressing it on harmless queries. Varying the suppression weight maps a relationship between the attack success and the over-refusal rates, supporting the hypothesis that the operator's
eigenvalue acts as a continuous dial for the instruction's influence. Moreover, this relationship holds relatively independently of how the Safety Loss is parameterized, yielding Pareto-improved safety instructions for appropriate values of suppression weight.
Read the full abstract ▾
All authors Benoit Dherin, Michael Munn, Xavier Gonzalvo, Adrian Goldwaser, Blaz Bratanic, Ananth Balashankar, Andrey Vlasov, Pinzhi Huang, Cecile Loge, Nicole Mitchell, Andre Fernandes, Trilok Acharya, Wendy Kan, Ziyue Wang, Hanna Mazzawi, Felipe Tiengo Ferreira, Mor Geva
Computer Vision
RA-CFGCache: From Branch-Level Criteria to Guided-Risk Control under Classifier-Free Guidance
Yiming Liu, Ben Wan, Tongxuan Liu et al. (+5 more)
Diffusion models enable high-quality visual generation, but iterative denoising remains computationally expensive, especially under classifier-free guidance (CFG), which requires both conditional and unconditional evaluations. Training-free caching reduces this cost by reuse of previously computed features or predictions. However, existing branch-local reuse criteria do not explicitly account for how cache errors combine under CFG or how local perturbations affect the final output. We identify two misalignments in cache control: a branch-guided mismatch, where guided error depends on both the magnitudes and
alignment of branch errors, and a local-final mismatch, where the downstream impact of a local error varies across timesteps. We propose RA-CFGCache, a Risk-Aligned Caching framework under CFG that incorporates both factors while keeping the sampling schedule and guidance rule fixed. CFG-aware Guided-Risk Composition combines existing branch-wise proxies using CFG coefficients and offline-calibrated cross-branch
alignment. Propagation-Aware Rescaling further weights the resulting guided-risk estimate with a timestep-dependent propagation prior calibrated from isolated reuse perturbations. An online threshold controller then determines when to jointly refresh or reuse both branches. Experiments on FLUX.1-dev, Wan2.1-T2V-1.3B, and CogVideoX-2B demonstrate improved efficiency–fidelity trade-offs over evaluated training-free caching baselines. Moreover, RA-CFGCache is compatible with diverse base proxy families, including TeaCache-, DiCache-, and MagCache-style
estimators, and consistently improves fidelity at nearly unchanged latency. Code is available at https://github.com/yiming-l21/RA-CFGCache.git.
Read the full abstract ▾
All authors Yiming Liu, Ben Wan, Tongxuan Liu, Ao Wang, Yuqi Xiong, Fan Zhang, Hui Chen, Guiguang Ding
Computer Vision
Temporal-Aware Fusion for Robust Outdoor LiDAR Localization
Minghang Zhu, Zhijing Wang, Yuxin Guo et al. (+5 more)
LiDAR relocalization aims to estimate the global 6-DoF pose of a sensor in the environment. However, existing regression-based approaches often encounter limitations in dynamic or ambiguous scenarios, as they typically prioritize single-frame inference, leaving the potential of spatio-temporal consistency across scans not fully explored. In this paper, we propose a Temporal-aware Localization framework (TempLoc) designed to enhance the robustness of outdoor localization by effectively modeling sequential consistency. Specifically, a Global Coordinate Estimation module is first introduced to predict point-wise global coordinates and associated uncertainties for each
LiDAR scan. A Prior Coordinate Generation module is then presented to estimate inter-frame point correspondences by the
attention mechanism. Lastly, an Uncertainty-Guided Coordinate Fusion module is deployed to integrate both predictions of point correspondence in an end-to-end fashion, yielding a more temporally consistent and accurate global 6-DoF pose. Experimental results on the NCLT and Oxford RobotCar
benchmarks show that our TempLoc outperforms state-of-the-art methods by a large margin, demonstrating the effectiveness of temporal-aware correspondence modeling in
LiDAR relocalization.
Read the full abstract ▾
All authors Minghang Zhu, Zhijing Wang, Yuxin Guo, Chen Liu, Yongshu Huang, Wen Li, Sheng Ao, Cheng Wang
Computer Vision
Towards Scalable Context-Aware Single-Cell Spatial Transcriptomics Prediction from Histology Images
Zijun Gao, Chunbin Gu, Jinxi Xiang et al. (+2 more)
Predicting gene expression from H&E-stained histology images offers a scalable alternative to costly spatial transcriptomics, yet most existing methods operate at the spot level, where signals from multiple
cells are aggregated and critical cellular heterogeneity is obscured. Extending this paradigm to single-
cell resolution is non-trivial. Naively applying pathology foundation models faces a scale mismatch: their patch-level representations mix multiple
cells, whereas per-
cell cropping or resizing distorts morphology and removes local context. Conversely,
segmentation-based models without strong pretrained visual encoders often lack the morphological representation capacity needed for accurate molecular prediction and inherit errors from imperfect
cell boundary masks. Here, we present CELLO, an efficient end-to-end framework that performs a single pathology foundation model forward pass per image and uses grid sampling to extract location-specific features for all
cells simultaneously. We further introduce a distance-decay cross-
attention module that refines each
cell representation using spatially biased local morphological context. Using 52 public Xenium-H&E pairs from HEST-1k that span 12 organs and approximately 10 million
cells, CELLO improves the average predictive accuracy over the evaluated baselines while reducing the mean whole-slide inference time compared to DeepSpot2
Cell, a 14.0x speed-up on average that excludes upstream
cell segmentation. Our work establishes a scalable foundation for single-
cell gene expression prediction from H&E images.
Read the full abstract ▾
All authors Zijun Gao, Chunbin Gu, Jinxi Xiang, Xiangde Luo, Pheng-Ann Heng
Robotics
Losing the name before the box: measuring and repairing what narrow fine-tuning costs a detector outside its deployment vocabulary
Trung Minh Bui, Jongsul Moon, YoungOuk Kim et al. (+2 more)
A detector pretrained on a broad corpus is fine-tuned on a narrow domain, its in-domain accuracy improves, and it ships. We ask what happens meanwhile to its coverage of objects the vocabulary never names, which in obstacle detection and inspection carry the risk. No in-domain test set holds an example of one. We give a longitudinal protocol: one pretrained checkpoint against its own fine-tuned descendants. It tracks held-out top-$K$ proposal coverage $C_\tau$: of categories pretraining covered and the vocabulary omits, the share of boxes a detector's top $K$ regions still cover. The quantity is the open-world proposal literature's; the longitudinal reading is not. $C_\tau$ falls while in-domain accuracy rises, on four architectures and three domains, by $5.12$ to $63.35$ points on boxes above $1024$ px$^2$. No in-domain number identifies the fall, and neither does detection average precision, which charges a missed and a misnamed box alike. On the one architecture scoring both, adaptation costs $87\%$ of the AP against a fifth of the coverage, and the naming goes first at all six depths of its freeze ladder, every run. What breaks is structured: three architectures sharing no pretraining run agree on which categories lose coverage, and those a model never learned do not lose any. A repair follows and needs no training: mixing a quarter of the pretrained state back, normalisation statistics included, raises coverage on every
cell swept for at most $2.47$ points of in-domain accuracy. Seeing it costs one extra evaluation pass.
Read the full abstract ▾
All authors Trung Minh Bui, Jongsul Moon, YoungOuk Kim, Jung-Hoon Hwang, Dongin Shin
Multiagent Systems
Agent-Based Evolutionary Dynamics for Mixed Autonomy Weaving Ramps
Sheryl Paul, Kexin Wang, Ruolin Li et al. (+1 more)
Existing models of mixed-autonomy weaving ramps characterize how altruistic connected and automated vehicles (CAVs) can improve traffic efficiency at the population level, but provide limited insight into how such behavior emerges from decentralized vehicle interactions or how it is affected by finite populations, heterogeneous preferences, and imperfect information. We develop an
agent-based model of a macroscopic weaving-ramp framework in which individual vehicles adapt their lane choices using an evolutionary game-theoretic update rule and altruism-based objectives providing a microscopic interpretation of the original Wardrop model. We prove convergence of the decentralized dynamics to the unique
equilibrium predicted by the macroscopic theory. Beyond reproducing aggregate
equilibrium behavior, the framework enables the study of deployment-level questions that cannot be addressed by static analysis. Simulation results demonstrate close agreement with the macroscopic predictions while revealing how convergence rates, adaptation to changing traffic conditions, heterogeneous altruism levels among CAVs, and imperfect state information influence system performance and the distribution of altruistic burden across vehicles. These results provide a bridge between
equilibrium traffic theory and decentralized mixed-autonomy deployment.
Read the full abstract ▾
All authors Sheryl Paul, Kexin Wang, Ruolin Li, Jyotirmoy V. Deshmukh
Systems and Control
Discrete-time polynomial systems with inputs and outputs: structure, reachability, observability, and minimal realizations
Eduardo D. Sontag
We study the realization problem for discrete-time input/output
polynomial systems. These are formalized using tools from commutative
algebra and algebraic geometry as systems whose state spaces are algebraic varieties, or more abstractly the set of k-points of an affine k-scheme, where k is an arbitrary infinite field. The input/output behaviors of such systems are described by "
polynomial response maps" in which outputs are
polynomial functions of past inputs. The main results show that every
polynomial response map admits a canonical (quasi-reachable and algebraically observable) realization, which is unique up to isomorphism. The key to the approach is to linearize dynamics by considering a "dual" system in which states are functions defined on states. Finite dimensionality of the canonical realization, and its polynomiality, are characterized in terms of the space,
algebra, and field of observables of the map, as well as in terms of algebraic input/output difference equations, and by a Jacobian rank criterion. A particular subclass consists of the maps that we call "bounded," defined by the property that their degree in the past inputs is uniformly bounded. Bounded maps are shown to be finitely realizable if and only if they are realizable by finite-dimensional state-affine systems, whose theory in turn reduces to that of rational formal power series. We also study the
lattice of quasi-reachable realizations of a given map, including normal realizations. This work is an update of the PhD thesis written by the author in 1976; connections to recent work are briefly discussed in the last section.
Read the full abstract ▾
Computational Complexity
Sample Complexity of Equivariant Reinforcement Learning
Rayan Mazouz, Haibo Zhao, Chris Hillar et al. (+1 more)
Reinforcement learning (RL) is a powerful framework for robotic control, yet its practical application is often hindered by high sample complexity. This is particularly restrictive in physical domains where interaction data is costly. While the world often exhibits
geometric and physical
symmetries, standard RL
algorithms typically fail to exploit this structure. In this paper, we demonstrate that exploiting group
symmetries significantly reduces the sample complexity of RL. Focusing on finite-horizon Markov decision processes, we find that leveraging homomorphisms induced by group
symmetries significantly reduces the theoretical upper and lower bounds on the number of environment interactions required to reach an optimal return. We further extend these bounds to continuous state and action spaces, providing corresponding sample-complexity guarantees under appropriate regularity assumptions. Beyond theory, we validate our findings through controlled experiments and demonstrate the advantages of
symmetry-aware policy learning on high-dimensional continuous robotic simulations. Our results show that integrating
symmetry into the learning pipeline yields substantial gains in sample efficiency and performance, offering a principled path toward more data-efficient robotics.
Read the full abstract ▾
All authors Rayan Mazouz, Haibo Zhao, Chris Hillar, Christian Shewmake
Numerical Analysis
A unified structure-preserving framework for geometric flows with coupled orientation and curvature dependence
Yulin Zhang
We develop a structure-preserving parametric finite element framework for
geometric flows whose energy density couples the unit normal and the
curvature. This class includes bending energies with orientation-dependent
rigidity or spontaneous
curvature. A common fully discrete formulation treats closed curves in two dimensions and closed surfaces in three dimensions, and accommodates the $L^2$ flow, curve or surface
diffusion, and the area- or volume-constrained $L^2$ flow. The formulation couples the
geometric and
curvature updates so that their contributions satisfy a discrete energy inequality. It uses continuous piecewise linear elements, a surface energy
matrix, and a mass-lumped projection of the
curvature derivative of the density. For positive densities satisfying a directional condition and convexity in
curvature, we prove energy dissipation without a time-step restriction. The
diffusion and constrained flows also preserve the enclosed area or volume exactly. The analysis allows non-even anisotropies and nonseparable dependence on orientation and
curvature. Numerical experiments exhibit approximately second-order convergence in the
manifold distance and confirm the discrete structural properties. Shape relaxation under an anisotropic Helfrich-type energy illustrates the use of the framework for coupled directional and bending effects.
Read the full abstract ▾
Artificial Intelligence
Support-Set Target Leakage in Relational Foundation Models during In-Context Learning: Model Dependence and Evaluation Reliability
Roshan Reddy Upendra, Alexandre Dorais, Joe Meyer et al. (+6 more)
Relational in-context learning (ICL) uses labeled support examples and their linked relational context to predict labels for new queries. This creates a failure mode when target-derived features are present in the support context but unavailable for the query. We study this setting as support-set target leakage. We construct 20 controlled target-derived features that vary in signal fidelity, representation, semantic transparency, coverage, and zero-, one-, and two-hop relational placement, and evaluate them across 13 RelBench tasks and five relational ICL configurations that vary the ICL head, message-passing depth, pretraining cohort, or relational encoder architecture. We evaluate matched 0-hop, 1-hop, and 2-hop leakage settings, together with a Full leakage condition containing all 20 leaker columns. Within the tested configurations, target-table (0-hop) and Full leakage produce the largest aggregate deviations from clean evaluation, while higher-hop effects are often weaker, consistent with differences in effective exposure associated with temporal reachability, sampling, and aggregation fidelity. Leakage effects are strongly task- and model-dependent and can reverse relative conclusions between model variants even when aggregate changes are small. For leaker detection, we compare an Integrated Gradients (IG)-based screening method with mutual information (MI) and leave-one-column-out (LOCO) on a common Baseline subset. Ranking quality is strongest in the high-impact 0-hop and Full leakage conditions, but detector-based removal does not consistently restore the clean evaluation. A four-task rel-salt case study further shows the same evaluation concern with native-schema leakage candidates from the original relational schema. These results identify the support/query information boundary as an important component of reliable relational ICL evaluation.
Read the full abstract ▾
All authors Roshan Reddy Upendra, Alexandre Dorais, Joe Meyer, Andrew Pouret, Anastasios Lambrianos Stappas, Dinesh Katupputhur Ramprasath, Viswanath Ganapathy, Tom Palczewski, Minghua Li
Robotics
FineART: Fine-Grained Annotated Robotic Trajectory Dataset and Vision-Language-Action Model for Bimanual Manipulation
Jade Choghari, Pepijn Kooijmans, Mansi Agarwal et al. (+8 more)
Robots operating in real-world environments must often execute complex, multi-step bimanual tasks over long horizons rather than single, isolated actions. Current manipulation
datasets provide limited support for this capability: although single-arm
datasets reach hundreds of thousands of trajectories, they typically provide only one high-level instruction per episode, while existing bimanual
datasets provide subtask annotations for only part of their recorded hours. We present FineART, a densely annotated bimanual manipulation
dataset comprising 40,543 episodes (1,718 hours) and 533,913 subtasks across 151 tasks. We also introduce FineART-VLA, a
vision-
language-action policy that predicts its own next subtask to guide its actions. Mid-training on FineART's subtask annotations improves FineART-VLA's success at following spatial instructions from 32.0% to 100.0%. With step-by-step human subtask guidance, success on unseen long-horizon tasks increases from 16.0% to 76.0%. Furthermore, FineART-VLA matches baseline performance on a new
robot with 10x less fine-tuning data and generalizes zero-shot to unseen tasks. We open-source the full
dataset, model weights, and training code.
Read the full abstract ▾
All authors Jade Choghari, Pepijn Kooijmans, Mansi Agarwal, Yusuf Umut Ciftci, Aseem Doriwala, Catherine Weaver, Mouli Sivapurapu, Kai Yang, Thomas Wolf, Jackson Lee, Pragna Mannam
Computation and Language
Eternal Sunshine of the Spotless Mind: Systematically Erasing LLM's Memories
Olga Ohrimenko
We consider persistent LLMs that accumulate
memories of their interactions with a user over time. Such LLMs maintain
memories using external storage, which they can query to overcome the limitations of a fixed context window. Such systems have numerous practical applications, as they can draw on all past interactions when responding to user queries. In this paper, we ask whether LLMs can forget information shared with them upon a user's request. We find that current LLMs fail to delete such information—even when they claim to have forgotten it and even when operating with a limited context. To this end, we consider a new direction of study: Deletion of LLM
Memories. We show that naively removing messages that match a user's deletion request is insufficient, since conversations naturally introduce message dependencies that cause information to persist. To correctly handle deletion requests, we propose the DeLLM framework. It dynamically constructs relevant context for each LLM query and maintains a provenance graph of messages to determine which ones must be removed during deletion. Our experiments show that DeLLM achieves a high deletion rate while maintaining utility.
Read the full abstract ▾
Robotics
One from Infinity: Actualizing Futures from Pretrained World Models into Robot Actions
Bang Du, Yichen Xie, Shuqi Zhao et al. (+3 more)
A pretrained
video world model admits many plausible futures for a scene, but a
robot must realize the exact task-conditioned one. To turn world models into executable
robot policies, existing methods fine-tune the heavy world model backbone using large-scale
robot data and computational resources. Challenging this status quo, we argue that the expensive part has already been paid in the world model pretraining since the representation space of a
video world model lays out the diverse potential futures. In this case, what remains is to select the future that accomplishes the task and to read out the actions that realize it. We formalize this task as actualization, which learns a task-conditioned selection and realization on top of a prior supplied by a frozen world model. This can be solved by a tiny actualizer model. We implement RoboActualizer with as few as 60M parameters on top of a frozen world model encoder. The actualizer is composed of two lightweight DiT experts that jointly predict future latents and actions by flow matching. The model can be trained entirely on a single GPU with 32 GB peak
memory. With up to 100x fewer trainable parameters than existing WAMs and VLAs, RoboActualizer reaches great performance on simulation
benchmarks including LIBERO, LIBERO-Plus, RoboTwin 2.0 and five tasks on two real-world platforms, with a low latency of 39 ms that allows real-time control.
Read the full abstract ▾
All authors Bang Du, Yichen Xie, Shuqi Zhao, Yuxin Chen, Menglin Wu, Masayoshi Tomizuka
Machine Learning
Longer Records, Broader Invariance: The Hidden Scaling Problem in Longitudinal Contrastive Learning
Rameen Mahmood, Xuhai "Orson" Xu, Zachary Beattie et al. (+2 more)
Longitudinal data are valuable because people change. Yet the objectives used to learn from these data can inadvertently erase that change. In person-level contrastive learning, observations from the same person are treated as positives; as records grow, those positives can span increasingly distant—and increasingly different—behavioral states. More history can therefore produce not only more data, but broader invariance. We show that this distinction is fundamental. We separate record span, how much history the learner sees, from supervision span, how far across that history positive-pair supervision reaches. Across in-home sensing records spanning up to 2.7 years, broader supervision systematically suppresses recoverable changing-state information, even when the available history is held fixed. At the broadest span, less than 10% of the information recoverable from an untrained encoder remains. Yet keeping positives local is not sufficient: as records grow, even distant states that are never paired become increasingly similar. Explicitly contrasting other observations from the same person reverses this loss without shortening the record, revealing a second route by which longitudinal scale can broaden invariance. Finally, we prospectively reproduce the supervision-span effect in 199 GLOBEM participants. Longitudinal scale therefore presents a choice: more history need not mean more invariance. By controlling what is held invariant as records grow, we can preserve the change that made the longitudinal data valuable in the first place.
Read the full abstract ▾
All authors Rameen Mahmood, Xuhai "Orson" Xu, Zachary Beattie, Jeffrey Kaye, Danny Yuxing Huang
Computer Vision
What Makes High-Magnification Knowledge Transferable? A Study of Cross-Resolution Distillation in Whole-Slide Imaging
Zhiyuan Yang, Jiahao Cheng and Mahdi S. Hosseini
Cross-resolution knowledge
distillation aims to improve low-magnification whole- slide analysis by transferring high-magnification representations, yet the conditions for useful transfer remain unclear. We develop a decomposition-based analysis of teacher access, representation loss, and model excess, motivating three questions: whether (a) teacher targets help the task, (b) low-magnification students can predict them, and (c) slide models benefit from those predictions. We investigate them through controlled experiments across ten pathology cohorts spanning classifi- cation, grading, and survival prediction. In the main comparison, providing teacher regional means alongside native low-magnification features improves downstream performance in all ten cohorts. Direct prediction achieves lower reconstruction error than residual prediction, yet the predicted features underrepresent variation in the teacher targets. Moreover, better reconstruction does not consistently improve downstream scores, and retaining native features changes performance even when the predicted teacher features are held fixed. Together, these findings expose a gap between reconstructing teacher representations and realizing their downstream value. They challenge the sufficiency of reconstruction error as a measure of cross-resolution transfer and provide a diagnostic framework for examining where that transfer breaks down. Future
distillation designs must account for both what students can predict and how slide models use those predictions.
Read the full abstract ▾
Artificial Intelligence
From Retrieval to Reasoning: Agentic Mechanism Prediction from Cell Painting Profiles
Jiayuan Chen, Botao Yu, Tianyu Liu et al. (+3 more)
Cell Painting is a high-content morphological profiling assay widely used for phenotype-based biological inference, with mechanism of action (MOA) prediction as a central application. Existing approaches largely formulate
Cell Painting-based inference as representation matching, assigning predictions from nearby reference perturbations in morphological feature space. However, retrieved neighbors are often noisy and partially misleading evidence due to batch effects, non-specific cytotoxicity, phenotypic convergence, and source-dependent variability. We reformulate
Cell Painting-based MOA prediction as a calibrated evidence
reasoning problem, where retrieved neighbors are treated as uncertain observations that must be evaluated, compared, and sometimes rejected before supporting a mechanistic conclusion. We propose PhenoAIR, a reliability-aware multi-
agent framework that maintains a candidate-centric evidence
memory and performs controller-guided refinement over phenotype- and mechanism-side evidence. PhenoAIR uses offline reference-set calibration to weight evidence by source reliability, phenotype stability, and mechanism-level confusion. We evaluate PhenoAIR on a
benchmark constructed from JUMP
Cell Painting profiles and annotations, covering controlled, realistic, and discovery-oriented open-world MOA prediction settings. PhenoAIR outperforms representation-matching and LLM-based baselines across all settings.
Read the full abstract ▾
All authors Jiayuan Chen, Botao Yu, Tianyu Liu, Thai-Hoang Pham, Meng Wu, Ping Zhang