Peer Revenue Publish or profit

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Computer Science

cs Watch Computer Science

Wednesday 14 October · still arriving

Computer Vision

MMKG-R1: Constructing Multimodal Knowledge Graphs via Reinforcement Learning

Zongyu Wu, Yilong Wang, Xiaochen Wang et al. (+5 more)

Multimodal knowledge graphs (MMKGs) organize information across different modalities to support multimodal Graph Retrieval-Augmented Generation (RAG), which aims to improve the performance of multimodal large language models (MLLMs). The effectiveness of these multimodal Graph RAG methods largely depends on the quality of the constructed MMKGs. However, most existing methods construct MMKGs without explicit guidance on graph quality, while reinforcement learning (RL)-based methods that rely on final-answer correctness may provide only indirect feedback on MMKG quality. To address these limitations, we propose MMKG-R1, a novel MMKG-quality-driven RL framework for training an MLLM as a multimodal knowledge constructor. To be specific, MMKG-R1 introduces multiple quality-aware rewards to directly evaluate both the quality of information extracted from individual knowledge sources and the quality of the assembled MMKG. These rewards are then used to guide the RL training. Experiments across multiple datasets show that MMKG-R1 constructs MMKGs that improve downstream question-answering accuracy.
Read the full abstract

All authors Zongyu Wu, Yilong Wang, Xiaochen Wang, Minhua Lin, Yuan Zhong, Fenglong Ma, Xiang Zhang, Suhang Wang

Artificial Intelligence

Can We Believe AI Peer Review?

Zongrui Yang, Haoyuan Li, Zhongsheng Wang et al. (+2 more)

As large language models become widely used in scientific review and manuscript revision, AI scores are increasingly serving as feedback for assessing paper quality and guiding edits. Yet whether a single AI score reliably reflects the reviewer's judgment, and whether revision gains based on that score are endorsed by other reviewers, remain understudied. Using 200 papers from ICLR 2025 and 2026 and approximately 14,000 AI reviews, we study score reliability, score-distribution estimation, and cross-model agreement on revision gains. Repeated reviews fluctuate even under identical conditions. Finer-grained scoring protocols reduce score variance but also systematically shift mean scores, while the gains assigned to manuscript revisions depend strongly on the reviewer.
Read the full abstract

All authors Zongrui Yang, Haoyuan Li, Zhongsheng Wang, Gillian Dobbie, Jiamou Liu

Human-Computer Interaction

Beyond Initial Creation: Two Decades of Community-based Chart Revisions from Wikimedia Commons

Jasmine Y. Shih, Chase Stokes, Melanie Tory et al. (+1 more)

Open online communities like Wikipedia offer a rich space for understanding the organic evolution of public-facing artifacts, but little is known about how these communities collectively edit chart images. We investigate the phenomenon of "community editing" of visualizations by exploring chart revision patterns on Wikimedia Commons. We introduce a novel dataset of 7,786 longitudinal chart revisions across 1,664 images, labeled with edit categories using an LLM pipeline. Initial data analysis uncovered several novel findings. Multi-contributor revision is associated with more edits to mark styling and text content compared to single-contributor revision. Temporally, mark and text edits occur in more "bursty" cadences than data edits. In addition, while SVG charts attract more unique contributors and revisions than raster charts, their edit histories involve more tightly scoped and less diverse changes. This dataset enables a wide range of analyses of chart change over time.
Read the full abstract

All authors Jasmine Y. Shih, Chase Stokes, Melanie Tory, Marti Hearst

Robotics

Probabilistic Physics Digital Twin for Human-Robot Interaction

Yushen Hu, Harshvardhan Chandirasekar, Apurv Kushwaha et al. (+2 more)

Digital twins are increasingly used in robotics to simulate interactions, predict action outcomes, and support planning. However, constructing accurate digital twins requires precise estimates of physical properties such as mass, friction, and center of mass, which are often difficult to obtain in real-world environments. We propose probabilistic digital twins, which represent uncertain physical properties as probability distributions rather than fixed values. By simulating actions across these distributions, the twin predicts a range of possible outcomes and their likelihoods. These probabilistic simulations can guide both human and robot actions by identifying actions that are effective despite physical uncertainty. We demonstrate this framework through human-robot interaction scenarios, including interfaces that communicate simulated outcomes to users and robot planning strategies that explicitly account for uncertainty.
Read the full abstract

All authors Yushen Hu, Harshvardhan Chandirasekar, Apurv Kushwaha, Jonathan King, Zhu-Tian Chen

Cryptography and Security · Cross-listed from stat.OT

Differentially Private Learning of Markov Chains: Near-Optimal Sample Complexity for Item and User Level Privacy

Devansh Gupta, Cristóbal Guzman and Enayat Ullah

Sequential data such as documents, conversations, and trajectories are naturally partitioned across users, motivating privacy guarantees that protect each user's entire trajectory. We initiate the study of privately estimating the transition matrix of a discrete ergodic Markov chain from N independent trajectories of length T, with arbitrary initial distributions, under maximum row-wise ell 1 error. Besides N, T, and the number of states, our bounds depend on the minimum stationary probability and the pseudo-spectral gap, which govern state coverage and mixing. Under item-level privacy, we give upper and lower bounds that separate the costs of sampling, state occupancy and mixing, and privacy. Under user-level privacy, we reduce the problem for sufficiently long trajectories to private product distribution estimation. Along the way, we establish optimal user-level private bounds for estimating products of discrete distributions, which may be of independent interest. In the user level setting, our bounds are nearly tight in the large T regime, whereas in the item-level setting, our bounds as tight up-to a non-private term.
Read the full abstract

Computation and Language

LLM-jp-4 Technical Report

LLM-jp: Chaoran Liu, Daisuke Kawahara, Hideyuki Tachibana et al. (+14 more)

This paper presents LLM-jp-4, a family of Japanese sovereign large language models (LLMs) developed with an emphasis on transparency, reproducibility, and practical usability. LLM-jp-4 comprises an 8B dense model and a 32B-A3B mixture-of-experts model, both trained from scratch on approximately 12 trillion tokens. To support independent verification and reproduction, we establish explicit criteria for selecting training data, external models, and software, and make the development process publicly available. Evaluation results demonstrate substantial improvements over previous LLM-jp models and competitive performance against open-weight models of similar scale. By documenting the development principles, training procedures, and empirical findings of LLM-jp-4, we aim to provide a reference for future sovereign AI development.
Read the full abstract

All authors LLM-jp: Chaoran Liu, Daisuke Kawahara, Hideyuki Tachibana, Hirokazu Kiyomaru, Jun Suzuki, Koshi Eguchi, Kouta Nakayama, Qianying Liu, Rio Yokota, Ryo Takahashi, Sadao Kurohashi, Sosuke Hosokawa, Taishi Nakamura, Takashi Kodama, Yun-Ang Wu, Yusuke Miyao, Yusuke Oda

Computer Vision

NavAble: A Large-Scale Dataset and Synthetic Data Generation Pipeline for Blind Navigation

Hochul Hwang, Jahir Sadik Monon, Soowan Yang et al. (+9 more)

Reliable recognition of accessibility-critical objects (e.g., audible pedestrian signals, door-activation buttons, and handrails) is essential for assistive navigation technologies that support safe, independent mobility for blind and low-vision (BLV) users. Yet these categories are underrepresented in existing large-scale vision datasets, and the few datasets that include them suffer from limited class coverage, inconsistent annotations, and poor diversity. We introduce NavAble, a large-scale dataset spanning 11 accessibility-critical object classes, combining 41K curated open-source images with 8K newly collected, densely annotated real-world images. To extend scale and diversity, we develop a synthetic data generation pipeline that renders high-fidelity 3D assets across 37 environments under diverse styles and viewpoints, yielding 452K frames with rich ground-truth annotations from 565 assets. The pipeline minimizes manual effort by combining automated web image crawling, VLM-based filtering, SAM 3D-based asset generation, and convenient camera trajectory configuration, enabling large-scale data generation at a diversity unattainable through manual collection. Experiments show that fine-tuning existing segmentation models on the NavAble Dataset achieves reliable accessibility-object segmentation (>80% mIoU on our held-out test set). We also demonstrate the effectiveness of synthetic data augmentation, which yields up to +3.5 mIoU over real-world-only fine-tuning. Beyond benchmark evaluation, validation on egocentric data from two mobility-assistive robots demonstrates improved perception reliability in real-world navigation scenarios. Together, NavAble establishes a scalable foundation for BLV navigation perception, directly addressing the data scarcity and embodiment gaps. The dataset and pipeline are publicly available at https://huggingface.co/datasets/NavAble/NeurIPS_2026_BLV.
Read the full abstract

All authors Hochul Hwang, Jahir Sadik Monon, Soowan Yang, Shiven Umeshbhai Patel, Anh Nhat Hong Nguyen, Kien Nguyen, Keshav Garg, Dylan Gage, Duretti Hordofaa, Anshu Anjna, Eshed Ohn-Bar, Donghyun Kim

Computer Vision

SonoWorld: Towards Interactive Pose-Controllable Ultrasound World Models

Zefan Yang, Zheyuan Zhang, Boris Maihe et al. (+7 more)

Robotic ultrasound can standardize image acquisition and reduce operator workload and variability, but policy learning remains limited by the lack of realistic, controllable simulation environments. While controllable world models have advanced rapidly in natural-image domains, ultrasound world modeling remains challenging because each frame captures only a cross-section of unobserved 3D anatomy, causing small probe motions to induce large appearance changes. Existing approaches rely on robot-tracked poses and latent-space prediction, but do not directly generate the observations required for interactive control. We propose SonoWorld, a causal ultrasound world model that learns video dynamics from unlabeled clinical videos and grounds them to 6-DoF control using a small set of freehand clinical scans tracked with an EM sensor. SonoWorld is trained in three stages: pretraining a latent-action-conditioned video diffusion model on around 6.9 million unlabeled multi-organ ultrasound frames; grounding the learned video dynamics to metric probe poses with a small EM-tracked thyroid dataset; and applying distribution matching distillation to enable causal, long-horizon autoregressive generation with selectively caching eviction tokens at inference time. We evaluate controllability through pose-following fidelity, round-trip consistency, and axis-specific control under synthetic actions on held-out tracked scans, and further validate it with a user study. In model-based action discrimination, SonoWorld correctly ranks the target-directed probe motion over an equal-magnitude opposite motion in up to 74.8% of trials. To our knowledge, SonoWorld is the first causal, pose-controllable ultrasound world model learned primarily from freehand clinical videos, providing a data-driven interactive environment for robotic ultrasound data acquisition and policy evaluation.
Read the full abstract

All authors Zefan Yang, Zheyuan Zhang, Boris Maihe, Yikang Liu, Hang Yu, Ankush Mukherjee, Xiao Chen, Pingkun Yan, Terrence Chen, Shanhui Sun

Numerical Analysis

Solving a sequence of eigenvalue problems using snapshot and row subsampling

Francesco Hrobat and Yuji Nakatsukasa

A sequence of eigenvalue problems requires solving $A_i\vec{x}_i= \lambda_i\vec{x}_i$ where $i \in \{1, \dots, N\}$, $N \gg 1$ and (usually) $\lambda_i$ is the largest or smallest eigenvalue of $A_i$. Often, the matrices $A_i$ come from a discretization of a smooth map depending on a parameter $p \rightarrow A(p)$, that is $A_i = A(p_i)$, so the matrices $A_i$ change gradually along the sequence. Under this latter assumption, we introduce an efficient algorithm for solving sequences of eigenvalue problems. Our algorithm combines the ideas of reduced order model and row subsampling, as done in [E. Jones and Y. Nakatsukasa, arXiv, 2025] for linear systems. In addition, in our approach, the same set of rows is guaranteed to work well independently of the value of the parameter $p$ or the matrix $A(p)$ as long as the snapshot matrix is unchanged and given that the row subset was chosen well at the beginning. Furthermore, we implement an update strategy of the snapshot matrix. Along these results, we present theoretical analysis of the snapshot matrix for eigenvalue problems, showing an exponential decay of the singular values, and we give bounds on the error on eigenvectors for unseen value of the parameter. This analysis generalizes that in [M. Guido, D. Kressner, and P. Ricci, J. Sci. Comput., 2024] and shows that, regardless of the distribution of the snapshot points, the convergence is always (asymptotically) exponential in the number of snapshot points if the function $A(p)$ admits an analytical extension. We further generalize our algorithm for computing largest or smallest singular values and relative singular vectors and to generalized eigenvalue problems. We conclude with numerical results showing the efficiency of our approach.
Read the full abstract

Cryptography and Security

AutoPoly-HE: Autonomous Design of Polynomial Activations for Privacy-Preserving Encrypted Inference

Zeinab Elkhatib, Kamrul Hasan, Ali Sekmen et al. (+2 more)

Privacy-preserving machine learning has become critical for deploying artificial intelligence in sensitive domains such as healthcare, finance, and infrastructure monitoring. Homomorphic encryption (HE) enables secure inference on encrypted data. Despite its advantages, HE is not compatible with nonlinear activation functions found in machine learning models. Existing methods mitigate this problem by manually designing polynomial approximations for these activation functions. However, these functions do not account for the differences in dataset characteristics or model architecture. Furthermore, these functions are also limited to the accuracies of convolutional neural network (CNN) models. In this work, we propose an autonomous framework that selects polynomial activation functions tailored to each dataset and model. Using this framework, we analyze the interdependence between polynomial activation design and model architecture by evaluating several CNN ResNet and Vision Transformer models across multiple datasets. Experimental results demonstrate that the proposed framework generalizes across multiple architectures, including ResNet-18, ResNet-34, ResNet-50, ViT-S/16, ViT-B/16, and ViT-L/16. Across all datasets, transformer-based models maintain higher accuracy under polynomial approximation, while convolutional models exhibit increased degradation as task complexity grows. In addition, latency analysis reveals that the proposed CLS token based inference pipeline significantly reduces computational overhead by eliminating depth-dependent operations, shifting the primary cost to encryption and decryption. These findings demonstrate that the proposed method not only adapts polynomial activations across architectures, but also enables efficientencrypted inference through both activation design and pipeline optimization.
Read the full abstract

All authors Zeinab Elkhatib, Kamrul Hasan, Ali Sekmen, Imtiaz Ahmed, Tariqul Islam

Machine Learning

Why don't Graph Transformers attend to the same token?

Alexandre Bismuth

In large language models (LLMs), attention concentrates heavily on the first token. This 'attention sink' is believed to protect deep models against over-mixing. Even though graph transformers face a related pressure, we find that a SoTA graph transformer, GRIT, rarely forms sinks on molecules, and that it sinks less when trained on larger graphs, the opposite of the trend in LLMs. In principle, a graph can prevent sinks: we prove and show empirically that its symmetry and locality limit the attention any node can receive. On molecules, however, with distinguishable atoms and global attention, neither limit prevents sinks; what stops them lies in the model. We find that the attention scoring function is the main cause. Replacing GRIT's gated logit with a dot product raises its sink rate from 0.09 to 0.68 on identical graphs. In Graph Transformers using dot-product attention, removing the sink gives about the same error as removing a random node, while GRIT's most-attended nodes matter to its predictions. Attention sinks are therefore a property of dot-product attention rather than of Transformers. In our experiments, they form when dot-product scoring has a distinguishable node to concentrate on.
Read the full abstract

Computation and Language

TopicMerge: A Divide and Conquer Approach to Topic Modeling

Matthew Raffel, Lizhong Chen, Sai Munikoti et al. (+1 more)

Recent advancements in topic modeling have split into two main paths: efficient embedding-based methods that offer scalability but are often outperformed on standard clustering metrics, and Large Language Model (LLM) approaches that achieve state-of-the-art clustering performance but at the cost of prohibitive computational expense and poor scalability. This paper introduces TopicMerge, a novel multi-stage pipeline that bridges these approaches to achieve both state-of-the-art clustering performance and computational efficiency. TopicMerge first employs an efficient clustering pipeline to deliberately over-segment a document corpus into a large set of fine-grained, semantically pure micro-clusters, which are then given initial semantic labels. It then introduces a novel, iterative merging algorithm where an LLM, guided by local neighborhood context, arbitrates on consolidating candidate pairs. These approved merges are then selectively executed and semantically relabeled by the LLM, hierarchically building broader, more coherent topics. Extensive experiments on the Bills and Wiki datasets show that TopicMerge achieves state-of-the-art clustering performance while reducing required LLM calls by over 92% compared to document-level methods, offering a practical and scalable solution.
Read the full abstract

All authors Matthew Raffel, Lizhong Chen, Sai Munikoti, Sameera Horawalavithana

Artificial Intelligence

Generalization Analysis of Biased Stochastic Gradient Methods for Minimax Problems

Shuang Zeng, Yunwen Lei and Yiming Ying

Minimax optimization refers to a class of optimization problems with some variables to minimize and other variables to maximize. Biased stochastic gradient methods have shown practical success in solving minimax problems to either improve robustness, enhance communication efficiency, or decrease computational costs. These successes motivate a lot of theoretical works to study the convergence of biased stochastic gradient methods, while their generalization analysis remains untouched. In this paper, we present the first framework to study the stability and generalization of biased stochastic gradient methods for minimax problems. We establish a connection between stability and generalization for minimax problems by relaxing the existing bounded gradient assumption to a bounded second moment condition. We then introduce a generalized Lipschitz-type condition on bias and gradient estimators, and derive a general stability bound to clarify the connection among bias, gradient estimators, and stability. We apply our general analysis to Zeroth-order and Clipped stochastic gradient descent ascent (SGDA), and derive stability bounds that match those of SGDA under appropriate smoothing/clipping parameters. We combine stability and convergence analyses together, and derive optimal excess risk bounds of order $1/\sqrt{n}$, where $n$ is the sample size.
Read the full abstract

Emerging Technologies

Hopfield Networks in Non-volatile Memory

Patrick Miller and Ulya R. Karpuzcu

Many important computational workloads are bound by the performance of the memory system that supplies data to conventional processors. Compute in Memory (CiM) promises to circumvent this bottleneck by performing the processing within the memory. Neural Networks are memory-bound, and a notable example of such is the Hopfield Network. This work explores the design space of non-volatile computing-in-memory (NV-CiM) architectures using an instruction-level simulator to evaluate application mappings and workload-driven architectural adaptations. The exploration examines the resulting design trade-offs and includes theoretical comparisons with representative CPU and GPU architectures.
Read the full abstract

Tuesday 13 October · still arriving

Cryptography and Security

Unmasking the Secret Code: Using Canaries to Show How Masked Context Controls Memorization in SFT

Avi Schwarzschild, Om Thakkar, Sahra Ghalebikesabi et al. (+2 more)

Canaries, or synthetic training records containing known secrets, are widely used to assess whether language models reveal information from their training data. But how much do these measurements depend on the canaries themselves? We study this question through controlled fine-tuning experiments with language and vision-language models, spanning synthetic tasks and tasks where training improves useful performance. Whether a secret is provided in the training input strongly changes what the model learns; for example, models can reproduce a secret in the input yet fail to recall it when it is removed, even after extensive repetition during training. When training instead requires predicting the secret from the surrounding context, later recall can become reliable. Repetition alone is therefore insufficient to characterize memorization. We also find that changing when a canary appears, or which other canaries accompany it, can substantially change measured memorization without increasing that canary's repetition count. These results show that canary measurements depend on how secrets are presented during training and tested afterward. Audits should test whether their conclusions hold across these choices before extending them to other training records.
Read the full abstract

All authors Avi Schwarzschild, Om Thakkar, Sahra Ghalebikesabi, Congzheng Song, Vinnie Monaco

Machine Learning · Cross-listed from physics.comp-ph

Global-in-time Spectral Duhamel-Picard Method for Post-Exposure Bake Simulation and Physics-Informed Neural Operators for EUV Lithography

Doyun Kim, Mircea Dusa, Werner Gilijns et al. (+1 more)

We present a global-in-time spectral Duhamel-Picard method with polynomial expansions for post-exposure bake (PEB) simulation. The coupled acid-quencher reaction-diffusion equations are reformulated as nonlinear Volterra integral equations using the Duhamel principle. Full trajectories are represented by finite polynomial expansions, with temporal coefficient fields determined over the bake interval by Anderson-accelerated Picard iteration, thereby avoiding sequential time marching. In 3-D PEB simulations, we employ Chebyshev and Legendre temporal bases while evaluating the Duhamel integrals using the corresponding Gaussian quadrature rules. We demonstrate that the proposed method simulates a large domain of $1200\times1200\times100$ voxels on the order of one minute. The representation is further incorporated into a physics-informed neural operator (PINO) with a mixed-transform variant of a factorized Fourier neural operator, which is trained without precomputed solution trajectories. Together, the integral formulation and its PINO extension enable accelerated, time-marching-free prediction of concentration fields at arbitrary times within the bake interval.
Read the full abstract

All authors Doyun Kim, Mircea Dusa, Werner Gilijns, Philippe Leray

Game Theory

Truthful and Expected-Payment-Optimal Model Routing in Graph-Based Agentic Workflows

Jiayuan Liu, Mingyu Guo, Jiarui Gan et al. (+1 more)

Assigning routine tasks to cheaper large language model (LLM) providers can reduce generation cost, but switching providers may require processing accumulated context again. Keeping dependent tasks with the same provider can instead reuse eligible context. Graph-based workflows therefore couple routing decisions across tasks. Independent providers may also privately know their real-time unit costs and strategically misreport them. We formulate this routing problem as a single-parameter procurement auction and propose STAR (Strategy-proof Topology-aware Agent Routing), which minimizes virtual-cost-weighted graph workloads. Its allocation rule and payments, computed by integrating workload over possible cost reports, make truthful reporting utility-maximizing regardless of other providers' reports and cover each truthful provider's execution cost. With independent cost distributions and non-decreasing virtual costs, STAR additionally minimizes the user's expected total payment to providers. Controlled synthetic experiments separate the execution-cost benefit of topology awareness from the payment benefit of virtual-cost optimization: the former grows with fan-out and transferred payload, while the latter increases gradually along the tested prior-shape heterogeneity path. A paired sensitivity study quantifies how payment savings vary with prior misspecification while strategy-proofness is preserved.
Read the full abstract

All authors Jiayuan Liu, Mingyu Guo, Jiarui Gan, Vincent Conitzer

Machine Learning

Historical-Feature Residual Adaptation for Sequential Screening

Yixiang Hong and Yuchen Zhang

Sequential screening policies must use acquired feedback effectively under a fixed experiment budget. We study acquisition-time residual adaptation: a small ranking correction built from fixed historical features and anchored to base scores recorded before feedback is revealed. The interface accepts a history-conditioned scorer and a candidate feature table; its primary instantiation augments AssayFormer's ten-dimensional representation with 64 historical directions while retaining the pretrained model. In retrospective evaluation on 20 official-test screens from 15 recorded sources, a comparison fixed before target-outcome access increased source-macro fraction of hits recovered from 22.58% to 28.61%, exceeding all six prespecified comparators. A same-feature ordinary ranking head achieved 25.73%. Supplementary strong logistic, rank-fusion, and unprojected controls achieved 26.42%, 26.02%, and 26.78%, respectively. A separate model-independent pilot on 13 reused development screens reconstructs the features without AssayFormer assets. Across three seeds, residual adaptation improves random-forest hit recovery from 28.87% to 31.27%, while logistic recovery changes from 30.22% to 29.78%. The results show effective pretrained-policy adaptation and useful transfer to an independent nonlinear base, with gains that depend on the base model.
Read the full abstract

Computation and Language

PersEvalBench: Can LLM Evaluators Judge Contextually Appropriate Use of Persona Information?

Xinyi Liu, Hooshang Nayyeri, Emine Yilmaz et al. (+3 more)

Large language model (LLM) evaluators are increasingly used to assess personalized assistants, yet whether they can reliably judge personalization has rarely been tested. We introduce PersEvalBench, a benchmark testing whether an evaluator can recognize appropriate and inappropriate uses of user history. It contains 403 multi-episode scenarios where user information should persist, transfer, or cease to apply. The benchmark covers nine controlled error subtypes across three families: under-personalization, over-personalization, and misapplied-state errors. Human audits confirm that generated errors match their intended types and that the prescribed uses of user information are reasonable. Across five LLM evaluators, recognition of controlled personalization errors remains limited even with the full dialogue history. On the human-evaluated subset, under-personalization shows the largest gap: humans recognize 94.4% of these errors, compared with 63.9% for the strongest reported evaluator. Evaluators often recognize that a response is problematic but reverse the error direction, mistaking failure to follow an applicable preference (UNDER) for use of an inapplicable preference (OVER). Easier evidence access offers little help: highlighting the relevant history turns raises error recognition only slightly, suggesting that the remaining difficulty lies in judging what role that information should play in the response.
Read the full abstract

All authors Xinyi Liu, Hooshang Nayyeri, Emine Yilmaz, Dilek Hakkani-Tur, Charith Peris, Hari Thadakamalla

Distributed Computing

Parallel Sparse Outputs for Tensor Compilers

Adrian Gushin, Pau Sum and Willow Ahrens

Sparse tensor compilers are commonly used high-performance tools that optimize tensor algebra programs over notoriously complex sparse data structures. Despite the ever-increasing size and scale of sparse datasets, there exists astonishingly little research on developing a parallel sparse tensor compiler. Existing approaches fail to parallelize or make assumptions about access patterns or output formats that limit generality. This paper presents Birdseed, the first tensor compiler to support all loop orderings over any standard sparse format in parallel. More specifically, Birdseed supports parallel sparse scatters and directly writes into list, bytemap, hash, and dense levels of fibertree sparse formats. Our approach exhibits asymptotic memory and time improvements over all previous work. We additionally demonstrate empirical speedups on challenging kernels to parallelize, such as histogram construction and SpMSpV.
Read the full abstract

Machine Learning

Did the LLM Answer the Question? A Semantic Reconstruction Game for Jailbreak Evaluation

Liam Davis, Romain Emil-Popescu and Suman Jana

Jailbreak attacks against large language models (LLMs) are evaluated using attack success rate (ASR), which depends on correctly identifying responses that fulfill prohibited requests. Existing evaluators can mistake refusals or topically related responses for successful attacks. We show that changing only the judge LLM can alter reported ASR by up to $82$ percentage points. We introduce the semantic reconstruction game, a training-free evaluator that decomposes compliance judgments into constrained semantic comparisons. Given only a response, an LLM reconstructs the request answered by its substantive content, excluding information merely repeated or refused. The game compares this reconstruction against similar decoy requests to determine whether it matches the original request, allowing abstention when neither candidate matches. Across three off-the-shelf LLM judges, the game achieves the lowest mean error against human labels among evaluated methods and reduces variation in reported ASR. It prioritizes avoiding false positives, even at the cost of missing some successful jailbreaks.
Read the full abstract

Computer Vision

Data-free Multistep Diffusion Distillation

Ngoc Duy Tran, Kien Do, Duc Thanh Nguyen et al. (+1 more)

One-step diffusion distillation achieves strong performance on small datasets but remains insufficient for high-resolution or complex text-to-image settings. Extending to multistep distillation often requires training data, suffers from early-step degradation, inconsistent outputs across sampling budgets, and incurs high computational cost. We identify that these limitations stem from simultaneous training, where a single adapted network approximates multiple score functions of different student distributions at once, causing biased distribution matching and unstable optimization. Therefore, we propose Progressive Multistep Data-free Distillation (PMDD), which assigns a dedicated adapted network per student distribution while avoiding the memory overhead of maintaining multiple adapted networks. Combined with a stage-wise unforget loss to preserve earlier-step generations, PMDD reduces optimization bias in the student, provides faster convergence, and a single unified model that supports multiple inference budgets. Extensive experiments show that PMDD matches or outperforms data-free few-step baselines at every multistep budget, reaches FID 1.99 on ImageNet 64x64 in two steps, below its 512-NFE teacher (2.32), and achieves FID 7.76 on zero-shot COCO 512x512, while requiring 4-5x less training compute than data-free DMDv2 to support three inference budgets.
Read the full abstract

All authors Ngoc Duy Tran, Kien Do, Duc Thanh Nguyen, Truyen Tran

Machine Learning

DTDD: Divergence-Triggered Dynamic Distillation for Reliable On-Policy Supervision

Qijia He, Zequn Li, Spencer Luo et al. (+6 more)

On-policy distillation (OPD) provides dense token-level supervision on student rollouts, but this supervision can become unreliable when the student and teacher diverge, either because of a large initial policy gap or compounding errors in long-horizon reasoning and tool use. We introduce Divergence-Triggered Dynamic Distillation (DTDD), an adaptive OPD framework that uses segment-level student-teacher divergence to trigger teacher takeover during training. When divergence triggers an intervention during reasoning, the student prefix continues to receive standard OPD supervision. The teacher then generates a recovery suffix, which the student learns through behavior cloning, using cross-entropy on the teacher-generated tokens. We evaluate DTDD in three regimes: long-horizon reasoning, distillation from a base initialization without offline supervised fine-tuning (SFT), and multi-turn agents. We also compare alternative trigger designs and teacher-intervention rates. DTDD reduces the post-warmup peak gradient norm by $8.6\times$ and, from a base initialization, reaches $21.3\%$ held-out math accuracy in $2.6\times$ fewer optimizer updates than OPD; both methods finish near $22\%$. It improves best-observed performance by $4.8$ and $4.1$ percentage points on ALFWorld with LoRA and agentic code, respectively. Ablations show that segment-averaged triggers outperform token-spike triggers. Behavior cloning is also more effective than the tested OPD surrogate for training teacher-generated recoveries. Across 23 experimental arms, takeover tends to help when the teacher substantially outperforms the OPD-trained student, provided that the teacher can recover from student-visited states, the student can learn those recoveries, and teacher intervention is neither excessive nor insufficient.
Read the full abstract

All authors Qijia He, Zequn Li, Spencer Luo, Tianle Chen, Pradyumna Narayana, Zixian Ma, Ranjay Krishna, Lan Nie, Sugato Basu

Artificial Intelligence

Beyond ICD Codes: Semantic IDs for Generative Medical Coding

Yihang Cheng, Qian Zhang, Yonghui Huang et al. (+2 more)

Generative approaches to automatic International Classification of Diseases (ICD) coding use large language models (LLMs) to predict ICD codes from clinical documents. However, original ICD code strings are not ideal generation targets because the ICD hierarchy does not always reflect clinical relationships, and LLMs struggle to interpret and generate these codes. We introduce SemICD, a framework that trains LLMs to generate Semantic IDs (SIDs) instead. SemICD constructs SIDs by hierarchically clustering embeddings of textual code profiles, allowing semantically related codes to share prefixes while each complete SID uniquely identifies one ICD code. To align the SID tokens with clinical semantics, we employ two auxiliary tasks for bidirectional generation between textual code profiles and SIDs. Experiments on MIMIC-III and MIMIC-IV with four LLMs show that SemICD outperforms direct code generation, improving Micro-F1 by 1.34-8.00 points. Our code is available at https://github.com/youwillbethebest/SemICD.
Read the full abstract

All authors Yihang Cheng, Qian Zhang, Yonghui Huang, Baoshuo Kan, Junyi Shen

Information Theory

The Capacity of the Noisy Torn Paper Channel and the Noisy Shuffling Channel

Junsheng Liu and Netanel Raviv

The torn paper channel breaks a codeword into fragments and returns all fragments without their original order. The noisy torn paper channel additionally introduces bit substitution errors to each fragment. Recovering the message therefore requires accounting for both bit errors and uncertainty about the fragment order. We establish a capacity upper bound for the noisy torn paper channel that matches the achievable rate of Ravi, Vahid, and Shomorony and resolves their conjecture. Our results also determine the capacity of the noisy shuffling channel of Shomorony and Heckel, in which all fragments have the same length, and resolve their open question.
Read the full abstract

Machine Learning

Verifier-Grounded Agentic RL for Provably Correct Code

George Ma, Yuhao Zhang, Bing He et al. (+5 more)

Reinforcement learning for code usually rewards passing tests, which a model can do while remaining wrong on the inputs the tests do not cover. A program verifier instead decides whether a program satisfies a formal specification for all inputs. Language models are almost always paired with verifiers at inference time only, and turning a verifier into a reward raises two difficulties. It reports only acceptance or rejection, which gives no gradient when every attempt fails, and a policy can cheat it, most simply by telling it to assume what the proof was supposed to establish. We build an agentic environment and reward that address both. The reward reads the verifier's diagnostics rather than its verdict, scoring a candidate by the weighted fraction of proof obligations it discharges, extended with hidden probes from a reference proof's annotations. An integrity audit outranks the verifier, rejecting any candidate that alters the specification or assumes what it was meant to prove. The model works as an agent against a live verifier over many turns, with a separately trained proof specialist as a frozen collaborator. Across Qwen3 models from 1.7B to 14B and a 35B Qwen3.6 mixture-of-experts model, held-out verification on code generation rises from 66.1% to 93.5% on DafnyBench and from 56.3% to 100% on SWE-Proof. This reward leads every other training signal we compare, including a binary pass-or-fail reward trained under otherwise identical settings, with the widest margin on code generation. When the agent must also deliver the repository patch, graded by each project's withheld tests, the fraction of held-out issues it resolves rises from 59.6% to 72.5%. Evaluated unchanged on three verification suites withheld from training, the checkpoints improve on all of them, including one whose language and verifier are absent from training.
Read the full abstract

All authors George Ma, Yuhao Zhang, Bing He, Zeren Shui, Ferhat Erata, Murali Krishna Ramanathan, Somayeh Sojoudi, Hao Zhou

Numerical Analysis

Variationally correct loss functions for nonlinear parametric problems and their neural approximations

Pablo Cortés Castillo, Wolfgang Dahmen and Jay Gopalakrishnan

We construct computable a posteriori error estimators for parametric families of nonlinear elliptic problems by measuring the residual of a variational formulation in the dual norm of a closed subspace of the test space. A single abstract theorem shows that the resulting estimator is bounded above and below by the error in the trial norm, i.e., it is a variationally correct loss function. Two instances are covered. The first considers an $L_2$-like test space, yielding a nonlinear first-order system least-squares loss function that certifies arbitrary, possibly meshless, approximations. The second breaks the test space, yielding a nonlinear discontinuous Petrov-Galerkin loss function computed by element-local problems. All assumptions are verified for quasilinear diffusion, including monotone fluxes and semilinear problems, and a nonlocal Kirchhoff model. Numerical studies train neural networks approximating the parameter-to-solution maps with the latter loss and measure the ratio of loss to squared error.
Read the full abstract

Robotics

Learning Keypoint Constraints from Visual Demonstrations for Safe Motion Planning

Zheng Qiu, Zhouyu Zhang, Ninghan Zhong et al. (+1 more)

We present Keypoint Constraint Learning (KCL), a framework for learning task-relevant motion constraints from visual demonstrations and for planning safely with the extracted constraints. To make constraint learning tractable from RGB-D observations, we represent demonstrations using sparse 3D object-centric keypoints and semantic mask centroids extracted through segmentation, visual feature clustering, and point tracking. In this reduced geometric space, we learn keypoint dynamics and infer distance- and orientation-based constraints that make the expert trajectories feasible and approximately locally optimal by enforcing KKT optimality conditions in the reduced state space. We then plan with the learned constraints under uncertainty, using robustness margins derived from accumulated prediction error, and provide a theoretical analysis of how keypoint perception error and learned dynamics error affect the recovered constraints. We evaluate KCL on visual demonstrations from navigation to rigid/deformable manipulation, showing that it recovers multi-stage geometric constraints and outperforms baselines in downstream safe planning.
Read the full abstract

All authors Zheng Qiu, Zhouyu Zhang, Ninghan Zhong, Glen Chou

Computation and Language

Qwenfant: Train Big Character-level BabyLMs on Concise Parallel Text

Lukas Edman and Alexander Fraser

This work concerns our submission for the Multilingual track of the 2026 BabyLM Challenge. We improve on our baseline using data that has been summarized and translated by an LLM, showing that higher-quality data can come from synthetic sources. We further improve with 2 simple changes: using a larger model, and using a character-level vocabulary. Both of these changes require LoRA-style fine-tuning for realizing their full potential. We dub our final Qwen3-based model Qwenfant, which, at the time of submission, ranks 1st in the Multilingual track.
Read the full abstract

Artificial Intelligence

Whose turn is it? Context-guided model routing for stateful agents

Anikesh Parashar, Avik Dutta, Vashu Chauhan et al. (+6 more)

Model routing can reduce the cost of LLM applications by assigning simpler requests to cheaper models without sacrificing the quality of the response. However, most routers treat requests as independent. Stateful agents do not operate under this assumption: a user turn may require multiple model and tool calls, its difficulty depends on the state produced by earlier turns, and changing models can sacrifice reuse of the conversational prompt cache. We study zero-shot, cost-aware routing at the user-turn boundary, systematically varying the routing rule, access to session history, candidate models that differ in size, reasoning effort, or model provider, and prompt-cache behavior. We evaluate on multi-turn versions of SheetCopilot and SpreadsheetBench 2, where persistent workbooks and cross-turn dependencies make isolated routing decisions especially brittle. At a matched offloading rate, giving the router session history improves task pass rate over the same router without history, with gains for every model pair; history-aware routing remains within one point of always using the expensive model. We also found that cost depends on execution dynamics rather than token prices alone: the cheaper model's per-task cost advantage narrows from 6.8x to 3.2x on longer sessions. A closed form cache analysis further shows that downward switches (expensive to cheap) recover their cache cost within two LLM calls, whereas upward switches do not break even under the evaluated conditions. Together, these results show that agent routing is a sequential decision problem: effective policies must account for what happened earlier in the session and for the asymmetric cost of switching models.
Read the full abstract

All authors Anikesh Parashar, Avik Dutta, Vashu Chauhan, Sumit Gulwani, Priyanshu Gupta, Hosein Hasanbeig, Vu Le, Ashish Tiwari, Jack Williams