Machine Learning
Justin Deschenaux, Alexandre Galashov, Andrew Campbell et al. (+4 more)
Diffusion models have revolutionized generative modeling for continuous data through the gradual refinement of a belief state. This iterative refinement has not yet carried over to discrete
diffusion models, which discard uncertainty at intermediate steps through categorical sampling (information collapse). We propose Simplex
Diffusion Models (SDMs), a framework that lifts the
diffusion process to the probability simplex to represent beliefs over categories. SDMs admit probability paths with closed-form reverse transitions and can be trained with a simple cross-
entropy loss. Contrary to earlier proposals such as Dirichlet Flow Matching which requires integrating an ordinary differential equation, we introduce a DDIM-like sampler with a tunable level of stochasticity. Because SDMs operate on samples on the simplex, they can carry uncertainty across denoising steps, which mitigates information collapse. On OpenWebText, SDMs are competitive with strong Discrete
Diffusion baselines, achieving $17.0$ GenPPL at $5.46$ unigram
entropy in 64 sampling steps, close to real validation data. Even without Self-Conditioning (SC), SDMs outperform masked and uniform
diffusion (with SC or predictor-corrector sampling) on code generation (TinyGSM, $T=0.1$; $49.0\%$ vs. $45.8\%$). Distilled down to 8 steps, SDMs solve $32.1\%$ of GSM8K problems, more than distilled Discrete
Diffusion models with 128 steps ($21.4\%$).
Read the full abstract ▾
All authors Justin Deschenaux, Alexandre Galashov, Andrew Campbell, Li Kevin Wenliang, James Thornton, Arnaud Doucet, Valentin De Bortoli
Cryptography and Security
INTCC: A Framework for Interactive Confidential Computing
Qingzhe Bing, Kaiyuan Zhang and Yinqian Zhang
Confidential computing leverages Trusted Execution Environments (TEEs) to ensure the confidentiality and integrity of data in use. However, TEEs rely on remote attestation to guarantee the integrity of their initial
memory state. This model is fundamentally at odds with interactive development workflows. In scenarios like LLM fine-tuning and exploratory data analysis, data processors need human-in-the-loop capabilities, including dynamic code injection, intermediate state inspection, and hyperparameter tuning, all of which inherently violate the static, one-time integrity guarantees of traditional remote attestation. To reconcile this tension, we propose the interactive confidential computing paradigm, a system architecture enabling untrusted data processors to execute dynamic, non-deterministic operations within TEEs without compromising data confidentiality. Driven by the insight that inherently unmeasurable human interaction must be excluded from the Trusted Computing Base (TCB), we logically partition the TEE into an interactive controller and a verifiable runtime. To realize this paradigm, we present INTCC, a framework featuring three key mechanisms: (1) a proxy-based dispatch system to preserve the native development experience; (2) a fine-grained information flow control mechanism based on a security
lattice to prevent data leakage; and (3) a
privacy-preserving verifiable execution mechanism to guarantee the runtime compliance of dynamic workflows. We implement INTCC on AMD SEV-SNP using Confidential Containers and evaluate it across diverse real-world workloads. Our experiments demonstrate that INTCC effectively balances security and interactivity, incurring a practical overhead of less than 5% for LLM fine-tuning and under 17% for data analysis relative to baseline execution.
Read the full abstract ▾
Artificial Intelligence
BaRe-Mem: Bayesian Reliability Memory for Robust and Adaptive Agent Consultation
Peilin Feng, Zhengyang Huang and Soujanya Poria
In multi-
agent systems, reliable consultation is challenging because advisor capabilities vary across tasks, and misleading information can make consultation worse than autonomous
reasoning. We introduce BaRe-Mem, an online
Bayesian reliability
memory for multi-
agent consultation. It estimates advisor reliability based on the central model's internal belief representations and updates these estimates from historical interactions. These estimates modulate the influence of advisor responses and guide the choice between consultation and autonomous
reasoning. Across nine
benchmarks and six central models, BaRe-Mem is more robust to misleading advisor information than debate and majority voting. On the more challenging tasks, it remains above autonomous
reasoning across all tested misleading levels. Moreover, we extend the BaRe-Mem mechanism to worker allocation in
agent teams. On the MuSiQue
benchmark, BaRe-Mem improves task completion over routing by historical success counts and identifies capable workers earlier.
Read the full abstract ▾
Artificial Intelligence
RareDx: Controlled Knowledge Integration and Graph-Grounded Policy Optimization for Rare-Disease Diagnosis
Bo Zhang, Yuchen Wang, Dongbai Li et al. (+6 more)
Rare-disease diagnosis is a long-tail
reasoning problem: phenotypes are incomplete, individual disorders are sparsely documented, and relevant evidence is distributed across ontologies, gene annotations, and biomedical text.
Language models consequently favor common conditions, miss rare candidates, or produce plausible but invalid names. We introduce RareDx, which couples controlled evidence use with knowledge-graph-grounded policy optimization. RareDx-Harness normalizes heterogeneous records into one ranked-diagnosis task and compares direct inference, static
retrieval, adaptive tools, and structured phenotype-gene-disease
reasoning over a shared knowledge layer. The training pipeline combines Top-10 post-training with RareDx-KGPO, our knowledge-graph-grounded policy optimization method. Its reward projects predictions into a canonical disease graph and integrates curated graded relevance, ontology proximity, biomedical similarity, and phenotype consistency. Vocabulary and output-budget constraints prevent dense partial credit from rewarding fabricated or overlong differentials. Across eight
benchmarks, the complete RareDx system centered on Qwen3.5-9B reaches 38.34 macro Hit@10, 1.60 points above GPT-5.5 under the archived protocol; a disjoint validation-selection audit retains a 6.80-point routing gain over Direct on held-out cases. The 27B system reaches 23.53/36.56/40.76 at Hit@1/5/10. Controlled ablations show that
retrieval is not uniformly helpful and that controlled routing is central to the gain. These results indicate that structured medical knowledge can turn a compact model into a competitive diagnostic ranker across heterogeneous long-tail settings in clinical practice.
Read the full abstract ▾
All authors Bo Zhang, Yuchen Wang, Dongbai Li, Matthew Yu Heng Wong, Qingkai Zeng, Lijun Wang, Tien-Yin Wong, Peng Cui, Tianyu Liu
Computational Complexity · Cross-listed from quant-ph
Sublinear Copies Suffice for Fidelity Estimation with Pauli Measurements
Jayadev Acharya, Abhilash Dharmavarapu, Yuhan Liu et al. (+1 more)
We present a protocol that estimates the
quantum fidelity, up to precision $\varepsilon$, between a known target state and unknown lab-prepared state with sublinear, $o(d^{0.9908}/\varepsilon^2)$, number of Pauli basis measurements.
Read the full abstract ▾
All authors Jayadev Acharya, Abhilash Dharmavarapu, Yuhan Liu, Nengkun Yu
Machine Learning
Graph World Models for Constrained Epidemic Policy Planning
Yiqi Su, Rashed Shelim, Lingyi Wang et al. (+2 more)
Epidemic policy planning often requires coordination between geographical regions, taking into account mobility-driven spillovers and how to make use of limited resources. Existing methods either lack action-conditioned models of coupled dynamics or cannot guarantee per-period feasibility. We present EpiMind, a graph world model framework for constrained epidemic policy planning across regions. A graph-factored recurrent state-space model generates joint policy-conditioned rollouts from regional latent beliefs, while graph-temporal ADMM optimizes regional interventions, enforces shared-resource feasibility through projection, and evaluates temporal specifications under the learned model. EpiMind reduces admission RMSE by 29% relative to graph-free dynamics modeling, plans within 1-5% of the best feasible constant policy with guaranteed shared-budget feasibility, and outperforms all deployable baselines across three resource budgets in real-context evaluation. These results demonstrate that graph-structured policy imagination with explicit constrained coordination supports effective epidemic interventions from learned dynamics.
Read the full abstract ▾
All authors Yiqi Su, Rashed Shelim, Lingyi Wang, Walid Saad, Naren Ramakrishnan
Computation and Language
Less Sycophancy, Stronger Refusal? Lessons for AI Safety from Mechanistic Interpretability
Xu Wang, Difan Zou and Xuansheng Wu
Reliable refusal of harmful requests is essential to the safe deployment of
language models. Because excessive eagerness to please users may undermine existing refusal capabilities, reducing sycophancy offers a potential route to stronger refusal beyond the harmful scenarios covered by safety training. We investigate this possibility using compensatory feature injection (CFI), a training technique designed to limit the acquisition of a target concept by supplying its associated activation during learning. Across three Qwen3.5 base models, we use sparse autoencoders (SAEs) to identify the top-ranked sycophancy feature from paired sycophantic and independent responses, then validate its behavioral influence through inference steering. We subsequently inject the selected feature during supervised fine-tuning on sycophantic targets. Positive injection reduces learned sycophancy after removal (by 62.0% relative to ordinary fine-tuning in 35B-A3B), whereas modest negative injection increases it. Unexpectedly, these reductions in sycophancy do not consistently improve direct refusal of harmful requests, motivating a narrower evaluation of the same harmful intents under user pressure. In this setting, ordinary fine-tuning on sycophantic responses substantially weakens refusal, while selected checkpoints trained with positive injection recover part of the loss, including approximately 95% in 35B-A3B. These findings show that persistent sycophancy reduction does not guarantee stronger direct refusal, while identifying recovery under user pressure as a distinct, conditional benefit of training intervention.
Read the full abstract ▾
Machine Learning
Learning the Robustness Mechanism with Bilevel Optimization
Yiyang Shen, Qihang Lin and Weiran Wang
We propose a distributionally robust learning framework where parameters defining the robustness mechanism are learned from held-out data instead of extensively tuned. Using bilevel optimization with both upper and lower level minimax problems, we create two instances of our framework to tackle setups with and without group labels in the training set. Theoretically, we provide sample complexity analysis for our robustness mechanism learning paradigm, showing that it achieves generalization guarantees comparable to exhaustive grid search while being more computationally efficient. Empirically, we evaluate our framework under a challenging setup when both intra-group and inter-group test distribution shifts occur at the same time, thereby demonstrating the efficacy and scalability of our method.
Read the full abstract ▾
Artificial Intelligence
Continuous Context Management
William Hoy, Jingxuan Fan, Nurcin Celik et al. (+1 more)
Long-horizon large
language model (LLM)
agents commonly retain their complete interaction history until compaction is triggered at a predefined threshold. We study Continuous Context Management (CCM), which performs compaction at every turn to prevent interaction history from accumulating in the active
prompt. At each turn, a CCM
agent emits an updated
memory together with an environment action; its next
prompt contains the original task, retained
memory, and newest observation rather than the complete transcript. We first evaluate CCM without fine-tuning on TerminalBench-2 using Claude Sonnet 4.6, Claude Opus 4.6, GLM-5, and Kimi K3. CCM substantially reduces cumulative input usage and active-
prompt size, although it lowers task success for most models while preserving performance for Kimi K3. We use GRPO with privileged full-history
distillation to improve CCM in open-weight models. A frozen copy of the student's initial model scores each sampled student action under the complete history reconstructed from that student's rollout, providing dense action-
token supervision without a separate teacher rollout or reference solution. On WebShop, this objective substantially improves CCM over GRPO at both evaluated model scales and surpasses full-history GRPO for Qwen3-4B-Instruct, though not for Qwen3-8B. On Endless Terminals, the augmented method provides a modest improvement over GRPO, with both CCM policies outperforming the untrained full-history baseline. These results demonstrate that CCM is a viable inference paradigm for
agents operating with substantially reduced retained context and that its performance can be improved through reinforcement learning with privileged full-history
distillation.
Read the full abstract ▾
All authors William Hoy, Jingxuan Fan, Nurcin Celik, Xu Pan
Computer Vision
Learning to Reason with Persistent Object States for Video Instance Segmentation
Yongxue Xu, Boxue Yang, Ziqian Liu et al. (+3 more)
Video segmentation models maintain object identities by carrying instance information across frames. Under prolonged occlusion, reappearance, or interactions between similar instances, however, an unreliable update can overwrite a valid history and cause persistent identity drift. We introduce POSReasoner, a trainable, plug-and-play framework that explicitly decides when an observation should change an object's state. Each persistent state records identity, confidence, and absence history. A sparse state-observation graph supports Propose-Verify
reasoning: provisional associations are revisited using object history, predicted presence, and competition among identities. The verified decisions determine whether to retain, update, reactivate, or suppress each state, while a learned gate controls the evidence written back to
memory. Only verified transitions update the persistent state used in subsequent frames. POSReasoner uses standard
video annotations and keeps the base model frozen, enabling integration with diverse VOS and VIS architectures. Experiments across long-term VOS and VIS
benchmarks show consistent improvements over strong baselines, with the largest gains under occlusion and object reappearance.
Read the full abstract ▾
All authors Yongxue Xu, Boxue Yang, Ziqian Liu, Shaoqiu Zhang, Rui Qian, Haopeng Chen
Numerical Analysis
Finite Element Approximation of a Hemivariational Inequality for Steady-State Heat Conduction: Double-Limit Convergence of Penalization and Discretization
Piotr Bartman-Szwarc, Anna Ochal and Domingo A. Tarzia
In this paper we study the numerical approximation and convergence analysis of a steady-state heat conduction problem with mixed boundary conditions. The physical model is governed by a hemivariational inequality depending on a heat transfer parameter $\alpha > 0$. We consider the finite element approximation of this penalized problem, as well as its corresponding limit problem with a prescribed constant temperature on a part of the boundary. The main theoretical result establishes the strong convergence of the discrete solutions. Specifically, we prove the double-limit convergence of the finite element approximations to the limit solution as the mesh size $h$ tends to zero and the penalization parameter $\alpha$ tends to infinity, independently and simultaneously. We further complement the convergence analysis with an error estimate of optimal order for the discrete limit problem, and we explain why an estimate uniform in the penalization parameter cannot be expected. The theoretical results are illustrated by numerical simulations on four examples verified by the method of manufactured solutions.
Read the full abstract ▾
Machine Learning
Optimal Networks for Agentic Information Aggregation
MohammadHossein Bateni, Zahra Hadizadeh, MohammadTaghi Hajiaghayi et al. (+2 more)
We study information aggregation in the networked learning model introduced by Kearns, Roth, and Ryu (SODA 2026). There is a fixed distribution over $d$ features and a common label.
Agents learn in topological order on a directed acyclic graph. Each observes a subset of the features and its parents' predictions, fits a linear predictor to minimize mean squared error, and passes only its prediction forward. The global predictor is the best linear predictor using all features. Kearns, Roth, and Ryu show that the output
agent's error approaches the global predictor's error along sufficiently deep paths with suitable feature coverage, while insufficient depth can prevent aggregation even in large networks. In contrast to their main focus on a given graph and feature allocation, we consider the limits of the model under two settings. In the adaptive designer setting, a designer chooses the graph, feature allocation, and output
agent knowing the distribution. In the oblivious designer setting, the designer fixes all three before an adversary chooses the distribution. Each
agent observes one feature and receives predictions from a limited number of parents. We call the aggregation exact when the output
agent matches the global predictor exactly. For $d\ge3$, we show that no finite depth guarantees exact aggregation for every distribution with one parent per
agent, even when the designer knows the distribution. In contrast, two parents per
agent suffice for exact aggregation even in the oblivious designer setting. A fixed graph, feature allocation, and output
agent achieve this for every distribution at depth $O(d\log d)$. Knowing the distribution reduces the depth to $O(d)$. Both constructions use $O(d^2)$
agents, with a very large constant for two parents. We show the bounds on the depth and number of
agents are all optimal up to constant factors.
Read the full abstract ▾
All authors MohammadHossein Bateni, Zahra Hadizadeh, MohammadTaghi Hajiaghayi, Mahdi JafariRaviz, Shayan Taherijam
Computer Vision
Look Before You Judge: Training-Free Region Mining for Grounded and Explainable Deepfake Detection
Chia-Ling Chen, Yu-Ting Ta, Jian-Yu Jiang-Lin et al. (+8 more)
Multimodal large
language models (MLLMs) can explain deepfake verdicts in natural
language, but such explanations are not necessarily visually grounded in the visual evidence underlying the prediction. A model may describe plausible artifacts inferred from
language priors rather than from image evidence. Existing grounding methods improve visual reliance through decoding or
attention interventions, but they generally strengthen grounding over the entire image, making them ill-suited for forensic artifacts that are subtle, spatially localized, and image-dependent. We propose Look Before You Judge, a training-free framework that formulates explainable deepfake detection as a sequential evidence acquisition process. Instead of directly predicting image authenticity from holistic visual
reasoning, our framework first identifies image-specific candidate evidence regions by contrasting the MLLM's decoder-to-visual
attention between an original image and its Gaussian-blurred counterpart. The identified regions are then inspected individually, and the resulting local evidence is integrated with the global image context before reaching a final verdict. The framework operates without manipulation masks, external forensic models, or parameter updates, making it directly applicable to off-the-shelf MLLMs. Across five open-source MLLMs on TriDF and MMTD-Set, our framework improves detection accuracy by up to 12.8%, reduces CHAIR by up to 33.4% and
hallucination rate by up to 21.3%, and outperforms representative training-free decoding and
attention methods.
Read the full abstract ▾
All authors Chia-Ling Chen, Yu-Ting Ta, Jian-Yu Jiang-Lin, Tai-Ming Huang, Ling Lo, Po-Ching Chen, Yan-Tsung Wang, Pei-Heng Li, Ling Zou, Hong-Han Shuai, Wen-Huang Cheng
Artificial Intelligence
ARISE: Adapting to Evolving Capability Gaps in Agentic Reinforcement Learning
Kun Feng, Yuchen Fang, Yiyang Tan et al. (+6 more)
As a long-horizon
agent improves through experience, previously observed weaknesses may recede while new limitations emerge, continually changing what it still needs to learn. Yet the learning process often remains tied to a static view of these needs: fixed behavioral criteria and training priorities can become misaligned with evolving
agent capabilities, while sparse task-level feedback makes such misalignment more difficult to detect. Even when capability gaps are identified, rollouts from the current policy may repeatedly reproduce the same failures rather than explore better alternatives. To address this, we introduce Adaptive Rubric-Skill Co-
Evolution (ARISE), a reinforcement learning framework that uses rollout evidence to continually adapt evaluation criteria, exploration guidance, and training priorities. Rubrics evolve to reward partial behavioral progress, while their paired skills are refined and selectively activated to guide exploration toward unresolved weaknesses. Alongside this co-
evolution, capability-based adaptive sampling prioritizes tasks that target behaviors needing further improvement. Experiments on two challenging long-horizon
agent benchmarks, SkillsBench and Terminal-Bench, demonstrate that ARISE successfully enhances both overall task performance and training efficiency. The project page is at https://foundation-model-research.github.io/ARISE .
Read the full abstract ▾
All authors Kun Feng, Yuchen Fang, Yiyang Tan, Shuqi Gu, Yongxiang Zhao, Yu Liu, Xingyu Lu, Lintao Ma, Kan Ren
Computational Engineering
Latent Twin Operator
Deepanshu Verma, Riley Chen and Matthias Chung
Surrogate models deployed on real physical systems rarely see data at fixed resolutions: sensor configurations vary across deployments and may evolve over time as sensing infrastructure changes. We introduce the Latent Twin Operator (LTO), a latent-space surrogate for time-evolving PDEs with a Convolutional Conditional Neural Process (ConvCNP)-style encoder and decoder that accepts a context set of $N$ sensor observations with arbitrary placement and can be queried at any resolution. A learned latent
evolution map advances the encoded state directly between arbitrary time points, decoupling temporal
evolution from the observation and query discretizations. We derive an explicit $\mathcal{O}(N^{-2/(3D)})$ rate for the context discretization error in spatial dimension $D$ under quasi-uniform refinement. We verify the predicted decay empirically on a 2D heat-equation
benchmark. Across time-dependent PDE
benchmarks, LTO achieves strong accuracy under one-step comparisons and transfers from native to coarser spatial resolutions with fixed parameters. On Navier–Stokes, its direct latent
evolution further reduces error over longer prediction horizons relative to recursive evaluation.
Read the full abstract ▾
Computer Vision
AutoRef: Harness Optimization for Agentic Multi-Reference Image Generation
Yuta Oshima, Ku Onoda, Yusuke Iwasawa et al. (+3 more)
Recent image generation models can take multiple reference images as input and combine them into a new image. However, multi-reference image generation remains challenging: models may omit or duplicate subjects from the references, or produce images in which multiple subjects appear unnaturally pasted. Recent work has proposed image generation
agents that combine image generation models,
reasoning models, and a harness, which is an executable program that specifies how reference images are interpreted, how generation is performed, how outputs are diagnosed, and how the final image is selected. In multi-reference generation, however, references play different roles and outputs must satisfy many criteria at once, such as fidelity to each reference and the naturalness of the whole image, so many parts of the harness could be improved, from how references are processed to how outputs are diagnosed. This makes it hard to predict which changes will improve performance and by how much, and good harnesses difficult to design by hand; indeed, human-written harnesses vary widely in performance. We therefore propose AutoRef, which optimizes the harness automatically while keeping both models frozen: a coding
agent iteratively rewrites the harness code. AutoRef separates the tasks whose feedback informs proposals from the tasks used to select candidates, and continues the search from a beam of the top-ranked harnesses on the selection tasks. Using this procedure, we discover AutoRef-Harness, which improves the open-weight FLUX.2 [klein] 4B from 5.72 to 7.37 on held-out four-reference tasks of the MultiBanana
benchmark, matching or exceeding proprietary models including Nano Banana Pro and GPT-Image-1.5. Without re-optimization, the same harness also improves results when the generator, number of references,
benchmark, evaluator, or
reasoning model differs from those used in the search.
Read the full abstract ▾
All authors Yuta Oshima, Ku Onoda, Yusuke Iwasawa, Masahiro Suzuki, Yutaka Matsuo, Hiroki Furuta
Machine Learning
Let the Neurons Die: Exploiting ReLU-Induced Model Degradation
Kexin Li, Wenjun Qiu, Joshua Abraham et al. (+2 more)
Rectified linear unit (ReLU) networks can suffer from dying
neurons, where units with persistently negative pre-activations produce zero outputs, blocking gradients through their activations. To exploit this failure mode, we present three training-time availability attacks based on data ordering and poisoning. We begin with the basic dynamic data-ordering attack (DOA), which greedily constructs a training prefix by selecting the next example that minimizes the target layer's post-update weight sum, aiming to push ReLU units toward negative pre-activations without modifying training samples or labels. We then develop two poisoning attacks, IG-DOA and IG-SKA, which use gradient inversion to synthesize class-conditioned samples by matching reference gradients in adverse model states constructed through data ordering or soft knockout, respectively. Soft knockout rearranges weights across adjacent layers to concentrate negative contributions. On a fully connected ReLU network trained on MNIST, ordering 100 of 60,000 training examples reduces test accuracy from 96% to 95% after only five epochs. Adding 200 poisoned samples from a single class reduces test accuracy to approximately 86-88% after five epochs in most evaluated conditions, compared with approximately 96% under clean training. These results demonstrate that ReLU-targeted data ordering and poisoning can impair learning without directly modifying the victim model's parameters.
Read the full abstract ▾
All authors Kexin Li, Wenjun Qiu, Joshua Abraham, Aditi Maheshwari, David Lie
Machine Learning
GeoGAE: Scalable Graph-Level Autoencoding via Hyperball Cloud Representations
Radosław Nowak, Anna Bielawska, Bogusz Stefańczyk et al. (+2 more)
Embedding structured objects into Euclidean spaces has enabled a wide range of successful machine learning applications. Such objects include words, documents, image patches, time series, and graph nodes. In contrast,
embedding entire graphs remains a challenging problem. Existing methods either sustain the original order of the graph nodes or match the output nodes to the input ones, both of which create scalability issues. In this work, we propose a graph representation as a cloud of hyperballs, which allows us to define a specific, typically unique, node ordering. Based on this representation, we propose GeoGAE, an autoencoder, in which the
Transformer encoder translates a hyperball cloud into a graph-level
embedding, and the
Transformer decoder translates the graph-level
embedding back into the graph. This formulation enables the model to capture both the global graph structure and local relational patterns. We evaluate our method on multiple graph
datasets, spanning various domains. The results demonstrate effectiveness of our method in encoding and reconstructing graphs from their
embeddings.
Read the full abstract ▾
All authors Radosław Nowak, Anna Bielawska, Bogusz Stefańczyk, Maciej Sanocki, Paweł Wawrzyński
Machine Learning
Deep Epistemic Value Functions for Optimistic Exploration
Leander Diaz-Bone, Marco Bagatella, Jonas Hübotter et al. (+1 more)
Principled exploration in reinforcement learning requires an
agent to quantify its epistemic uncertainty and act to resolve it. Uncertainty over the value function provides a natural signal for exploration, yet existing deep approximations remain brittle and perform inconsistently. The central challenge is therefore to scale these ideas robustly. We conduct a systematic empirical study of how epistemic uncertainty is represented, propagated, and optimized in deep epistemic value functions, and uncover distinct failure modes along each of these axes. These findings motivate DEVOTE, a model-free reinforcement learning
algorithm that controls how uncertainty generalizes beyond observed data, stabilizes its temporal propagation, and preserves adaptation to the resulting non-stationary exploration objective. Across reward-free exploration and challenging continuous-control tasks, DEVOTE reaches novel states more effectively and achieves higher task return than strong model-free and model-based exploration baselines. These results provide evidence that deep epistemic value functions are a promising path toward scalable, principled exploration.
Read the full abstract ▾
All authors Leander Diaz-Bone, Marco Bagatella, Jonas Hübotter, Andreas Krause
Computation and Language
Beyond Token Scale: Chunk-Level Sparse Autoencoders for Reliable Semantic Feature Discovery
Xu Wang, Yifan Yang, TingHao YU et al. (+1 more)
Sparse autoencoders (SAEs) expose features that help us understand and steer
language models, but faithful reconstruction does not guarantee informative concepts.
Token-level objectives reward lexical and formatting details alongside semantic content, all competing for a limited sparse budget. We introduce a family of chunk-level SAEs that encode mean-pooled activations over chunks, each a contiguous span of
tokens: Mean-Chunk reconstructs the observed chunk, Cross-Chunk predicts an independently processed neighbor, and Joint-Chunk combines both targets. These designs separate the effect of a larger observation unit from that of predicting information shared across passages. With matched training data, chunk-level SAEs remain powerful interpretability tools while learning reliable semantic features that capture high-level concepts and respond selectively to relevant content. Their strengths are complementary: Mean-Chunk improves high-level feature discovery,
reasoning detection beyond surface cues, and steering; Cross-Chunk leads document
retrieval and classification transfer while producing selective, persistent features. Changing what an SAE sees and predicts yields reliable semantic features for more meaningful tasks. We demonstrate their practical value through gains across downstream tasks such as
retrieval,
reasoning detection, and steering.
Read the full abstract ▾
All authors Xu Wang, Yifan Yang, TingHao YU, Difan Zou
Machine Learning
Reward-Aligned Reweighting for On-Policy Distillation
Haofeng Xu, Junwei Su, Lansong Diao et al. (+2 more)
On-policy
distillation (OPD) trains a student
language model with dense feedback from a stronger teacher on student-generated trajectories. Yet standard OPD weights
token-level
distillation terms uniformly, implicitly treating local teacher preference as a proxy for correction utility. A decision's task value, however, depends on how the student completes the subsequent
reasoning. This mismatch can cause imitation to suppress viable student strategies or reinforce paths the student cannot reliably execute. Verified trajectory outcomes provide complementary evidence about continuation quality, but do not directly identify the utility of individual decisions. We introduce Reward-Aligned Reweighting for On-Policy
Distillation (R$^{2}$-OPD), which uses outcome agreement and the magnitude of teacher–student disagreement to continuously reallocate teacher supervision. It gives reward-aligned corrections greater relative influence while retaining dense feedback, moving beyond uniform imitation and hard filtering. Our analysis formalizes the mismatch between local teacher preference and student continuation value and establishes sufficient conditions for reallocation to improve first-order task progress over uniform OPD. Across seven mathematical
reasoning benchmarks, R$^{2}$-OPD achieves the highest average accuracy among the compared training methods in both cross-size and same-size
distillation. It outperforms standard OPD on all seven
benchmarks, with average gains of 3.5 and 2.4 percentage points for 1.7B and 4B students, respectively. An extension to code generation yields an average gain of 1.6 percentage points over standard OPD. These results highlight outcome-guided supervision allocation as an effective way to translate dense teacher feedback into stronger student performance across model scales and task domains.
Read the full abstract ▾
All authors Haofeng Xu, Junwei Su, Lansong Diao, Wenchao Zhou, Chuan Wu
Robotics
Inspection-SPARS: Task-Oriented Sparse Roadmaps for Inspection Planning
Adir Morgan, Oren Salzman and Kiril Solovey
Inspection planning seeks a minimum-length collision-free
robot tour that observes a given set of points of interest (POIs). Sampling-based methods reduce this continuous problem to a graph inspection planning (GIP) problem over a discrete roadmap, which is then solved using combinatorial solvers. Dense roadmaps capture diverse inspection viewpoints and motion shortcuts, and thus admit higher-quality solutions, but they induce large combinatorial search spaces on which state-of-the-art GIP solvers struggle to find good solutions within practical time budgets. Roadmap sparsification—restructuring a dense roadmap into a compact representation that preserves connectivity and path lengths—can alleviate this burden. However, existing sparsification approaches are either agnostic to the underlying inspection task, or strive to ensure coverage of the POIs without accounting for the quality of the resulting inspection plan. We present Inspection-SPARS, which is, to our knowledge, the first inspection-roadmap sparsifier with POI coverage and path-quality guarantees relative to the dense roadmap. To this end, we generalize the SPARS framework, a popular task-agnostic sparsifier, from purely
geometric criteria to task-oriented ones, introducing an inspection-aware vertex admission mechanism that treats POI coverage as a first-class sparsification criterion alongside connectivity and path quality. Experiments in realistic 3D environments show that Inspection-SPARS reduces vertex and edge counts by 4-8x while preserving coverage, allowing the GIP solver to compute tours up to 25% shorter than with the dense roadmap or state-of-the-art inspection roadmap. More broadly, Inspection-SPARS shows that sparsification can be made task-aware without sacrificing guarantees on solution quality.
Read the full abstract ▾
Artificial Intelligence
MechBench: Can AI Scientific Agents Discover Mechanisms Beyond Phenomenal Laws?
Zihan Yu, Jiadong Zhang, Jialin Cheng et al. (+2 more)
Scientific discovery requires not only recovering mathematical laws that describe observable behavior, but also identifying the mechanisms that generate them. Existing
benchmarks for symbolic regression and scientific
agents primarily evaluate phenomenal-law recovery, leaving mechanism discovery largely untested. We introduce MechBench, a
benchmark that explicitly separates these two capabilities. Each task is defined by a mechanistic model, a structured set of scientifically meaningful relations whose joint consequences entail an observable phenomenal law, while
agents receive only observational data and scientific context. We evaluate mechanism recovery through mechanism probes, which query internal scientific consequences that cannot be inferred from the phenomenal law alone. To reduce reliance on memorized textbook mechanisms, we construct unfamiliar variants through controlled, scientifically interpretable mutations of canonical mechanisms, and screen for mechanistic indistinguishability to exclude ambiguous instances admitting comparable competing mechanisms. Experiments across representative scientific
agents reveal a substantial phenomenal–mechanism recovery gap: for Codex with GPT-5.6-sol, phenomenal-law accuracy reaches 35.00% on the Core-set while mechanism accuracy is only 13.75%, with mechanism recovery failing in 64.29% of cases where the phenomenal law is correctly recovered. The gap widens as mechanisms become increasingly mutated, and even providing the correct phenomenal law leaves mechanism recovery below 50%. These results reveal a substantial generalization gap in mechanistic
reasoning and establish mechanism discovery as a distinct challenge beyond recovering observable scientific laws.
Read the full abstract ▾
All authors Zihan Yu, Jiadong Zhang, Jialin Cheng, Jingtao Ding, Yong Li
Machine Learning
One Proposal for Every Margin: Zero-Shot Amortized Sequential Importance Sampling for Binary Matrices
Ruishuo Chen, Weijia Li, Xun Wang et al. (+3 more)
In ecology, psychometrics, and the analysis of social and financial networks, binary
matrices are often analyzed conditional on their observed row and column sums, which restricts the problem to a finite sample space of
matrices with the same margins. Two fundamental problems are to count this space and to sample uniformly from it. Sequential importance sampling (SIS) addresses both with independent weighted samples and an unbiased count
estimator, but its efficiency depends critically on the proposal distribution. Existing proposals are analytically designed, and their accuracy can vary substantially with the margins. We show that the ideal SIS proposal, under which every weight equals the count and the variance vanishes, is exactly the policy of a generative flow network (GFlowNet) with unit reward on every
matrix that has the given margins. We therefore propose MarginFlow, a framework that turns the design of the proposal into a learning problem and amortizes it across margins by exploiting their self-similarity. Every partial
matrix is itself an instance with reduced margins, so one set
transformer that reads the remaining margins serves every margin. We train MarginFlow on a pool of 1904 margins and evaluate it zero-shot on 1190 held-out margins, synthetic and real, from $3\times3$ to $870\times6$. On 1187 of the 1190 margins it matches or beats the best of 31 analytically designed configurations, chosen post hoc for each margin, and its median effective sample fraction is 99.8%. On the 56 margins where that best loses more than one nat of effective sample size, MarginFlow wins every one and raises the median effective sample fraction from 10.3% to 94.1%.
Read the full abstract ▾
All authors Ruishuo Chen, Weijia Li, Xun Wang, Yu Chen, Leheng Cai, Longbo Huang
Machine Learning
Improving Generative Model Self-Training with Geometrically Modified Outputs
Patrick Batsell, Thomas Walker and Richard Baraniuk
Self-training generative models - the continued improvement of a model using its own outputs - is becoming increasingly important as high-quality training data becomes scarce. However, naively finetuning on model-generated samples leads to degradation through model collapse and the model autophagy disorder. Negative-guidance self-training methods turn this degradation into a useful signal, using a model finetuned on its own outputs to guide the original model toward improved generation. Existing methods, however, take the negative signal in standard model outputs as given. We instead ask whether this signal can be explicitly strengthened. We introduce Geometrically Modified Outputs (GMOs), which reweight the singular values of the generator's input-output Jacobian to increase the influence of its leading singular directions. This
geometric modification amplifies the mode-seeking behavior and distortions of standard outputs, providing a stronger and more targeted negative signal for self-training. Across a range of one-step generative models, GMOs consistently improve the performance of negative-guidance methods, including Neon and SIMS, compared with using standard model outputs.
Read the full abstract ▾
Operating Systems
Argus: Agentic, Reference-Calibrated, Tree-Guided, System-Software-Level Bottleneck Localization
Vlad-Petru Nitu, Harsh Songara, Konstantinos Sgouras et al. (+3 more)
Operating system (OS) code can account for a substantial share of CPU execution time. First, as application logic is offloaded to heterogeneous accelerators (e.g., GPUs), the CPU increasingly acts as an orchestrator, spending cycles in driver calls, data movement, and synchronization rather than in application code. Second, workloads such as serverless functions frequently invoke OS services. At the same time, the OS is a complex codebase spanning many subsystems (e.g.,
memory management, networking), making it hard to localize the specific code path responsible for a slowdown. Existing profilers expose measurements that require interpretation(e.g., perf and Intel VTune) or can perturb short operations when extensively instrumented (e.g., ftrace). Diagnosing OS bottlenecks can therefore require repeated kernel instrumentation and manual interpretation. We introduce Argus, an agentic LLM-based profiler that produces instrumentation code and autonomously reasons over potential OS-level bottlenecks. Argus integrates two key mechanisms: (i) a calibration methodology that involves collecting a measurement from an idle system and using it as a reference point to discover potential bottlenecks, and (ii) a tree-based data structure that represents the different OS execution paths, improving the
agent's bottleneck localization accuracy. Argus aims to identify a specific kernel code path rather than stop at a subsystem-level diagnosis. In two case studies, we employ Argus to autonomously discover bottlenecks present in the
memory management subsystem caused by (i) a THP aggressor co-running with other applications, and (ii) applications that incur different types of page faults. Argus produces 19 times fewer incorrect deep-path diagnoses than the strongest evaluated LLM-based baseline, which lacks reference calibration, while preserving low time-to-diagnosis (approximately 31 s)
Read the full abstract ▾
All authors Vlad-Petru Nitu, Harsh Songara, Konstantinos Sgouras, Spiros Galanopoulos, Konstantinos Kanellopoulos, Onur Mutlu
Computer Vision
ReVA: A Scene-Centric Dataset Beyond Repetition for Remote Sensing Video Question Answering
Zhen Yao, Likai Wang, Yuming Yang et al. (+9 more)
Multimodal Large
Language Models (MLLMs) have demonstrated remarkable advances in remote sensing. However, existing remote sensing
multimodal reasoning benchmarks exhibit two critical limitations: they rely on (i) template-driven questions, which causes repetitive questions; and (ii) static images that fail to capture the inherent temporal nature of drone/
UAV videos. This leaves systematic evaluation of remote sensing
video reasoning largely unexplored. To address this gap, we introduce ReVA, a new
dataset for remote sensing
video question answering, designed to assess spatiotemporal, scene-centric, and
reasoning-oriented capabilities of MLLMs. ReVA comprises 2,438 drone
videos spanning 18 cities worldwide (580K frames) and 22K high-quality question-answer pairs across 11 challenging QA tasks. We develop a semi-automatic annotation pipeline that leverages Text LLMs and MLLMs for question-answer generation with human verification. We evaluate 23 proprietary and open-source
Video LLMs on ReVA, exposing fundamental limitations of current models. These findings position ReVA as a critical
benchmark toward better remote sensing
video understanding and temporal
reasoning capabilities for real-world deployments. Our code and
dataset are available at: https://github.com/zyaocoder/ReVA
Read the full abstract ▾
All authors Zhen Yao, Likai Wang, Yuming Yang, Zhihao Zheng, Bo Lang, Qiuyu Tang, Jialu Sheng, Jingqi Xu, Yuehai Yang, Jumal Barker, Xiaowen Ying, Mooi Choo Chuah
Cryptography and Security
Detecting False Data Injection and Unstable Operation in Smart Grid via System-Aware Graph Boundary Learning
Emad Efatinasab, Denis Donadel, Mirco Rampazzo et al. (+1 more)
Cyber-physical power systems increasingly rely on data-driven tools to detect instability and support reliable grid operation. However, reliable stability prediction is difficult when unstable operating configurations are rare, sensitive, or unavailable during model development, since collecting such data safely and at scale is often impractical. At the same time, False Data Injection (FDI) attacks can manipulate reported system parameters to trigger false instability alarms or conceal unsafe operation, and such threats in Decentral Smart Grid Control (DSGC) systems remain largely unexplored. These challenges are rarely addressed jointly, leaving a gap between stability prediction and attack detection that this work aims to close. In this paper, we introduce StarGNN, a graph learning framework that learns the stable operating region exclusively from clean stable configurations and uses a single abnormality score to flag reported configurations that should not be trusted as evidence of safe operation, covering both genuine instability and unseen FDI manipulations. Each configuration is represented as a producer-consumer
star graph and processed by a role aware graph neural network, with physics constrained pseudo-negatives generated by perturbing reaction time and price response parameters standing in for the unavailable unstable and attack data. A single threshold, calibrated only on held-out stable data, is used without task or attack specific adjustment. Evaluated on nine unseen FDI scenarios, StarGNN detects between 0.780 and 0.973 of attacks on stable configurations and retains a post attack instability recall between 0.972 and 0.999 on unstable ones, with 0.899 recall against a stronger adaptive attacker, showing that stable-only boundary learning can support both stability prediction and attack detection without access to genuine unstable labels or attack samples during training.
Read the full abstract ▾
All authors Emad Efatinasab, Denis Donadel, Mirco Rampazzo, Chuadhry Mujeeb Ahmed
Machine Learning
An RL View of OPD: Least Square Policy Distillation for Sample-Efficient LLM Reasoning
Shangzhe Li, Yuxiao Yang, Tianrun Yu et al. (+4 more)
We study on-policy
distillation (OPD) through the lens of reinforcement learning, establishing a connection between the reverse-KL objective in OPD and KL-regularized policy optimization. Building on this connection, we introduce Least-Square Policy
Distillation (LSPD), an RL-inspired framework that brings optimistic exploration and off-policy data reuse from value-based RL into policy
distillation. LSPD preserves policy diversity through exploration while improving rollout efficiency by repeatedly learning from previously collected trajectories. Our theoretical analysis connects LSPD to optimistic value-based learning and shows that its idealized formulation achieves a sharp $\tilde{\mathcal O}(\log K)$ regret bound under online exploration. Empirically, LSPD consistently outperforms existing
distillation baselines across six mathematical
reasoning benchmarks and diverse teacher-student settings, with average gains of +1.59 points in Avg@16. Remarkably, through Pass@k evaluations up to k=64, we found that LSPD better preserves policy diversity by achieving stronger performance as k grows. Its fully off-policy variant achieves comparable performance to vanilla OPD using only the first 25% of rollout batches. Together, these results provide an RL perspective on OPD that offers both a principled interpretation and a practical route toward more effective and rollout-efficient
language model
distillation.
Read the full abstract ▾
All authors Shangzhe Li, Yuxiao Yang, Tianrun Yu, Kaixiang Zhao, Xiaoyun Wang, Taylor W. Killian, Weitong Zhang
Computer Vision
SolveEdit: Benchmarking Visual Problem Solving in Generative Models
Wenjie Shu, Yexin Liu, Harold Haodong Chen et al. (+9 more)
Machine intelligence is often evaluated through abstract
reasoning problems, yet many real-world problems are visual, such as arranging objects, repairing layouts, or tracing routes. Solving these problems requires understanding a scene, inferring what must change to achieve a goal, and realizing that change without disturbing unrelated content. However, existing
benchmarks mainly evaluate perception, generation, or explicitly specified transformations, leaving goal-driven visual problem solving underexplored. To bridge this gap, we introduce SolveEpIT, a
benchmark for visual problem solving through scene transformation. Given an image and a goal, a model must infer a valid transformation from the request, the scene, or a visually expressed rule, then execute it while preserving unrelated content. SoLvEEDrr contains 2,728 cases. Atomic transition contracts specify required and protected conditions, enabling SoLvEScoRE to measure completion and unintended changes without a single reference output. The strongest evaluated model achieves only57.0% SolvEScore. We further introduce SolveEdiT-PLAN, a two-stage visual planner that instantiates the transition before generation. Under matched single-generation evaluation, it improves SoLvEScoRE by 9.1 points on average across three tested generators, including a gain from 57.0% to 71.6% for GPT-Image-2, without modifying the editor.
Read the full abstract ▾
All authors Wenjie Shu, Yexin Liu, Harold Haodong Chen, Xuerui Qiu, Zehan Wang, Yidi Zhang, Yizhan Chen, Zunwei Wang, Minghao Liu, Qi Chen, Harry Yang, Xiaogang Xu