Large-scale empirical tuning and comparison of default optimizers for variational inference
Trevor Campbell, Jonathan H. Huggins, Kyurae Kim, Charles C. Margossian
AI 速览：作者在1092个贝叶斯推断优化问题上大规模评测56种随机梯度优化算法，发现没有单一方法占优，但选5种算法组合即可接近最佳观测性能。
Black-box variational inference (BBVI) is a methodology for posterior approximation that relies on stochastic optimization. In practice, the stochastic optimizers underpinning BBVI generally require extensive problem-specific tuning, which undermines its promise as a truly "black box" inference algorithm. However, over the past decade, many new adaptive stochastic optimization algorithms have been developed that reduce or remove entirely the need for tuning. In this work, we investigate this new collection of adaptive methods in the context of BBVI, with the goal of establishing the current state of the art in tuning-free optimization-based inference. In particular, we present a large-scale empirical evaluation of 56 stochastic gradient-based optimization algorithms applied to 1092 Bayesian inference optimization problems, involving over 550,000 individual optimization runs and 15 core-years of compute. The optimization algorithms we evaluate are chosen to represent a wide spectrum of recent approaches and the benchmark problems are chosen to span a range of difficulty, with posterior target dimension 1-10^4, condition number 1-10^8, and a range of variational families. Our results show that no single method dominates, but running a selection of 5 algorithms suffices to reliably get close to the best-possible observed performance. We thus provide a strong baseline for applications where expert tuning is not possible and for comparison when developing new stochastic optimization algorithms.
ENFORCE: Nonlinear Constrained Learning with Adaptive-depth Neural Projection
Giacomo Lastrucci, Artur M. Schweidtmann
AI 速览：提出 ENFORCE 架构，通过自适应投影模块 AdaNP 在预测中施加非线性等式与不等式约束，并在基准任务上报告满足给定容差 ε 的约束。
Ensuring neural networks adhere to domain-specific constraints is crucial for addressing safety and trustworthiness while also enhancing inference accuracy. Despite the nonlinear nature of most real-world tasks, the majority of existing methods are limited to affine (equality) or convex (inequality) constraints. We introduce ENFORCE, a neural network architecture that uses an adaptive projection module (AdaNP) to enforce nonlinear equality and inequality constraints in the predictions up to a specified tolerance $\varepsilon$, and exactly in the affine-in-$y$ case. For affine constraint sets, we prove that the associated projection mapping is non-expansive (1-Lipschitz), ensuring stable gradient propagation. For nonlinear constraints, we establish local convergence analysis under standard regularity conditions. We evaluate ENFORCE on multiple tasks, including function fitting, real-world engineering case studies, and learning optimization problems. For the latter, we introduce a class of scalable optimization problems as a benchmark for nonlinear constrained learning. In the benchmarks, the predictions of our architecture satisfy nonlinear equality and inequality constraints up to a prescribed tolerance $\varepsilon$, while maintaining scalability with tractable computational complexity at training and inference time.
VA-TDM: A Unified View for Improving Few-Step Diffusion Models by Velocity Adversarial Distillation
Yihong Luo, Tianyu Pang, Tianyang Hu
AI 速览：提出速度对抗蒸馏（VAD）框架并构建VA-TDM，以轨迹分布匹配和中间转移监督提升少步扩散/流模型的生成质量。
Few-step distillation offers efficient generation with diffusion and flow models, but preserving quality with only a few sampling steps remains challenging. Distribution-matching methods via score distillation provide a strong foundation, yet they rely on accurately tracking the student's changing distribution. At intermediate generation steps in few-step sampling, student-score estimation can be difficult, and commonly used estimators may describe a different distribution from the one used to update the student. Flow matching allows a more flexible approach to supervision through velocity fields constructed from the student's sampling process. Building on this perspective, we introduce Velocity Adversarial Distillation (VAD), which uses a learned critic to compare student and teacher fields and turn their disagreement into feedback for learning. VAD unifies score distillation, Fisher-based matching, and moment matching, also explaining existing heuristic MMD updates and delivering new training objectives. We further develop VA-TDM, which follows trajectory distribution matching: providing supervision for partial denoising along the student trajectory. Beyond this, VA-TDM combines transition-aware online velocity learning and dual-form VAD supervision across intermediate student transitions. Extensive results show that VA-TDM is powerful for learning few-step generators, achieving state-of-the-art four-step performance in both text-to-image and text-to-video generation across various backbones and scaling effectively to 33B MiniMax-H3. In particular, our VA-TDM-H3 notably outperforms the few-step Fast-H3 and is closer to the original many-step H3 according to human preference.
What Should Data Teach? Moving Bottlenecks Across Circuit, Store, and Use
Yixiao Chen, Ke Cheng, Yue Liu, Jun Zhang, Jie Jiang
AI 速览：作者提出一种“电路视角”的数据诊断原则，将语言模型训练中的瓶颈分为计算形成、内容可用与路径选择三类，并报告相应数据干预在350M模型实验中的效果。
What should data teach a language model at a particular point in training? A circuit view reveals three distinct bottlenecks: forming a computation, making its required content available, and selecting among available routes. A shared diagnosis-to-data principle connects them: localize the missing operation, preserve its causal relation, vary shortcut-bearing context, and re-audit the residual. Formation-sensitive selection and prerequisite ordering accelerate a binding-matching-transport path; a brief early prefix from the same training multiset retains a validation advantage through 100B tokens. Availability counterfactuals then distinguish writing content from invoking available memory, while paired supervision and context-opportunity ranking improve matched route decisions and long-context answer likelihood. A continuous 350M-model experiment connects the three interventions on the same facts: early circuit training improves subsequent learning, and the complete sequence outperforms stage-replacement controls on facts withheld from Use teaching. Independent query surfaces and opposed-source decisions expose conditional arbitration as the remaining frontier. Together, these results show why a change in the limiting operation calls for a change in supervision, not merely a new ranking of difficult examples.
Transforming Integer Grids to Spatial Intuition: Perception, Not Planning, Governs ARC-AGI-3
Harsidak Singh Banwait
AI 速览：作者提出开放权重智能体 SARBLOH，主张在 ARC-AGI-3 交互推理中起决定作用的是空间感知而非规划能力，并用零参数感知引擎把整数网格帧转成对象清单与几何简报。
Sample-efficient fluid intelligence requires agents that rapidly form working models of unfamiliar interactive environments. ARC-AGI-3 tests this across uninstructed visual games scored by Relative Human Action Efficiency, which penalizes exploratory waste quadratically. When frontier models read raw integer grids, serialization destroys spatial topology, and they spend hundreds of actions rediscovering basic boundaries and symmetries. We introduce SARBLOH, an open-weight agent built on the premise that spatial perception, not planning capacity, governs interactive reasoning on ARC-AGI-3. A zero-parameter O(HW) perception engine turns each frame, in 35 ms, into an object inventory with bipartite temporal tracking, a geometric briefing that separates proven invariants from affordance hypotheses, and an annotated canvas. Paired with a code-as-action runtime, the agent writes algorithmic hypotheses instead of low-level moves. Across eight presentations, grounded perception raises the rubric score from 3/28 to 25/28 points while cutting token use by 89.4%. On 23 held-out games, geometric invariants generalize with 100% accuracy, while single-frame affordance guesses fall to 6-50%, so active verification probes remain necessary. On the public set, Run A (briefing off) solved 143 of 183 levels in 8.2 hours at 59.4% single-pass efficiency; Run B (briefing on) reached 63.5%. Within the nine-hour budget on one Kaggle RTX PRO 6000 GPU, SARBLOH scored 16.53 on the hidden Kaggle evaluation, 13.7x the first Milestone winner.
Reverse to Advance: Teleoperation-Cost Effective Hard Policy Learning from Reversed Easy Tasks
Qiyuan Qiao, Ge Yuan, Can Wang, Dong Xu
AI 速览：提出一种通过时间反转简单任务来自动生成并精炼监督数据、从而低成本学习高难度操作策略的框架。
High-quality teleoperation datasets are costly to collect, particularly for hard tasks. We observe that many tasks exhibit directional asymmetry: completing the forward hard task is difficult, whereas reversing it by relaxing or disrupting the environment is comparatively easy. This suggests that reversed easy-task trajectories can serve as a scalable supervision signal for the hard task, reducing the cost of manual demonstration collection. However, reversed data can be noisy, and directly training on it may yield suboptimal policies. To enable largely automated acquisition and effective use of reversed data, we propose a teleoperation-cost effective framework for hard policy learning via temporal reversal of easy tasks, consisting of three key components: a closed-loop data collection pipeline that alternates between hard-task and easy-task policies to autonomously reset the environment and generate diverse trajectories; a hierarchical data refinement pipeline that temporally inverts easy-task rollouts and filters low-quality motion using kinematic priors and a critic-guided advantage filter; and an iterative policy learning method that trains the hard-task policy using both initial reversed easy-task demonstrations and the filtered reversed data in a continuous online learning loop. By combining automated collection, hierarchical refinement, and iterative learning, our method enables scalable, reliable training of complex, high-precision manipulation tasks. Across two simulated benchmarks and real-robot experiments, we demonstrate that our method improves hard-task success rates with higher data efficiency and more stable training compared to reversal-based and reinforcement-learning baselines, without requiring extensive hard-task teleoperation.
Using profiles of cognitive capability to assess AI suitability for workplace tasks
Jonathan Prunty, Marko Tešić, Patrick Quinn, José Hernández-Orallo, Lucy Cheke
AI 速览：提出一套基于共同认知能力维度的画像流程，将AI系统能力与工作任务需求对齐，用于判断哪些任务适合自动化、保留给人类或人机协作。
Organisations deploying AI face a scoping problem: which tasks can be automated, which should remain with humans, and which are best shared between the two. Aggregate benchmark scores provide little insight into where systems will succeed or fail in practice, while human judgements of model capabilities quickly become outdated. We introduce a pipeline that profiles agents and tasks using a shared set of core cognitive capabilities. Cognitive capability profiling infers an agent's capabilities from performance on a benchmark battery annotated for the cognitive demands of each item. Task requirements weighting elicits from domain experts the relative importance of these same capabilities for their work. As both use a common set of cognitive dimensions, they can be updated independently as models and roles change, and combined to estimate AI suitability at the level of a domain, organisation, role, or individual duty. We validate capability recovery on synthetic agents, profile six AI systems, and elicit task requirements from 410 employees across six occupational domains. AI systems differed more across cognitive dimensions than across model families, while workplace activities converged on a shared cognitive core. The resulting scores provide a comparative scoping tool for identifying promising candidates for piloting and areas where current systems are unlikely to be well suited. We discuss extending the framework to profile human workers alongside AI systems, moving from AI suitability towards human-machine task allocation.
Enterprises run business transactions through multi-agent AI systems spanning clouds, large language models (LLMs), orchestration harnesses, and systems of record. Finance receives AI cost only at settlement grain, aggregated by SKU and charge period. The business question (what did this order, claim, or case cost in AI?) cannot be answered from billing data, because providers never see enterprise transaction boundaries. The FinOps Open Cost and Usage Specification (FOCUS) gives billing data one schema across providers but remains settlement-grain even as it adds AI token and requester attribution. We formalize this grain mismatch, derive seven requirements, and propose the Two-Ledger Model: a settlement ledger of FOCUS-conformant billing data joined to an attribution ledger of invocation-grain usage records built on OpenTelemetry GenAI telemetry. A layered correlation spine (trace context, A2A/MCP metadata, business-key correlation, probabilistic matching) assigns each record to its business transaction with recorded method and confidence; a versioned rate card prices usage in near real time; period reconciliation ties both ledgers to the invoice, and the unexplained residual identifies shadow AI. Costs are reported per order, claim, or outcome with agentic cost metrics. A simulation with known ground truth shows the spine sustaining cost coverage above 99% as trace propagation falls to zero, the reconciliation residual recovering planted unmetered spend within 0.24 percentage points while flagging provider billing errors, and mid-flight gating cutting p99 transaction cost by up to 61% at a 2-4% false-pause rate. We propose the attribution ledger as an invocation-grain FOCUS profile for standardization.
AI 速览：AdaLoop 提出一种按问题条件进行视觉循环计算的框架，让冻结的预训练视觉编码器在多次迭代中复用其变换，以适配不同问题所需的视觉处理深度。
Most vision-language models (VLMs) treat the vision encoder as a static feature extractor: it encodes an image once, independently of the question, and leaves subsequent reasoning to the language model. Yet different questions may require different visual operations and processing depths. We present AdaLoop, a framework for question-conditioned visual recurrence in VLMs. AdaLoop adapts visual computation to the question and reuses the encoder's transformations across multiple iterations, with the pretrained vision weights frozen. We also introduce VisRO, a procedural dataset and generation pipeline for training and evaluating vision-centric reasoning. On six VisRO tasks, AdaLoop outperforms standard fine-tuning and question-conditioned single-pass baselines with matched training data and optimizer updates. These gains hold across three VLM families on both in-distribution and out-of-distribution examples. Our findings suggest that pretrained vision encoders can contribute to reasoning through task-directed, iterative visual computation.
A fast acceptance classifier could select revisions for a slower evaluator. We ask whether agreement on existing papers supports that use. We compare a classifier of word and document features with two released text-only acceptance predictors fine-tuned from the same language model with three billion parameters. For 198 papers from the International Conference on Learning Representations, we compare model rankings of the original papers and of paragraph rewrites screened automatically for unchanged facts, claim strength, and attribution. The classifier ranks original papers closely with one predictor and moderately with the other, but shows no detectable agreement with either on which rewrites raise scores. The tests specified in advance used reduced numerical precision. Rounding error dominated the small score responses. We therefore rescored the released weights in full 32-bit precision as a post hoc check; edit agreement remained near zero. At this precision, when each predictor chooses among five rewrites per paper, its average raw score is under two points out of 100 higher than with random choice. The evidence covers presentation rewrites and two open predictors from one family; transfer to costly evaluators remains untested. Revision selectors should therefore be checked against the intended evaluator on the revisions they rank.
Phevos Paschalidis, Constantinos Daskalakis, Devavrat Shah
We study causal inference under outcome interference for sequential, observational settings. Specifically, we consider settings where the binary outcomes over N units are Markovian across T time steps. At each time step, the outcomes of N units have dependencies captured through an Ising model; each outcome is also impacted through an external field capturing the effects of its treatment as well as latent confounders. Similar to panel data literature, these latent confounders are modeled to have a low-rank factor structure. Our data is a single sample from this high-dimensional distribution. To estimate causal quantities of interest, we provide a computationally efficient method based on Maximum Pseudo-Likelihood Estimation (MPLE) for learning the model parameters. Under mild assumptions, we establish non-asymptotic consistency for parameter estimation and show this translates to faithful estimation of causal quantities of interest after sampling from the learned model. We demonstrate the efficacy of the method through synthetic experiments as well as a real-world case-study investigating causal effects of vaccine rates on COVID-19 death rates within US counties nationwide.
Farouk Mokhtar, Joosep Pata, Michael Kagan, Javier Duarte
The workflow from particle collision to physics analysis passes through a series of reconstruction steps that are traditionally modular and disconnected, with no shared representation linking low-level detector data to high-level analysis tasks. We show that casting event reconstruction as a machine learning problem naturally produces such a shared representation. We repurpose a machine learning model trained for particle-flow reconstruction (MLPF) to perform three distinct analysis tasks: jet flavor identification, jet energy regression, and missing momentum regression. By appending the per-particle latent representations learned during reconstruction as additional input features, we substantially improve over baselines that use conventional particle-level features alone. We further demonstrate that a single linear layer trained using only the latent representations achieves competitive performance against state-of-the-art baseline architectures, and outperforms the baseline for missing momentum regression with approximately 35 times fewer parameters. These results demonstrate that the latent representations learned during reconstruction encode essential physics information needed for downstream analysis, establishing MLPF as a foundation model and offering a concrete step toward an end-to-end pipeline from detector data to physics analysis.
Quan Minh Nguyen, Hoang M. Ngo, Bhakti Chowkwale, Yasaman Keshtkarjahromi, My T. Thai
One-shot and few-round federated learning (FL) reduce communication burden by constructing a global model within one or a few communication rounds, but often achieve lower performance than conventional multi-round FL. Without frequent synchronization, client models may drift apart, causing neurons that learn similar features to occupy different positions across clients. At the same time, local data heterogeneity can cause corresponding parameters to differ in their importance to each client's predictions. Together, these challenges cause direct averaging to combine mismatched neurons and assign equal weight to parameters despite differences in their importance. We propose FedFiOT, a Fisher-weighted optimal transport framework for one-shot and few-round FL that leverages Fisher information to quantify parameter importance and guide neuron alignment and model aggregation. FedFiOT jointly optimizes neuron correspondences and global model parameters through alternating minimization of a shared Fisher-weighted objective. Theoretically, we show that FedFiOT's optimization procedure monotonically decreases the layer-wise merging objective and establish a stationarity bound that quantifies the effects of client heterogeneity and nonuniform Fisher weighting.Experiments across multiple datasets demonstrate consistent improvements over competing baselines in both one-shot and few-round federated learning settings.
Xuanming Cui, Jaiminkumar Ashokbhai Bhoi, Chionh Wei Peng, Adriel Kuek, Ser Nam Lim
Dynamic Scene Graph Generation (DSGG) aims to capture objects and their evolving relations in videos. Despite recent progress, the practicality and quality of generated scene graphs remain limited compared to the rapid advances in Multimodal Large Language Models (MLLMs). In this work, we revisit DSGG from two fundamental perspectives: task setup and model design. From the task setup perspective, we identify two key limitations of the current recall-oriented evaluation protocol: (i) a severe precision-recall trade-off, and (ii) uninformative and redundant relation generation. To better assess practical usefulness, we introduce five additional metrics that provide a more comprehensive evaluation of the quality of generated dynamic scene graphs. From the model design perspective, we explore directly using MLLMs for scene graph generation and establish a strong MLLM-based DSGG baseline through three design changes. First, we replace the conventional bottom-up pipeline with a top-down reason-then-locate strategy. Second, we reformulate frame-wise dynamic graphs as Temporal Relation Set (TRS) prediction, improving both efficiency and performance. Third, we introduce Importance-Aware Finetuning (IAF) to encourage more relevant and diverse relation generation. Extensive experiments on Action Genome, VidVRD, and PVSG show that our approach consistently achieves state-of-the-art performance.
Chang Liu, Ke Han, Niccolò Biondi, Gloria Isotton, Elisa Ricci, Davide Talon, Nicu Sebe
Sign language generation (SLG) maps written sentences into signing motion while preserving both lexical content and fine-grained articulation. This requires a model to decide what signs should be expressed and to realize them through precise handshape, trajectory, facial expression and body coordination. Existing pose-based systems and reconstruction-only motion tokenizers often entangle these two decisions, leaving the generator without an explicit text-aligned intermediate target for what to sign before refining how to sign it. To address this issue, we introduce Semantic Residual Vector Quantization (SRVQ), a framework for SLG that frames generation as a two-level process moving from semantics to kinematics. The newly introduced SRV Quantizer learns a hierarchy in which the first residual level encodes sign content and subsequent levels refine pose, handshape, trajectory, and body coordination, all within a single quantizer and without separate semantic representations or external planning stages. Through word-aware conditioning for hierarchical code generation, a masked semantic generator predicts semantic codes and a kinematic generator completes the residual codes for full SMPL-X motion synthesis. We evaluate using retrieval and FID in a shared text-motion embedding space, alongside standard geometric metrics. On CSL-Daily and Phoenix-2014T, SRVQ achieves R@1 of 82.3 and 70.3, outperforming the previous SOTA by 30 and 46 points, respectively.
Algorithm-driven services, including search, recommendation, and advertising, rely on multi-stage model pipelines. While research prioritizes model optimization, post-deployment monitoring focuses on operational reliability. Yet operationally healthy systems may still produce poor-quality results, exposing a gap in existing evaluation and monitoring practices. However, a systematic problem formulation and benchmark for this task remain lacking. To bridge this gap, we formulate System Quality Diagnosis (SQD) for algorithm-driven services, a new task that extends quality evaluation beyond identifying poor outcomes to attributing their causes across system components. Focusing on industrial search, we propose WISE, an agentic framework for end-to-end quality diagnosis that discovers user-facing issues and traces their origins to underlying algorithmic components. To improve diagnostic reliability, we introduce ReTrace for constructing evidence-consistent trajectory supervision and Adaptive Harness-Policy Optimization (AHPO) for failure-guided optimization of the agent harness and model policy. To enable systematic evaluation, we construct WISE-BENCH, a real-world multimodal search benchmark that combines user-facing evidence, internal pipeline traces, and expert diagnosis annotations. Extensive experiments show consistent gains across model scales, with WISE achieving 75.20% end-to-end accuracy on Qwen3.5-35B-A3B. In production, WISE supports quality auditing, online quality assurance, and model optimization, reducing diagnosis costs and improving downstream relevance models.
LLMs on edge devices are constrained by limited memory and bandwidth. Although 1.58-bit ternary weights greatly reduce model storage, long-context inference is increasingly bottlenecked by KV-cache traffic and attention. Existing accelerators also use separate datapaths for ternary linear layers and attention, incurring hardware and execution overhead. We present SPARTA, a sparse ternary LLM accelerator for efficient long-context edge inference. SPARTA integrates: (1) a phase-aware sparse dataflow exploiting heterogeneous sparsity across compute-bound prefill and memory-bound decode; (2) a unified compute fabric mapping ternary SpMM/SpMV and bit-serial attention onto shared processing elements to exploit zero-weight and zero-bit sparsity; and (3) an adaptive two-stage attention engine combining page-level Top-k selection to reduce K-cache accesses with token-level Top-p pruning to reduce V-cache traffic. SPARTA reduces KV traffic by 4.1-4.8x over full attention and 1.3-1.5x over a state-of-the-art sparse-attention scheme, with task-dependent quality loss. Compared with NVIDIA Jetson AGX Orin, SPARTA achieves 3.9x higher generation throughput and 8.3x higher end-to-end energy efficiency on average; versus AMD Ryzen 7 9700X, the gains are 5.0x and 81.7x. It also delivers 1.6-12.0x higher throughput, 2.0-4.2x higher energy efficiency, and 1.3-51.4x higher area efficiency than FIGLUT, TeLLMe, and MCBP.
Reinforcement learning with verifiable rewards (RLVR) often adopts GRPO-style group-relative updates, sampling multiple rollouts per prompt to construct normalized learning signals. However, merely increasing the number of rollouts does not reliably strengthen learning: under GRPO-style group normalization, per-rollout policy-gradient features can concentrate into a low-rank, signed geometry, causing substantial cancellation during aggregation and weakening the effective update. We address this failure mode with SALT, a Subspace-Adaptive geometry pLug-in componenT that uses sample-wise gradient geometry to reweight the coefficients of group-relative updates. SALT estimates a dominant shared subspace from the mini-batch Gram geometry, decomposes group-relative coefficients into shared and residual channels, and adaptively amplifies the residual channel when signed cancellation is severe. Across diverse reasoning-oriented RLVR benchmarks and model scales, SALT improves effective update geometry and performance without modifying the reward model or the rollout sampling procedure.
Forecasting foundation models are increasingly used as frozen feature extractors for time series classification, but it remains unclear whether their pretrained representations offer an advantage over strong classifiers that do not require pretraining. This raises an important question: do these representations provide useful information beyond what the model architecture alone can capture, particularly when labelled data are scarce? We present a controlled benchmarking of two forecasting models, Reverso and Chronos-Bolt, on six large datasets from the MONSTER benchmark, using 1%, 10%, and 100% of the available training labels. We compare their frozen representations against four established time series classification methods and random-weight versions of the same forecasting models. Our results show that pretraining improves classification accuracy over random-weight counterparts on five of the six datasets, with gains of up to 23 percentage points. However, the frozen representations do not reliably outperform the strongest non-pretrained baseline in any setting: they perform worse in 15 of 17 settings and tie in the remaining two. Further experiments show that restoring amplitude information recovers part of the performance gap, while fine-tuning and forecast-error features do not close it. These findings highlight the importance of strong task specific baselines and random weight controls when evaluating forecasting foundation models for classification. Our code is available at https://github.com/Rachet2006/frozen-forecasters-monster.
Fine-grained visual reasoning is essential for remote sensing in urban management, yet remains underexplored in the evaluation of multimodal large language models (MLLMs). We introduce UMBench, a fine-grained urban remote sensing benchmark built from unmanned aerial vehicle (UAV) images for multiple urban management tasks such as traffic monitoring, urban construction, environmental quality, and disaster warning. Evaluations on UMBench reveal that current MLLMs often rely on scene-level priors and overlook fine-grained visual evidence. To address this issue, we further propose a novel evidence-first chain-of-thought method that directs models to inspect fine-grained visual evidence in a training-free manner. Experiments show that our method can improve the performance of both open-source and commercial MLLMs on UMBench.
Coding agents powered by models such as GPT-6-Astra connect spatial reasoning with executable robot behavior, raising the question of how such agents can produce lasting improvements in robot capabilities from physical interaction. In this work, we study embodied recursive self-improvement (RSI) through real-world exploration, code refinement, and deployment using PyRUA-REAL, a framework in which coding agents retain interaction experience as persistent memory and reusable executable skills that inform subsequent exploration and improvement. Our empirical study places physical manipulation at the center of the improvement process, examining how agents collaborate, what kinds of execution feedback support effective code refinement, and how task specifications and previously acquired skills influence robustness and task transfer. We further investigate the efficiency and capability of the resulting programs when deployed without involving an online agent for decision-making. Across eight real-world tasks, PyRUA-REAL achieves an average success rate of 91.1%. In complementary evaluation on 42 tasks from the RoboDojo benchmark, it achieves an average success rate of 84.4%. This study provides empirical guidance for turning physical exploration into reusable robot skills that support continued improvement and robust, efficient deployment.
Kareem Amin, Rudrajit Das, Alessandro Epasto, Adel Javanmard, Dennis Kraft, Mónica Ribero, Sergei Vassilvitskii
The rapid adoption of generative AI and Large Language Models (LLMs) has spurred interest in synthetic data as a privacy-preserving alternative to sensitive real-world datasets. However, generating high-utility synthetic data often carries the risk of memorizing and regurgitating private information from the training corpus. In this work, we present a customizable empirical auditing framework designed to detect and explain such data disclosures. Our framework introduces a mechanism to distinguish between "true disclosures"-where the system directly reproduces a user's information-and "phantom disclosures''-where the system incidentally generates a user's data. By partitioning input data into training and holdout sets and applying rigorous statistical hypothesis testing, we determine if observed disclosures are consistent with strict privacy baselines, such as zero-learning or specific Differential Privacy (DP) bounds. Crucially, this approach requires no model access, no canary insertion, and no reference model training -only the synthetic output and a held-out control set. We demonstrate that this framework effectively functions as a membership inference attack, providing empirical lower bounds on privacy leakage that are tighter than prior data-based auditing methods. Our approach is model-agnostic, applies to any synthetic data generation mechanism, and requires orders of magnitude fewer computational resources than shadow-model or canary-based alternatives.
Yuchi Wang, Rongyao Fang, Weihao Yu, Xilin Wei, Tiancheng Gu, Dingkang Yang, Hang Zhang, Dayiheng Liu 等 10 位作者
Beyond basic tool use and coding skills, effectively incorporating visual references into agentic generation is equally essential for applying AI agents to real-world productivity tasks. We study this crucial capability through webpage development, a practical and readily evaluable setting in which existing benchmarks remain largely idealized: they typically focus on reconstructing visual targets or generating webpages from textual instructions alone, overlooking the more common workflow of developing from visual examples. Specifically, we introduce RefCraft, a benchmark comprising 300 realistic multimodal webpage-development tasks that combine natural-language instructions with screenshots from one to five reference webpages. Agents must selectively interpret and utilize these references while rendering, visually inspecting, and iteratively refining their outputs. We evaluate the resulting open-ended artifacts using independently verifiable keypoints and complementary programmatic and VLM judges. Across nine frontier (M)LLMs, the strongest system achieves an overall score of only 52.12 out of 100, revealing that visual-reference-conditioned agentic generation remains a substantial challenge for current models in practical webpage-development settings.
Sign Language Production (SLP) generates continuous signing from a linguistic input and drives accessibility technologies such as signing avatars. Current generators, diffusion models included, regress joint coordinates as an unstructured vector, so nothing ensures that the generated body respects anatomy or the mechanics of movement. Physics has entered SLP only through geometric loss terms, whose benefit for intelligibility has not been verified, and the dynamics of the signing body has not been modelled. We present PIDiffSign, a diffusion model informed by physics at two levels. For kinematics, a forward-kinematics head outputs bone directions and lengths and composes them into an anatomically exact skeleton. For dynamics, a rigid-body model of the upper body with anthropometric parameters supports two training priors. A physical-plausibility suite that includes two laws of motor control evaluates the generated signing. Under a controlled protocol, the head lowers back-translation word error rate by about five points on PHOENIX14T and CSL-Daily, whereas geometric penalties have no measurable effect; a control experiment traces the gain to the kinematic-chain representation and not to bone-length conservation. The dynamics priors gain about one further point over matched non-physical controls on PHOENIX14T, but not on CSL-Daily. Generated signing follows the laws of motor control as closely as real signing but carries markedly less kinetic energy. The exact skeletons can drive rigged avatars directly, and the suite offers a physical benchmark for future SLP systems.
Scientific image-analysis workflows require agents to select among diverse capabilities like description, detection, counting, and segmentation. Current multi-tool architectures conflate intent interpretation with visual execution, masking whether a failure arises from flawed reasoning (misrouting) or failed perception (invalid tool execution). This lack of auditability renders agents unreliable for high-stakes scientific automation. We introduce ZETA, an extensible architecture that operationalizes the decoupling of routing policy from action execution. ZETA resolves scientific intent before invoking specialized visual tools, treating measurement as a deterministic operation grounded in explicit geometry rather than generative output. This architecture exposes the "black box" of agent decision-making, allowing routing correctness and downstream execution to be evaluated as independent failure modes. To our knowledge, we provide the first controlled study in scientific vision-agent routing that isolates explicit routing-policy specification while holding the legal action interface fixed. Evaluation on a zero-shot benchmark of 51 scientific-intent queries shows that explicit routing semantics can substantially improve routing accuracy, particularly at semantically adjacent action boundaries, although improvements vary across models. For Qwen 3.8, explicit routing-policy specification improves segmentation routing accuracy from 6.2% to 75.0% and detection routing accuracy from 60.0% to 100%. ZETA's execution traces make routing decisions and downstream artifacts separately observable, providing a basis for distinguishing intent-selection errors from visual execution failures. ZETA establishes the necessary infrastructure to isolate execution errors from interpretive ones, a critical requirement for deploying reliable scientific vision agents.
We study logistic regression for continual learning with linear predictors in a high-dimensional statistical setting. Existing theoretical work has largely focused on continual linear regression with variants of L2 regularization, motivated by their analytical simplicity and connection to the kernel regime of neural networks. We first derive the optimal L2 regularization strength for continual linear classification and characterize its generalization performance, recovering behavior analogous to prior results for continual linear regression. Next, we examine the effect of L2 regularization in deeper models using a minimal two-layer diagonal linear network. In the single-task setting, L2 regularization on the weights of such models is known to promote sparsity in the resulting predictor, yielding a rich, sparse feature-learning regime beyond the kernel regime (a similar effect arises in two-layer ReLU networks). However, in continual learning, where L2 regularization is typically centered on previous-task weights, this sparsity bias is no longer preserved. We therefore propose a weighted variant of L2 regularization that extends this beneficial sparsity bias to continual learning. We show theoretically and empirically that, for a sufficiently sparse teacher, our variant converges substantially faster than directly learning a linear predictor with conventional continual L2 or L1 regularization, trading some long-horizon bias for markedly improved performance over moderate task horizons.
Dawid Wolkiewicz, Anastasiya Pechko, Przemysław Spurek, Piotr Syga
The growing adoption of photorealistic 3D facial avatars, particularly those utilizing efficient 3D Gaussian Splatting representations, introduces new risks of online identity theft, especially in systems that rely on biometric authentication. While effective adversarial masking methods have been developed for 2D images, a significant gap remains in achieving robust, viewpoint-consistent identity protection for dynamic 3D avatars. To address this, we present AEGIS, the first privacy-preserving identity masking framework for 3D Gaussian avatars, concealing identity from automated face recognition while preserving the avatar's perceptual realism and animation usability. AEGIS applies adversarial perturbations to the base (DC) color coefficients of the Gaussian primitives, guided by an ensemble of pre-trained face verification networks, embedding the protection directly in the 3D representation so that it applies across viewpoints and animations, without retraining or modifying the avatar's geometry. Evaluated against black-box recognizers, AEGIS reduces face retrieval and verification accuracy to near-zero while maintaining high perceptual fidelity. It also preserves key facial attributes such as age, gender, and emotion, demonstrating strong privacy protection with minimal visual distortion. We further introduce a viewpoint-robust leakage metric that quantifies residual identity leakage across an avatar's rendered viewpoints, which single-view evaluation cannot capture.
Quan Minh Nguyen, Min-Seon Kim, Hoang M. Ngo, Trong Nghia Hoang, Hyuk-Yoon Kwon, My T. Thai
Membership inference attacks (MIAs) pose a serious privacy threat in federated learning (FL). While MIAs have been extensively studied in standard FL, the recent shift toward federated fine-tuning introduces new and largely unexplored attack surfaces. In this work, we show that federated prompt-tuning, which adapts pre-trained foundation models using lightweight input prefixes, exposes a novel and effective vector for membership inference. We propose PromptMIA, a membership inference attack tailored to federated prompt-tuning, in which a malicious server introduces adversarially crafted prompts and exploits their updates during collaborative training to determine whether a target data point belongs to a client's private dataset. We formalize this threat via a security game and demonstrate that PromptMIA achieves consistently high attack advantage across diverse benchmark datasets, substantially outperforming current SOTA federated MIAs. We also provide a theoretical lower bound on the attack advantage that explains the observed empirical behavior. Finally, we show that existing MIA defenses are often ineffective against PromptMIA, highlighting the need for defense mechanisms specifically tailored to prompt-tuning in federated settings.
Alessio Giorlandino, Sebastian Goldt, Antoine Maillard
Large language models excel at factual recall, yet the fundamental limits of storing input--output associations with neural networks remain unclear. We study these limits in a minimal setting: a linear associative memory that maps $p$ input embeddings in $\mathbb{R}^d$ to their $d$-dimensional targets via a linear layer, requiring each mapped input to score its own target above all other outputs. Since all inputs compete for the same outputs, these constraints are strongly correlated, which makes the storage capacity $p_c$ --- the largest number of associations that can be stored --- hard to characterise. We address this in three steps. First, we introduce a decoupled model, in which each input competes with its own independent set of outputs, and provide numerical and analytical evidence that it is equivalent to the original model, both in capacity and in the structure of the learnt solution. Second, using tools from statistical physics, we show that the decoupled model has capacity $p_c \log p_c / d^2 = 1/2$, extend this computation to linear two-layer networks, and go beyond the capacity by predicting the spectrum of the learnt weights and the storage mechanism. The latter explains how the optimal solution outperforms a naïve Hebbian rule: rather than boosting input--output alignments with broad fluctuations, it raises the correct scores just above the extreme-value threshold set by the competing outputs. Third, we prove the decoupled model's capacity thresholds rigorously, including in the non-convex two-layer case; combined with a recent proof for the original model, this shows that both models have the same capacity for full-rank weights. These findings give a sharp characterisation of factual storage in linear networks, and suggest decoupling as a general tool for memorisation and learning problems in which each input must be matched to one of an extensive number of outputs.
In this paper we introduce token time continuous diffusion (TTCD), a new diffusion language model which (a) operates in continuous space, deterministically mapping Gaussian noise to a final token canvas with no further sampling, and crucially (b) incorporates a new notion of per-token times, with some tokens proceeding from noise to token at a faster rate than others. Continuous space modeling helps TTCD avoid the parallel sampling of multiple tokens, which is a key source of inaccuracy at high speedups for models that iterate purely in discrete space. The notion of per-token times helps TTCD to better model conditional generation, allows for more sure tokens to proceed at a faster rate, and allows for differentiated inter-token influences during refinement. TTCD outperforms discrete models at high speedups. We train a 160M parameter TTCD model on OpenWebText, and then self-distill it; we find that at high speedups we are comparable in unconditional generation quality, and outperform in conditional generation, several existing models of similar size trained, on the same data, and self-distilled. We achieve similar gains in Sudoku solving as well.