overfeed.news
Anuncio

Agendator

Reservas online gratis para tu barbería o salón.

Página de reservas, agenda y recordatorios por email. Sin cookies ni anuncios para tus clientes.

Crear mi página gratis

Tema

Seguridad y regulación

2441documentos

179últimos 7 días

arXiv AI Papers

Knowing When to Intervene: Causal Evaluation of AI Agents

Current evaluations of agentic AI systems primarily rely on benchmarks. However, the static nature of benchmarks makes them uninformative about the value of auditing interventions at deployment time. For reliable deployment, we often need to determine within an agent's trajectory whether or not to intervene, e.g., by verifying the agent's current output or steering it toward different behavior, using either human review or more costly models or time-consuming agentic supervision. The value of such interventions is fundamentally a causal inference question, asking whether the causal effect of such an intervention is beneficial for the final outcome. Estimating this causal effect is particularly challenging due to the dynamic nature of the problem: the impact of an intervention depends on the current context and environment, as well as the agent's future behavior. We develop the first methodologically grounded framework for evaluating the causal effects of randomized interventions within agent trajectories. By leveraging a relatively small dataset of agent trajectories collected under randomized interventions such as verification or screening, we can estimate the dynamic causal effect of a rich class of (contextual) intervention policies. The structural parameters that we estimate can also be used to construct safety leaderboards that capture the risk of various agentic actions, which can be used by practitioners for audit prioritization. Applying our methods to tau-bench Retail, ToolSandbox, and ImpossibleBench, we find that intervention effects vary substantially across agent actions and states. More broadly, our work suggests a rich research agenda at the intersection of causal inference and agentic evaluation, which we believe is essential for the reliable and safe deployment of AI agents in the real world.

en
arXiv AI Papers

Safety Is Π_1, Provability Is Σ_1: One Framework for the Hugging Face Breach and OpenAI's Mathematical Proofs

OpenAI's agents breached Hugging Face in July 2026, and the same company released more than seven hundred AI-generated manuscripts of mathematical proofs in October 2026. The first event is read as a failure in security, while the second is read as a success in mathematics. We read both through the hybrid framework of our paper (Kim and Kang, 2026, arXiv:2611.01914), in which coding, proving, and learning from data are derivations by one and the same deductive mechanism, distinguished by the kinds of premises from which they are derived. Under this framework, the AI systems behind the two events ran the same loop, in which a statistical proposer produces candidate exploits or proofs and a deductive verifier settles each one by a derivation. What separates the events is the logical form of the statement wanted by each of the two domains, security and mathematics. The safety of a program is a Π_1 statement, for it says that no exploit exists, which is not assured by any finite list of tests that have been passed by the program. By contrast, the provability of a sentence is a Σ_1 statement, for it says that a proof exists, which is assured by a single instance, namely a proof that has been checked by the deductive verifier. The theorem behind both is Turing's Undecidability Theorem: the halting set is Σ_1 but not decidable, so its complement is Π_1 and not Σ_1, which is to say that membership has a finite witness but non-membership has none. We prove that attaching a statistical proposer to a deductive verifier does not change this undecidability result, however good the proposer is. We then discuss what does assure safety and where the residual risk goes.

en
arXiv AI Papers

Intent-Aware Caching for Efficient LLM Serving

Modern LLM serving systems use prefix caching to accelerate multi-turn conversational workloads. They cache the key-value states of previous conversation prefixes and reuse them when subsequent requests share the same prefix, to avoid redundant prefill computation and reduce serving latency. However, achieving a high cache hit rate remains challenging because cache reuse depends on whether a conversation continues and how soon the next turn arrives. Our analysis of real workloads shows that user intent is a useful signal for this behavior, as different intents exhibit different continuation probabilities and inter-turn delays. Based on this key insight, we propose Intent-Aware Caching (IAC), an online eviction policy that classifies requests by intent, learns lightweight reuse statistics from recent serving logs and leverages these statistics to guide an oracle-inspired eviction strategy. Extensive evaluation on real workloads (WildChat and ShareChat) shows that IAC reaches a 34.6% cache hit rate. This substantially outperforms both the previous learning-based approach (18.8%) and traditional methods such as least recently used (LRU, 12.2%), while reducing average serving latency by up to 46.1%.

en
arXiv AI Papers

Ordered Policy Optimization

Proximal Policy Optimization (PPO) stabilizes policy updates by clipping the importance ratio between the new policy and the data-collecting policy. The standard rule uses one fixed clipping range for all samples. While this choice offers simplicity and effectiveness, it fails to distinguish statistically harmful extreme ratio values from large, yet tolerable changes. We study how to adapt the clipping radius to the reliability of the induced importance ratio distribution. Our case study demonstrates that reliable updates require controlling these extreme ratios, rather than merely verifying that individual ratios fall within a fixed interval. Motivated by this observation, we propose Ordered Policy Optimization (OPO). In each minibatch, OPO sorts the importance ratios, selects the extreme ratios that potentially have the largest effect on the update, and assigns them adaptive sample-wise radii instead of using the fixed clipping radius. These radii are derived from a prescribed tail profile, which specifies desired target values for the selected extreme ratios. The resulting objective remains first-order and requires only minibatch sorting and a modified policy loss. Experiments conducted on MuJoCo and Atari benchmarks demonstrate that this simple modification achieves strong performance relative to established baselines.

en
arXiv AI Papers

Composable Causality: A Toolkit for Systematic Time-Series Causal Discovery and Treatment-Effect Benchmarking

Real-world multivariate time series rarely come with verified ground-truth causal structure, so research on time-series causal discovery and treatment-effect estimation depends on synthetic data. Existing tools are either fixed datasets that cannot be modified after release or script-based generators in which each script fixes one combination of mechanism, confounding, and missingness. Neither supports controlled experiments that vary a single property, such as sparsity, noise distribution, or missingness mechanism, while holding the others fixed. Counterfactual evaluation has the same gap, since existing simulators are built per paper with hardcoded mechanisms or produce observational data only. We introduce DTCD (Datasets and Toolkit for Causal Discovery), a data generation library that produces multivariate time series with ground-truth contemporaneous and lagged adjacency matrices. Topology, mechanism, noise, observational mask, and intervention policy are independent components that compose through configuration alone. Paired counterfactual twins, produced by caching the stochastic realization during the factual rollout and re-injecting it during the interventional rollout, recover individual treatment effects to numerical precision. A reference benchmark of eight discovery algorithms on eighty-one configurations and four treatment-effect estimators on eighteen configurations exposes failure modes that aggregate metrics hide. These include the inability of standard discovery methods to handle missing values, regardless of whether the missingness mechanism is informative, and a marginal-variance shortcut (varsortability) that we measure and remove by standardization. Code and data: https://github.com/hferdous/Composable_Causality

en
arXiv AI Papers

HomeLoop: A Simulation Framework and Benchmark for Robot-IoT Coordination in Everyday Domestic Assistance and Safety

Recent progress in embodied robotics is bringing robots closer to deployment in homes, as vision-language models (VLMs) are demonstrating significant advances in reasoning, making VLM-driven robots promising home assistant agents. These agents will often be introduced into homes that already have cameras, Internet of Things (IoT) devices, and residents with wearables. To fulfill their potential, they need to integrate with existing smart infrastructure and reason through ambiguous scenarios: a smoke alarm may come from a pan about to ignite, a pan already on fire, a resident cooking or a faulty sensor, and each calls for a different response. To that end, we introduce HomeLoop, a simulation framework and benchmark built on Isaac Sim that combines robot bodies, networked devices and wearables, simulated residents and visitors, faulty sensors, and hazards that evolve over time. Our benchmark is composed of 20 tasks across five families grounded in real-world household hazards and needs, and designed so that the same alert can have different causes. A runtime monitor scores each episode with temporal-logic checks on the final outcome and the actions along the way, and every action carries a predefined cost. We evaluate six open-weight and proprietary VLMs. The best achieves a family-balanced score of 61.9%, and across models we see a pattern of over-reaction: failed episodes involving prohibited interventions outnumber those involving missed required interventions by more than two to one, with models often shutting off the stove before investigating the alarm's cause. HomeLoop is an extensible foundation intended for use in research on robot-IoT coordination, long-term memory, proactive assistance, and privacy-safety trade-offs. HomeLoop will be released with 2,100 home layouts that can be automatically furnished in diverse configurations, five robot bodies and 20 networked device types.

en
arXiv AI Papers

Task-Contribution Dynamics of Adam in Physics-Informed Neural Networks: Slow-Manifold Analysis and Gradient Modulation

Physics-informed neural networks (PINNs) often suffer from imbalanced optimization of competing loss components, particularly when approximating solutions with high-frequency or multiscale features. Although existing balancing strategies have proven effective in alleviating this difficulty, the mechanisms governing the evolution of individual task contributions to Adam's parameter updates remain insufficiently understood. We introduce task-wise indicators to characterize this evolution and identify three distinct training regimes: smooth variation, intermittent spikes, and persistent oscillations. Focusing on the early smooth regime, we combine discrete-time analysis with a continuous-time singular perturbation framework under the full-batch setting and the parameter regime (1-β_2)/(1-β_1)1. A first-order approximation of the locally attracting slow manifold reveals two factors governing the step-to-step redistribution of task contributions: curvature-related alignment disparities and differences in the coordinatewise distributions of preconditioned task gradients. Both factors involve Adam's shared second-moment estimate. Inspired by AdaTask, we propose Gradient-Modulated Adam (GMAdam), which modulates incoming task gradients using task-specific second-moment estimates. Numerical experiments on high-frequency Poisson and wave equations, the Allen-Cahn equation, and two-dimensional lid-driven cavity flow (Re=100,400) demonstrate consistent improvements in predictive accuracy over AdaTask at comparable computational cost.

en
arXiv AI Papers

Geometry-Aware Score-Repellent Monte Carlo

The recently proposed Score-Repellent Monte Carlo (SRMC) method improves MCMC sampling by tilting the target density π in {R}^d with the factor (-α\,θ_n^s(x)), where the score function s(x) = _x π(x), and θ_n is the score average from past samples. However, a single scalar repellence strength α0 cannot handle both the steep and the flat directions in anisotropic energy landscapes at once. We introduce Geometry-Aware SRMC (GA-SRMC), which generalizes the Euclidean alignment to θ^M s(x) for any symmetric positive-definite matrix M. A stochastic-approximation central limit theorem under a trace constraint on M identifies the trace-normalized inverse Fisher M_{opt}S^{-1} as the unique minimizer of the worst-direction covariance bound, elevating the Fisher information S=Cov_π(s,s) from a proposal-side preconditioner (e.g., FisherMALA, natural gradient) to the bound-optimal repulsion-side matrix, with a single online estimate S_n serving both roles at no additional computational cost. The full-Fisher online metric M_n converges almost surely to M_{opt} and satisfies a CLT, providing a theoretical guarantee for the online implementation of the GA-SRMC framework. On an 8-target simulation spanning the condition number κ[1.3,100] and dimension d[2,785], GA-SRMC dominates SRMC in ESS by {4.3} on anisotropic Gaussian (κ=100) and {6.8} on MNIST under the MALA baseline, by {16--32\%} on three well-conditioned logistic-regression posteriors, and by up to {+48\%} under FisherMALA; gains peak on the most anisotropic targets, and posterior-mean MSE is lower than with SRMC in all 16 full-Fisher implementations.

en
arXiv AI Papers

Learning to Restart: Differentiable Restart Policies for Optimization Dynamics

Adaptive restart mechanisms, which reset the momentum or other auxiliary state of an iterative optimizer, play a central role in the empirical performance of accelerated first-order methods, yet existing restart rules remain almost entirely hand-designed. The obstruction to learning restart policies is that, unlike the typical Learning to Optimize (L2O) setting, the discrete restart events make the resulting trajectory non-differentiable in the algorithm parameters and thus block gradient-based optimization of the restart policy. We resolve this obstruction by developing a differentiable framework for restart policies through their continuous-time limit. We introduce surrogate gradients for the restart events and optimization trajectories that are efficient to compute and provably converge to their continuous-time counterparts. We validate the effectiveness of our surrogate gradients in L2O settings, training restart policies that improve upon standard handcrafted adaptive restart criteria on quadratic and linear programs, and that transfer zero-shot from synthetic quadratics to logistic regression on real datasets.

en
arXiv AI Papers

Adaptive Scheduling Pipeline for Multi-Instance Asynchronous Reinforcement Learning

We present the Adaptive Scheduling Pipeline (ASP), an offline rollout scheduler for multi-instance asynchronous reinforcement learning (RL) of large language models (LLMs). ASP jointly reduces three sources of rollout underutilization - padding bubbles from intra-minibatch length heterogeneity, skewness bubbles from long-tail outputs, and sorting bubbles from fluctuations in training-queue occupancy induced by length-sorted dispatch - through a single offline preprocessing pass. This enables efficient multi-instance rollout that overcomes the early scaling plateau of single-instance inference while smoothing the arrival of generated samples at the trainer. We integrate the scheduler into an asynchronous architecture that decouples rollout from training while keeping policy lag observable and bounded. Complete ASP improves end-to-end throughput by up to 3.22 times and reduces iteration time by up to 76.8% relative to MindSpeed-RL, while reducing iteration time by up to 62.5% relative to verl. Within ASP, scheduling reduces mean iteration time by 10.9%. ASP also scales more robustly with additional generation resources than these baselines and shows no visible convergence degradation in the measured bounded-lag regime. ASP has since been used in an internal telecom-domain RL post-training workflow, reducing the reported training period from more than three months to about two months and producing a domain-enhanced model for knowledge answering and tool calling.

en
arXiv AI Papers

STRIDE: Learning Humanoid Locomotion in Minutes with Acceleration-Based Rewards

Learning locomotion directly on a humanoid robot is a promising direction to avoid the sim-to-real gap. However, locomotion rewards in Reinforcement Learning (RL) typically track the base velocity, which cannot be measured onboard without an external motion capture system or well-calibrated state estimator. This is why we introduce {}, which learns humanoid locomotion from minutes of samples with a reward that only requires the onboard Inertial Measurement Unit (IMU). Instead of tracking the base velocity, STRIDE rewards the robot for following a commanded acceleration profile. To learn efficiently from this reward, we use the off-policy RL algorithm XQC and combine a high gain setting for the joint position controllers with left-right symmetry augmentation. In simulation, STRIDE learns to walk on the Unitree G1 and Booster T1 humanoids within 14 minutes of samples. Compared to learning without these two components, it reduces the falls during training by 81%, while without them the robot learns only in over an hour.

en
arXiv AI Papers

Solver-Admitted Metamorphic Testing of Learned Network Optimization Surrogates

Equivalent encodings of an optimization problem can produce different learned predictions. Interpreting these differences requires accounting for numerical error and multiple optima. We construct metamorphic tests for optimization solver surrogates with explicit eligibility conditions, output alignment, and numerical thresholds. Reference-solver agreement and independent constraint checks determine which sampled pairs enter the audit. We study alternating-current optimal power flow (AC-OPF), the traveling salesman problem (TSP), the capacitated vehicle routing problem (CVRP), and maximum independent set (MIS). Across five grids, GridSFM-Open violates each of 13 relations on at least one grid, and 12 by more than twice the threshold. Randomly initialized copies of the grid models fail nearly the same relations, so the failures do not require training. Source patches and input canonicalization remove selected discrepancies, but no intervention establishes preserved accuracy and feasibility within the declared margin on every grid. In routing, Policy Optimization with Multiple Optima (POMO) has a single-pass rotation/reflection violation rate that rises from 12.7% to 89.5% between 20 and 100 cities. Graph-Based Diffusion Solvers for Combinatorial Optimization (DIFUSCO) violates the MIS disjoint-union relation on 35.6% of draws. Post hoc normalization within connected components removes these violations on the tested graphs. These findings support reporting metamorphic consistency alongside optimality, feasibility, and accuracy, and validating repair quality separately from consistency.

en
arXiv AI Papers

High-Dimensional Off-Policy Evaluation without Sample Splitting using an Onsager Correction

Reward models used for off-policy evaluation (OPE) are commonly selected using held-out data. When sample size and context dimension grow proportionally, the difference in test risk between models fitted on the full sample and on a fixed fraction of it need not vanish. Held-out evaluation can therefore consistently estimate the test risk of the model trained on the smaller sample, but because the relative ranking of candidate reward models can change after full-data refitting, held-out selection can be suboptimal. We develop an Onsager-corrected, target-policy-weighted criterion that estimates the full-data reward model's out-of-sample risk under the target policy using only its training observations. This enables early stopping along the full-data gradient-descent trajectory without sample splitting. For neural networks with fixed depth and hidden width under Gaussian contexts, single-index rewards, context-dependent multi-index target policies, and known context-independent logging, we prove uniform consistency of our criterion over a fixed finite set of gradient-descent iterates and a near-oracle guarantee for the selected reward model. The same correction yields a doubly robust estimator that consistently estimates the true policy value without sample splitting, even when reward-model prediction error does not vanish. Experiments show decreasing risk-estimation error as dimension grows at a fixed sample-to-dimension ratio, and improved reward-model selection relative to held-out baselines. Empirically, our correction also reduces the root mean squared error of doubly robust value estimates computed from the same data used to fit and select the reward models.

en