Search arXiv⌕ Search

arXiv subjects

Search papers

Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.

At least 865 records · Page 48Linked to original sources
arXiv2026

Policy-Verified Self-Training: Learning Agent Compliance from Verifiable Policy Rewards

Runtime policy verifiers can prevent agents from executing disallowed actions, but their decisions are typically used only for enforcement and discarded afterward. We ask whether these decisions can instead become verifiable policy rewards for training the agents they guard. We introduce Policy-Verified Self-Training (PVST), a post-training paradigm in which an agent generates its own interaction trajectories and an executable policy verifier, here a deterministic gate compiled from a written policy, evaluates whether each action complies. Compliant trajectories are reinforced, and violations become negatives for preference optimization; beyond the gate and the benchmark's task-success signal, PVST needs no stronger teacher model, no human preference labels and no attack-specific supervision, although rollouts are collected in the benchmark's attacked environment. On held-out AgentDojo tasks under prompt injection, policy-verified preference training of Qwen3-8B reduces violating trajectories from 37.2% to 10.3 $\pm$ 0.9% (three seeds) and successful attacks from 12.4% to 0.5% at equal or higher pooled task utility, although utility falls on travel, compared with 22.0% when negatives are selected by an attack oracle and 31.9 $\pm$ 0.9% with negatives that contain no violation. The learned behavior holds under an attack template never used in training (37.7% to 11.0%) and on a suite excluded from training (34.8% to 16.7%), and the retained verifier's intervention rate falls from 39.2% to 8.8% (seed 0). Our analysis identifies two conditions under which the verdict is most informative: success must not already imply compliance, and guarded actions must appear in both permitted and forbidden contexts.

cs.LG↗
arXiv2026

Evaluating Compressed Memory: Dependence, Utility, and Content Benefit

Does a language model answer better because its compressed memory contains useful document information? Replacing memory vectors with zeros can lower a trained model's score without showing that it outperforms a model trained without memory. We combine zeroing with separately trained models that cannot access stored text or receive the same constant input for every document. Across nine systems, some models rely on memory yet perform worse than the trained alternative. For example, in experiments that predict states in generated stories, zeroing memory worsens a model's prediction loss, yet a model trained without memory achieves lower loss than the original model with intact memory. On TriviaQA, giving xRAG memory from another question's passage raises F1, which measures word overlap with accepted answers, from 19.38 without a document to 24.78. Its effect on how often outputs contain an accepted answer is statistically uncertain. The question's retrieved passage scores higher still (29.88 F1). Thus a gain over no document alone does not establish that the gain requires the question's passage. Evaluations must therefore specify the comparison model, absolute performance, and generation and scoring conditions.

cs.CL↗
arXiv2026

A Critical Review of Grid Integration Challenges and Flexibility Opportunities for Large-Scale Data Centers

The rapid proliferation of cloud computing and artificial intelligence workloads is fundamentally reshaping the magnitude, spatial concentration, and electrical characteristics of data center power demand. Data centers differ from conventional commercial loads because they combine high rack densities, converter-dominated interfaces, thermally coupled cooling, uninterruptible power supplies, programmable workloads, and increasingly complex behind-the-meter generation and storage within a single facility. This review examines their grid integration from utility and sustainable-energy perspectives across resource planning, large-load interconnection, distribution feeders, market operation, real-time dynamics, power quality, and asset management. Particular attention is given to distribution-level issues that receive limited treatment in broader energy reviews: bidirectional operation on renewable-rich feeders, harmonic propagation and resonance, ride-through and reconnection, transformer derating, cable ampacity, neutral-conductor heating, protection, and measurement. Static, composite, physics-based dynamic, calibrated, impedance, electromagnetic transient, and hybrid load models are compared according to study purpose and data requirements, and utility coordination platforms, uninterruptible power supply and battery control, workload flexibility, renewable co-location, cogeneration, fuel cells, thermal storage, and clean firm supply are assessed. The central finding is that requested peak demand and annual energy matching alone cannot establish reliable or sustainable integration. Credible assessment must consider hourly and locational supply, network deliverability, marginal emissions, lifecycle and water impacts, backup obligations, rebound, asset health, and verified operating capability.

eess.SY↗
arXiv2026

One-Step Prediction Fidelity Is a Poor Proxy for Planning Performance in JEPA World Models

Latent world models are routinely selected, checkpointed and ablated on one-step prediction loss, on the assumption that a model which predicts the next latent better is a better model. In model-based RL this assumption is known to be unreliable; we ask whether it also fails for latent world models, where the training target is the model's own embedding and the loss is not expressed in interpretable units. We introduce a frozen-decoder protocol reporting one-step error in physical units, decomposed into the part already present in the encoder's representation and the part the predictor adds. Applying it to a recent end-to-end JEPA world model across three continuous-control environments, we move the metric by fine-tuning and measure goal-conditioned planning on the same checkpoints, with three training seeds and paired planning seeds throughout. Within one environment, a per-epoch ladder shows the relationship saturating after a single epoch: the first epoch improves fidelity by 32% and planning by 4.4 pp, but across the fifteen checkpoints that follow -- a further 15% fidelity improvement, monotone in every training seed -- the correlation between one-step error and planning success is r = -0.010. This is not a task-ceiling artifact: stratifying by a model-independent measure of training-data support, planning in the lowest-support stratum starts at 80%, leaving ample headroom, and is nonetheless flat from epoch 1 onward. Across environments the exchange rate is equally unstable: a 93% fidelity reduction on TwoRoom yields -0.7 pp of planning, while a 40% reduction on PushT yields +4.9 pp [+1.1, +8.9]. All results replicate under a decoder-free measure, latent drift. We also show that a matched continued-training control absorbs the entire fidelity gain we had attributed to an auxiliary objective.

cs.LG↗
arXiv2026

DeLoop: Repeated Actions as Self-Supervision for Long-Horizon Agents

Language-model agents that run long tasks often redo work they have already done: they re-issue a tool call they made earlier and get back exactly the same observation, a step that adds no new information. We identify this failure as a loop. Loops appear natively in Qwen, Llama, and Gemini agents and become far more common once the agent's history is compressed to fit its context window. To reduce these loops, we propose DeLoop, a self-supervised training method. DeLoop builds on a simple observation: the agent's own rollout can provide the supervision needed to learn from its loops. We use a looping action as a negative example and the first later non-looping action in the same rollout as a positive example. DeLoop learns from these pairs through direct preference optimization and does not require task rewards, gold answers, or LLM judges. On CodeGym under compression, DeLoop reduces the share of Qwen3-8B tasks containing loops from 57.3% to 6.1%. It also uses about one third as many steps per task, with similar task success. On Qwen3-32B, DeLoop reduces loops and steps while raising task success from 37.7% to 44.8%.

cs.AI↗
arXiv2026

Emission spectroscopy of KELT-2A b with Gemini-North/IGRINS-2: hints of a stellar C/O ratio and sub-stellar metallicity

We present the results of high-resolution emission spectroscopy of the hot Jupiter KELT-2A~b using IGRINS-2 on the Gemini-North telescope, covering the $H$- and $K$-bands at a spectral resolution of $R \sim 45{,}000$. Observations were carried out over five nights, covering both pre- and post-eclipse orbital phases. Stellar and telluric contaminations are removed via singular value decomposition. We detect $\rm H_2O$ at a signal-to-noise ratio of {6.57}, confirming the previous result at a higher significance. While $\rm ^{12}CO$ is identified in the log-likelihood map near the expected planetary velocity, a definitive detection of $\rm ^{12}CO$ remains elusive given the current data quality. We perform atmospheric retrieval both with IGRINS-2 data alone and jointly with archival \textit{Spitzer}/IRAC photometric data points. The retrieved temperature--pressure profile is non-inverted, consistent with prior observations. {From the joint fit we constrain the water abundance to} $\rm log(H_2O) = {-3.78^{+0.27}_{-0.20}}$ {with the free chemistry model}. {The chemical equilibrium model yields} $\rm C/O={ 0.19^{+0.12}_{-0.07}}$, $\rm [C/H]=-0.81^{+0.41}_{-0.34}$, and $\rm [O/H] = -0.37^{+0.27}_{-0.17}${: the C/O ratio agrees with the stellar value, while the carbon and oxygen abundances are moderately depleted relative to the host star, though still consistent at the $\sim2σ$ level.} If confirmed, {this combination of a stellar C/O ratio with a sub-stellar metallicity is difficult to reconcile with gas accretion between the $\rm{H_2O}$ and CO snowlines, which predicts a super-stellar C/O. It instead points to the accretion of metal-depleted gas outside the CO snowline with only limited planetesimal enrichment, and hence to a disk-free, high-eccentricity migration history.}

astro-ph.EP↗
arXiv2026

Robust Explanatory Analysis for Structured Optimization Networks (REASON) in Learning

We seek a hierarchical resolution of residual error under a composite learning objective as a principled mechanism for identifying and exploiting additional information, particularly under distribution shift and perturbation. Inspired by the System 1/System 2 paradigm, we formulate this principle as a two-loop mechanism serving a common inference goal. A fast, System 1-like loop resolves the residual addressable by the available evidence; the remaining epistemic residual triggers a medium speed, System 2-like loop that actively acquires additional evidence and measures its informational value through residual reduction. Over nested solution spaces, the medium-loop optimum is provably no worse than the fast-loop optimum. Experiments show that the resulting two-loop model consistently improves upon the fast loop at approximately half the computation of the medium loop, including under unseen corruptions, and, when provided a cleaner observation, outperforms staged classifiers and learning-to-defer methods at equal computational cost.

cs.LG↗
arXiv2026

Simulations of Dust Content and Attenuation in $z=5-15$ Galaxies

Correctly modeling the dust attenuation law in galaxies is essential to properly inferring their growth and physical properties. With the rapid discovery of $z>6$ galaxies by JWST and lack of constraints on the properties of dust at these redshifts, we must turn to theoretical models to predict the properties of dust at these redshifts. In this paper, we develop a theoretical foundation for the nature of high-redshift attenuation laws. We analyze $107$ cosmological zoom-in simulations of galaxies run with the SMUGGLE galaxy formation model. SMUGGLE includes an explicitly modeled multiphase interstellar medium and a state-of-the-art dust model with a self-consistently evolving dust grain size distribution. We forward model these simulations with the radiative transfer package POWDERDAY to produce synthetic spectra and attenuation curves for a set of $z=5-15$, $M_*\sim10^{6.5}-10^{10}\; \mathrm{M}_{\odot}$ galaxies. We find evidence for flatter attenuation curves than the commonly used locally-calibrated literature attenuation laws assumed when modeling high-redshift galaxies. The vast majority of our attenuation curves also lack the 2175 Angstrom absorption feature. Our attenuation laws do not evolve strongly over $z=15-5$, which we attribute to relatively constant fractions of small grains up to $z=4$ in our simulations. We present a set of simulation-informed priors on dust attenuation law parameterization that could be implemented in SED fitting codes. There is an anti-correlation between galaxy mass and attenuation curve slope and positive correlation between galaxy mass and $A_V$, which we attribute to dust growth becoming efficient in our massive galaxies with $\log{M_*/\mathrm{M}_{\odot}}\gtrsim9$ and with ISM conditions above a critical metallicity threshold.

astro-ph.GA↗
arXiv2026

Learning What to Keep: Transfer-Aware Personal Skill Learning from User Feedback

User feedback enables LLMs to refine reusable personalized skills, but often mixes persistent preferences with request-specific corrections. Indiscriminately retaining both risks negative transfer, motivating selective retention at the level of individual requirements. We frame this challenge as selective transfer over atomic rules and propose SkillGrain, a framework for learning reusable personalized skills from user feedback. SkillGrain decomposes feedback into independently actionable rules, predicts their transferability, and retains only those suitable for future requests. To learn these decisions, we combine rubric-guided skill rewards with fine-grained credit assignment that separately evaluates rule content and retention decisions. We further jointly optimize skill induction and skill-conditioned generation through reinforcement learning, ensuring that learned skills support downstream personalization. We construct a benchmark spanning five personalized writing tasks to evaluate cross-request generalization. Experiments across three base models in both in-domain and out-of-domain settings show that SkillGrain achieves the best overall average performance among competing methods.

cs.CL↗
arXiv2026

EvoWeave: Composing Reusable Agent Experience through Operation Level Graphs

Different tasks often require interactive agents to perform similar operations at intermediate steps. Retrieving an entire past workflow can introduce irrelevant steps, whereas retrieving isolated operations can discard the dependencies needed to reuse them. We introduce EvoWeave, a framework for experience memory that represents abstract states as nodes and reusable operations as directed edges with input and output contracts, provenance, and capability references. A planner decomposes the current task into a dependency graph; operation retrieval and bounded directed search then assemble relevant experience fragments. An execution agent can inspect these fragments and their referenced capabilities, while reflection separates concrete execution evidence from reusable abstractions. We study the framework across four benchmarks, five evaluation settings, and three language model backbones. Across backbones, AppWorld Normal and Challenge task goal completion average 51.91 and 27.99, respectively, exceeding the baseline without memory by 6.99 and 4.99 percentage points. An AppWorld ablation associates directed connections with an additional improvement of 2.08 points over independent operation cards. The results motivate composition of reusable operations while exposing tradeoffs in retrieval overhead and conditions for valid reuse.

cs.AI↗
arXiv2026

ixi-sLM: On-Device sLM for AI Call Agent Service

Running an AI call agent on-device makes it private, network-independent, and free of per-query cloud cost, but shipping one as a real-world service imposes constraints a cloud deployment never faces: ~1GiB memory budget, in-call latency and thermal limits, static-integer NPU execution, Korean STT input far from sLM pretraining, and cloud-level quality expectations. We bring ixi-sLM, a 1.2B Korean sLM, to mobile NPUs through ON-DEVICE MODELING: domain-adaptive modeling trims the vocabulary by call frequency to shrink the embedding and LM head that take nearly half the model under INT4 transformer weights and refits the tokenizer to call speech (18% fewer tokens on call traffic), while task-aware quantization calibrates on real calls and assigns per-module mixed precision by activation sensitivity. A single domain-adapted model runs ixi-O's two production services, AI Call Summary and AI Call Answering, at 94% of a state-of-the-art cloud LLM's task-weighted quality; our on-device modeling carries that quality onto the NPU with only a marginal drop in overall score, in an 803MiB model (28% of FP16) at 23% of the CPU's per-call energy. The same recipe deploys across Qualcomm and Apple NPUs and serves several million ixi-O subscribers in production.

cs.AI↗
arXiv2026

Phase Stability and Anisotropic Superconductivity in Ternary M2B4C2 (M=Ca, Mg) Monolayer

In this work, we investigate the phase stability, electronic structure, and superconducting properties of the two-dimensional ternary borocarbides MB$_4$C$_2$ (M = Ca, Mg) using first-principles calculations and fully anisotropic Migdal-Eliashberg theory. Ab initio molecular dynamics, phonon dispersion, and elastic constant calculations confirm that both monolayers are dynamically, thermally, and mechanically stable. Electronic structure analysis reveals an intrinsic metallic nature with a multiband Fermi surface, where states near the $Γ$ point are dominated by in-plane B-$p_x$, $p_y$ orbitals, while zone-boundary states are primarily derived from out-of-plane B-$p_z$ orbitals. The electron-phonon coupling is determined to be $λ= 0.64$ for both materials, driven by the rigid covalent B-B and B-C honeycomb networks. By solving the anisotropic Eliashberg equations, we predict that CaB$_4$C$_2$ and MgB$_4$C$_2$ are nodeless two-gap superconductors with critical temperatures ($T_c$) of 33.4 K and 33.8 K, respectively, at a Coulomb pseudopotential of $μ= 0.10$. Furthermore, the superconducting state remains robust under increased Coulomb repulsion ($μ= 0.13$), retaining elevated $T_c$ values of 31.2 K and 30.9 K, respectively. These findings establish MB$_4$C$_2$ monolayers as promising candidates for nanoscale superconducting applications and enrich the family of two-dimensional multigap superconductors.

cond-mat.supr-con↗
arXiv2026

MMA-SafetyBench: A Benchmark for Multimodal Agent Safety Evaluation

Multimodal agents increasingly rely on visual interfaces to perceive state and select actions, creating an attack surface not captured by text-only safety evaluations. We study Visual Semantic Hijacking, in which deceptive but semantically meaningful interface elements, such as forged system alerts, cause an agent to depart from its authorized task. We propose a trajectory-guided procedure that identifies candidate injection points and inserts reproducible visual payloads into the observation stream. Using this procedure, we construct MMA-SafetyBench, which contains 1,083 adversarial trajectories across web automation, desktop GUI control, mobile navigation, document understanding, and video processing. Across nine evaluated models, Attack Success Rate (ASR) reaches 87.96%, while Semantic Compromise Rate (SCR) reaches 98.09%. A 1--3--5 scoring scheme distinguishes target-matching outputs, semantic disruption without a target match, and successful preservation of the benign task. Controls covering all nine models and five domains show that attacks persist under neutral prompts. A safety-priority prompt lowers ASR without eliminating the failure mode.

cs.CR↗
arXiv2026

RA$^2$: Retain-Anchored Attraction Preserves Forget-Adjacent Utility in LLM Unlearning

Large Language Model (LLM) unlearning is often evaluated primarily by target suppression and clean retain utility. These metrics can miss a local failure mode: a model may perform well on standard retain prompts, yet degrade on benign reasoning when prompts lie near the forgotten domain. This phenomenon is termed forget-adjacent utility loss. A fundamental question is: where should forget representations go after they leave the removed behavior? Successful unlearning depends not only on whether forget-conditioned representations depart from the harmful computation, but also on the direction in which they are redirected afterward. We propose RA$^2$ (Retain-Anchored Attraction), a latent-space unlearning method that redirects forget-token states toward nearby retain-supported representations while preserving higher-layer behavior on retain data. On WMDP and MUSE, RA$^2$ yields higher forget-adjacent utility than strong unlearning baselines at similar broad-utility levels. These results show that the migration direction of forget representations materially affects benign behavior near the forget boundary. Our code is available at https://github.com/ZacLing/retain_anchored_attraction.

cs.LG↗
arXiv2026

STRATA: Self-Maintaining Structured Experience Graphs for Self-Evolving Language Agents

Large language model (LLM) agents can self-evolve at test time by accumulating experience: retrieving relevant memories before acting and writing new lessons back afterwards. Yet most are single-level: trajectories, distilled workflows and reasoning items all land in one undifferentiated pool indexed by embedding similarity, with no typed abstraction hierarchy. They never forget or deduplicate, and are prone to negative transfer, where a superficially similar but directionally opposite experience is recalled and its concrete steps are copied verbatim, actively harming the agent. We introduce STRATA, a structured, outcome-grounded, self-maintaining memory that organizes experience into a three-layer graph of atomic skills and code snippets, task workflows and generalized meta-frameworks, connected by typed edges. STRATA maintains itself through five mechanisms sharing one similarity geometry: a multi-dimensional similarity that scores retrieval by operation and target; a typed reflector crystallizing procedures from successes and pitfalls from failures; community detection that induces meta-frameworks; pressure-triggered garbage collection with articulation-point protection; and contrast edges that flag confusable, direction-opposite experiences. Across four interactive benchmarks (AppWorld, WebArena, ALFWorld, ScienceWorld), two frozen backbones and seven baselines, STRATA attains the best average score while simultaneously using the lowest latency and fewest environment steps of any memory-augmented method, and 44% fewer lines of injected memory than the most context-hungry baseline. We isolate two failure modes of prior memories: single-level memory degrades sharply as the task set grows, and single-level concrete memories suffer catastrophic negative transfer, which contrast edges repair, lifting ScienceWorld from 78.7 to 96.7 within the same system.

cs.AI↗
arXiv2026

The Sum of All Summaries: PROPALIGN for Graded Proposition Salience Evaluation

Identification and ranking of salient propositions in documents is indirectly involved in many AI tasks, but few papers define proposition salience scores clearly, make diverse data available, or evaluate models directly on the task. This paper proposes PROPALIGN, a multiple-summary-based approach for propositional salience score estimation, and releases PROPALIGNBANK, an accompanying dataset for the method with 138K alignments of over 35K propositions. We compare several approaches to salient proposition extraction in English across a broad range of spoken and written genres, and demonstrate that naively asking models for judgments leads to inconsistencies. By contrast, graded, summary-based metrics are more stable. We further observe that while single models perform worse than humans, an ensemble of models incorporating linguistic features approaches manual annotation quality ceilings. In addition, analyses of the model outputs suggest insights into how models can extract salient information with and without the use of summaries.

cs.CL↗
arXiv2026

ELARSpeech: A 30K-Hour Dataset Across 675 Underrepresented Languages for Long-Tail Multilingual Speech Technologies

Massively multilingual speech modeling has made remarkable progress in recent years, yet existing training resources and benchmarks still underrepresent much of the linguistic long tail. To address this gap, we introduce ELARSpeech, a large-scale multilingual speech resource comprising 30,202 hours of speech across 675 underrepresented languages, curated from the Endangered Languages Archive. Combining human-annotated and unlabeled recordings, the corpus spans diverse recording conditions, domains, and language families. Beyond corpus construction, our experiments show how ELARSpeech supports extending ASR and TTS capabilities to diverse underrepresented languages, with further benefits from pseudo-labeling unlabeled data. At the same time, our speech representation analysis reveals remaining limitations in transferring existing multilingual models to the linguistic long tail. Together, ELARSpeech provides both a resource and a testbed for extending multilingual speech technology further into the linguistic long tail. We publicly release the dataset, all trained ASR and TTS models, and the complete data construction pipeline to facilitate research on long-tail multilingual speech technologies.

cs.CL↗
arXiv2026

Physical Processes Driving Year-to-Year Surface Temperature Variability Changes in a Warming Climate

Under anthropogenic greenhouse forcing, climate models project a decrease in year-to-year temperature variability in the mid-to-high latitudes and an increase in the tropics and subtropics. Here, we examine year-to-year temperature variability in the CESM Large Ensemble in concert with an idealized stochastic moist energy balance model (SMEBM) with effective surface heat capacity and local feedback parameter specified based on CESM. We find that the zonal-mean CESM variability change pattern largely persists after filtering out the impact of El Niño. The SMEBM largely captures this change pattern. In the SMEBM, changes in the mid-to-high latitudes are driven by changes in the effective heat capacity due to sea ice retreat, combined with changes in anomalous moist static energy convergence modulated by the polar-amplified time-mean temperature change. The CESM local feedback parameter evolves to be less stabilizing in the low latitudes, which causes tropical and subtropical variability to increase in the SMEBM.

physics.ao-ph↗