每天知道
你的领域
出了什么

Security Operations Centers (SOCs) process large volumes of security events, requiring analysts to accurately detect and assess ongoing cyberattacks under time pressure. Recent advances in Large Language Models (LLMs) suggest potential benefits for security operations, yet their practical suitability for real-world SOC workflows remains poorly understood. To address this gap, we conducted 25 semi-structured interviews with SOC practitioners who had prior experience with LLMs, complemented by interactive scenarios to anticipate challenges and identify opportunities for the responsible integration of LLM-based tools into SOC workflows. We identified 15 LLM use cases grouped into six functional categories. While LLMs are valued for automating repetitive, low-level tasks such as report automation, practitioners rate high-impact tasks such as incident analysis as not yet feasible, reporting limitations in technical depth, context awareness, and organization-specific knowledge. They locate these limitations less in the models than in the readiness of their SOCs and human factors driving over-reliance. Despite concerns, practitioners express a strong willingness to adopt LLMs, describing competitive pressure that leaves few alternatives. This work contributes an empirical, practitioner-driven analysis of LLM use across SOC roles and organizations and derives concrete design and integration requirements for human-centered, operationally safe LLM-assisted security operations.

Large language models (LLMs) can generate fluent reasoning traces that nevertheless lead to incorrect answers, making response-level uncertainty estimation important for abstention, human review, and adaptive compute allocation. Existing approaches generally fall into three categories: passive single-trace methods use token-level confidence signals, sampling-based methods compare multiple complete traces at higher generation cost, and active prefix-based methods probe partial traces to study answer stabilization or preference transitions. However, none actively re-elicits an answer from a completed reasoning trace to measure its consistency with and support for the original answer. To address this gap, we introduce Trace-Conditioned Answer Consistency (TrAC), a correctness-supervised uncertainty quantification framework that combines active and passive signals anchored to one completed reasoning trace. Its active component, Prefix-Conditioned Elicitation (PCE), re-elicits a short answer conditioned on the completed trace and represents both its consistency with the original answer and its token-level probabilistic support. Its passive component, Trace Uncertainty Profile (TUP), summarizes how token-level uncertainty evolves throughout the original generation without additional decoding. A lightweight head then integrates the two representations into a response-correctness score. Across five mathematical reasoning benchmarks and three LLM families, TrAC improves macro AUROC by 1.8% and reduces AURC by 3.4% relative to eight-sample self-consistency, while using one complete reasoning trace and a short cached answer probe. When eight samples are already available, augmenting sample consensus with re-elicitation further improves macro AUROC by 4.3% and reduces AURC by 8.3%, without additional full-trace generation.

Recent advancements of zero-shot Named Entity Recognition (NER) establish strong baselines by formulating sequence labeling into question answering where Large Language Models (LLMs) can be naturally adopted. However, existing LLM-based zero-shot NER methods suffer from the limitations of prompt and demonstration engineering. To address these issues with minimal human interventions, we introduce DE-NER, a dialogue elicitation framework which elicits the chatting ability of LLMs to fully extract the knowledge encoded in LLMs. Our experiments demonstrate that the proposed method outperform the competitive baselines in zero-shot settings across multiple benchmarks, with an average improvement of 3.75% F1 points. Codes are released in https://github.com/kkkenshi/DE-NER.

Diffusion large language models (dLLMs) have emerged as a promising alternative to autoregressive LLMs, offering efficient generation through block-wise progressive unmasking. However, their strong general-purpose performance does not necessarily translate into reliable mathematical reasoning, where correctness depends on preserving coherent numerical-symbolic reasoning trajectories. In this work, we analyze the decoding trajectories of LLaDA 2.0 and identify a recurring diffusion confidence trap: local token confidence can become misaligned with global reasoning correctness during progressive block decoding. Our analysis reveals two representative failure regimes: sampling-sensitive failures, where correct paths exist but are unstable, and sampling-consistent failures, where repeated sampling converges to repetitive high-confidence but incorrect continuations. Motivated by this observation, we propose Evolutionary Decoding, a training-free test-time scaling framework that views diffusion decoding as an evolutionary process over candidate reasoning states. The framework combines step-wise selection, which preserves useful numerical-symbolic signals and suppresses repetitive patterns, with block-wise mutation, which introduces structured alternatives to escape incorrect high-confidence basins. Experiments on multiple benchmarks show that Evolutionary Decoding improves LLaDA 2.0 over confidence-based decoding, leading to more reliable mathematical reasoning.

The rapid development of Large Language Models (LLMs) has opened new avenues for Automated Heuristic Design (AHD) for solving NP-hard combinatorial optimization problems (COPs). However, existing LLM-driven AHD methods are largely confined to rigid solver templates, relegating the search process to isolated module tuning. Transitioning to fully autonomous, system-level algorithm design is essential but fraught with low reliability of generated operators, extremely large search spaces, and ineffective credit assignment. To overcome these drawbacks, this paper proposes a Directed Graph-Guided Automated Algorithm Design framework, termed DGA$_2$D. It structures the open-ended program space as a directed graph, where each node represents a functional operator that can be instantiated using one of multiple candidate code implementations, while directed walks constitute complete algorithmic pipelines. A first-order path-dependent credit assignment mechanism is introduced to evaluate code variations strictly based on their topological context. Extensive experiments across 12 distinct COPs, ranging from complex scheduling to routing, demonstrate the consistent empirical advantages of DGA$_2$D. It reduces the average normalized gap by up to 10.96 percentage points compared to state-of-the-art LLM baselines.

Although large language models (LLMs) can produce fluent spatial reasoning traces, their intermediate relations may fail to support the final conclusion, making token-level confidence insufficient for final-answer reliability estimation. Existing formal verifiers provide stronger semantic evidence, but their applicability is partial: a parsed claim need not yield a definite semantic verdict. To address this issue, we introduce SymboUQ, a symbolic uncertainty quantification framework that estimates final-answer reliability from reasoning traces by distinguishing symbolizability, whether a claim can be represented in the verifier's formal language, from semantic determinacy, whether its execution yields an entailed or contradicted verdict rather than an unknown or not-evaluable outcome. SymboUQ comprises (i) a Layout Auditor that executes ordered spatial claims and extracts feasibility, conflict, and repair evidence; (ii) a label-free Determinacy Profile that characterizes effective executable coverage; and (iii) a Determinacy-Aware Reliability Composer that integrates constraint-based, representation-based, and decoding-based scores according to verifier applicability. Extensive experiments on five spatial reasoning benchmarks with four frozen LLM backbones show that SymboUQ achieves approximately an 8% relative improvement in AUROC and a 7% relative reduction in class-balanced Brier loss over the strongest baseline.

Large Language Model (LLM) agents rely on external tools to perform multistage tasks. Existing agent frameworks typically assume that tool calls are atomic and return binary success or failure signals. However, real-world systems exhibit non-atomic behaviors such as timeouts after dispatch, delayed visibility, and partial state updates. These mismatches lead to reliability issues including duplicate actions, task success, and unnecessary tool executions. A lightweight, verification-aware tool wrapper is introduced that augments tool calls with postcondition verification, verify-before-retry logic, and idempotency keys. The approach is evaluated in a controlled simulated environment with injected non-atomic failures across multiple task templates. The results demonstrate that the proposed method significantly reduces duplicate actions, while maintaining comparable task success rates. Overall, the findings suggest that strengthening tool interaction semantics is a promising direction for improving LLM agent reliability without requiring modifications to the underlying language model.

Computer-use agents (CUAs) operate real desktop and web interfaces through screenshots, mouse and keyboard actions, and stateful UI feedback, yet their failures remain difficult to diagnose and repair. Unlike text-only agents, CUA failures arise from coupled visual perception, spatial grounding, low-level interaction, task reasoning, and environment dynamics, making debugging a distinctive multimodal causal localization problem. We introduce CUADebug, a framework for diagnosing and repairing CUA failures. CUADebug includes a CUA-specific error taxonomy, CUAErrorBench, a human-annotated OSWorld failure benchmark, and CUADebugger, a tool-augmented debugger. Instead of prompting over the full trajectory once, CUADebugger actively inspects suspicious steps with paired before/after screenshots and action traces, then submits a structured diagnosis containing the root-cause step, error type, grounded evidence, and corrective strategy for re-execution. Human annotations over 204 failed trajectories show that task reasoning and control is the largest failure family (110/204), followed by perception (36), grounding/interaction (25), external/system (13), and an others category of 20 OSWorld infeasible-task cases. On the main Claude-agent split, CUADebugger improves joint subtype-and-step diagnosis from 11.2% to 19.6% with Gemini 2.5 Pro and improves consistently across debugger backbones. In single re-execution package evaluation, RCA-based conditions achieve higher task completion than history-only continuation (28.47% with machine RCA and 29.90% with our method, versus 13.89%); in continual re-execution, our method improves success from 12.2% to 25.86%, while human-oracle guidance reaches 29.21%. These results show that CUA root-cause diagnosis can provide actionable repair signals rather than merely post-hoc explanations.

Large language model (LLM) agents increasingly rely on external tools to complete complex real-world tasks. However, reliable tool-use planning remains challenging due to the limitations of implicit reasoning and the evolving nature of real-world execution environments. Existing tool-use agents typically rely on LLMs to infer tool compositions from textual descriptions, which can lead to inefficient exploration and unreliable execution in complex tasks. To address these challenges, we model tool relations at the schema level and construct a directed Tool--Schema Hypergraph, in which tools are represented as hyperedges from their required input-schema nodes to their output-schema nodes. Furthermore, we propose HyperAgent, a Tool--Schema Hypergraph-guided framework for dynamic planning and execution. Given a task, HyperAgent first extracts a task-relevant tool context graph and uses it to guide the construction of a schema-aware Task DAG. During execution, HyperAgent dynamically realizes each subtask by constructing a state-conditioned tool support graph through deficit-oriented expansion, which identifies unresolved requirements and retrieves supporting producer tools according to the current agent state. Experiments on AppWorld demonstrate that HyperAgent improves task completion performance while reducing redundant API calls, LLM interactions, and token consumption compared with existing agent baselines.

Recent radiology-adapted vision-language models have achieved strong performance on standard report generation benchmarks, yet their robustness and generalization remain constrained by imperfect alignment and correlation between visual and textual features. Existing methods connect image and text either implicitly through autoregressive report supervision or explicitly through contrastive learning. However, autoregressive supervision alone is insufficient to establish reliable image-text alignment, while contrastive learning can push apart unpaired reports that describe related pathologies simply because they are not paired with the same image. This is problematic in radiology, where different reports may share compatible pathology semantics rather than being true negatives. As a result, the learned representation may fail to organize images and reports around shared pathology concepts, causing the decoder to rely on pretrained language priors and generate clinically plausible reports that are not fully supported by radiographic evidence. To address this issue, we propose PALM, a pathology-aware alignment framework for radiology report generation. Instead of directly matching each image-report pair while separating all others, PALM aligns visual and textual features through shared pathology prototypes. These prototypes provide a clinically meaningful bridge between radiographic evidence and textual findings, allowing cases with similar pathology semantics to move toward common concepts without separating compatible cases. In addition, we introduce Masked Evidence Modeling to strengthen the image encoder sensitivity to local radiographic evidence by learning semantic changes caused by masked image regions. Experiments on MIMIC-CXR, IU X-Ray, and MIMIC-ABN show that PALM consistently improves both report generation and abnormality-focused robustness.

Deep reinforcement learning (RL) agents achieve strong performance by optimizing scalar reward functions. However, once deployed, the policies of these RL agents are often rigid and costly to adapt to new performance criteria. For instance, an agent trained to maximize expected cumulative reward may not accommodate previously unknown stakeholder preferences. Existing approaches to achieve fairness, a type of preference, in RL typically assume that such preferences are known a priori and require complete retraining of the policy under a fairness-oriented metric. Inspired by inference-time alignment in large language models, we investigate the problem of steering a pretrained RL policy toward welfare-based fairness objectives at inference time without updating the base policy's parameters. We formalize inference-time fairness alignment as a policy shaping problem and propose a multiplicative policy shaping framework that adjusts action probabilities using action-dependent welfare scores, thus requiring no modification to the base policy. Our framework is general and compatible with any deep RL agent. Through extensive experiments across multiple domains, we demonstrate that inference-time policy shaping substantially improves welfare-based fairness objectives while preserving core task performance.

Image restoration quality can be evaluated along three complementary facets: pixel-level fidelity, human perception, and downstream machine preference. However, existing lossy compression restoration methods optimize for at most one of these criteria: fidelity-oriented models often regress toward conditional means and produce over-smoothed outputs, while generative approaches hallucinate plausible but factually incorrect textures that degrade both ground-truth fidelity and downstream task accuracy. To navigate this three-way tradeoff, we propose FDIR, a two-stage architecture that decouples the conflicting demands through complementary inductive biases: Quality-Guided One-Step Flow Matching (QO-Flow) recovers global semantic structure in latent space via a single forward pass, while Flow-Conditioned Detail Refinement (FCDR) deterministically restores high-frequency textures and suppresses generative hallucinations in pixel space. Extensive experiments demonstrate that FDIR achieves superior fidelity, with a favorable perceptual-fidelity balance and competitive machine preference.

3D scene understanding requires reasoning about entity existence, spatial layout, and object relations, yet RGB images alone often provide insufficient 3D cues. Existing 3D-VLMs commonly rely on depth or 3D-position-aware inputs at inference time, introducing additional acquisition, reconstruction, or annotation costs that limit RGB-only deployment. We therefore study how training-time 3D evidence can be converted into spatial reasoning capabilities retained under RGB-only inference. We propose a privileged-evidence distillation framework that constructs a distillable teacher through a unified evidence interface and controlled residual injection, and transfers its knowledge to a deployable student receiving only RGB images and questions through logit and structured representation distillation. To avoid imitating teacher signals unsupported by RGB, we further introduce evidence-sensitivity-guided distillation, which uses corrupted evidence to identify highly evidence-dependent targets and down-weight their supervision. We also define a recoverability decomposition based on the matched baseline, teacher, and student, separating privileged gains into RGB-recoverable improvements and residual teacher advantages. Across four benchmarks, the teacher achieves the best result on 7 of 11 reported metrics among the compared methods. The RGB-only student outperforms its matched baseline on all 11 metrics, including gains of 10.4 ScanQA CIDEr and 19.1 Scan2Cap CIDEr@0.5, without additional inference-time inputs. These results validate the effectiveness of training-time privileged 3D evidence distillation for both teacher performance and deployable RGB-only spatial reasoning. Separately, our matched baseline-teacher-student analysis characterizes privileged-gain transfer across evidence types and spatial skills.

Knowledge distillation (KD) is a widely utilized technique for transferring knowledge from a large model (the teacher) to a smaller model (the student). Owing to its flexibility and broad applicability, KD has been extensively applied in the compression of server-side models to meet the Quality of Service (QoS) requirements of client users. Despite significant advancements, the performance of distillation is substantially compromised when a large disparity exists between the capabilities of the server and the requirements of the client. To alleviate this problem, we propose a novel distillation approach, named Progressive$^2$, which operates through the combination of a progressively stronger teacher and a progressively smaller student. On the side of the teacher, rather than involving all layers simultaneously, we progressively select additional layers for distillation following a raw-to-rich semantic progression, establishing a systematic learning curriculum. Furthermore, we design a teacher-side multi-feature fusion adapter for the teacher to improve training stability, which is theoretically supported by the framework of Lipschitz continuity. On the side of the student, rather than directly training a tiny model, we gradually reduce the size of the network to facilitate an iterative co-evolution with the teacher. Progressive$^2$ serves as a flexible framework; the progressive strategy of the teacher can be deployed independently to achieve an optimal balance between accuracy and training efficiency, while the joint integration of the teacher and the student yields further improvements in overall performance.

World models and generative simulators are emerging as interactive testing infrastructure for autonomous driving because they can react to the ego planner and produce counterfactual, rare, and safety-critical rollouts. This changes a test scenario from a fixed replayed trajectory into an interactive scenario family whose realized evolution depends on the planner under test. The unresolved question is therefore not only whether dangerous rollouts can be generated, but what valid closed-loop evidence is enough to support a specified testing intent and stopping decision. This paper formulates interactive world-model-style testing adequacy and introduces WM-Cov, a provider-agnostic evaluation layer that converts raw provider outputs into requested, realized, and valid evidence. WM-Cov reports adequacy through coverage growth, valid-failure discovery, failure-mode diversity, realism, artifact suppression, duplicate accounting, and valid-evidence precision. Studies on executed TeraSim/SUMO events, WM-like mixed trace pools, and a real DriveArena TrafficManager--WorldDreamer matrix show that dangerous-looking events can include valid ADS failures, duplicates, partial realizations, and artifacts. The DriveArena matrix evaluates two planners, two horizons, six prompt conditions, and 360 ego-route requests; 304 attempts become fully realized evidence and 56 remain partial. A disjoint 80-request route-slice check yields 74 fully realized and 6 partial attempts. The results support evaluating world-model-style testing by convergence of valid interactive evidence under budget, rather than by raw generated failures or prompt coverage alone.

Spatial atomic layer deposition (SALD) is a leading atmospheric-pressure, high-throughput route to industrial ALD, but design and control are limited by the cost of predicting surface coverage: high-fidelity CFD is far too slow for operating-window scans, while analytic models miss transport modulation such as the gas curtain. We present a physics-chemistry-informed neural network (PCINN), a hybrid surrogate with CFD-level accuracy at real-time speed: a query returns coverage in about 7 ms, roughly 5x10^4 times faster than a CFD solve, reaching a test R^2_log = 0.998 (leave-one-out R^2_raw = 0.974) from only 30 training cases spanning four orders of magnitude in coverage. The architecture is not a black box: a small network learns only the operating-condition to near-wall concentration closure, while the known surface kinetics is a hard-coded, trainable chemistry layer integrated along the substrate trajectory. This single-scalar bottleneck keeps it accurate under sparse data, interpretable and invertible. We add a full identifiability analysis (Fisher information, profile likelihood). The adsorption energy E_ads and desorption rate k_des are robustly identifiable; k_ads is not separately identifiable at a single temperature (only k_ads*c_wall is). Across four temperatures the prefactor nu and E_ads bind along a weakly identifiable degeneracy valley of slope 0.065 eV/decade, derived analytically as k_B T_eff ln(10) and turned into a reliability diagnostic: a seven-chemistry mismatch matrix shows it is invariant under any single-Arrhenius mismatch and shifts only when a second thermally activated process appears, so a slope departure flags unmodelled site heterogeneity. Data come from simulation with known ground truth inverted by the same kinetic form, so the study verifies pipeline self-consistency and the identifiability boundary, not real parameters.

The push for broader coverage in future cellular networks depends on reliable service, yet this is increasingly harder to do as we encounter more instances of extreme weather conditions. In extreme weather conditions, we have difficulty evaluating coverage due to limited access to channel measurements. In this paper, we generate channel state information (CSI) in low and moderate weather conditions to synthesize realistic MIMO CSI under adverse weather conditions. Our primary contributions are to (1) synthesize MIMO channel datasets incorporating three weather types, each with three intensity levels, representative of practical 5G/6G scenarios; (2) train a diffusion model conditioned on weather using channel samples obtained through conventional pilot-based estimation under low and moderate weather intensities, and subsequently use it to generate channel realizations for severe weather conditions; and (3) evaluate the downlink Bit Error Rate (BER) and Outage Probability measures using the generated channels. The results show that diffusion-based generative models provide a scalable, data-driven alternative for channel modeling in harsh environments and can generalize to severe weather conditions using only low- and moderate-intensity training data.

Agentic systems must repeatedly decide whether to answer directly, decompose a task, invoke a tool, execute code, delegate to a specialist, verify an intermediate result, or recover from failure. These meta-decisions affect not only task success but also operating cost and latency, yet they are often embedded inside an orchestration framework and evaluated only through aggregate task accuracy. We present MetaRoute-Bench, an open, inspectable framework for comparing meta-decision policies under a shared execution model. The initial benchmark contains 180 synthetic task profiles spanning data analysis, research, and document processing, eight routing policies, and 30 paired random seeds. Across 43,200 traces, a task-aware compositional policy achieves 79.4% success compared with 76.7% for a strong workload-specific static policy, 67.4% for one-shot task routing, and 52.9% for direct answering. Relative to the static policy, this is a 2.7 percentage-point improvement with paired 95% CI of plus or minus 2.0 points, at 4.7% higher mean cost and 6.4% higher latency. Ablations show the largest losses when route composition is restricted to one operation and when verification is removed. These results are generated by a seeded offline execution model rather than a live deployment; accordingly, the primary contribution is a reproducible evaluation method and an analysis of routing-policy tradeoffs, not evidence of production effectiveness. We release task generation, policies, traces, tests, and analysis artifacts to support live-system validation.

Compact web agents can reduce deployment cost, but training them poses challenges in both data collection and post-SFT reinforcement learning (RL). Successful trajectories are expensive to collect and often contain inefficient detours. After supervised fine-tuning (SFT), full trajectory corpora are dominated by routine states; moreover, when group-relative RL is applied to web actions, inadequately designed action-level rewards can yield weak or misleading relative updates, while groups rejected as unsuitable for such updates receive no fallback learning signal. We present RMSWeb, a three-part recipe for Qwen3-VL-Instruct at 8B and 32B. Reflection-conditioned retries increase collection yield and shorten successful trajectories; failure-mode mining concentrates offline RL on critical states exposed by the SFT policy; and Salvage-DS combines an action-semantic polarized reward, contrast-and-competence-gated dynamic sampling, and an action-only anchor for rejected groups. Policies trained with reflection-collected data use up to 19.7% fewer action steps on solved tasks. On WebVoyager, Online-Mind2Web, and WebTailBench, RMSWeb improves over SFT by 2.4-7.0 points at 8B and 1.2-7.7 points at 32B. Our 8B model also achieves the strongest reported Online-Mind2Web result among similarly sized open-weight models in our comparison and a leading reported accuracy-cost trade-off on WebVoyager and WebTailBench, with the caveat that external evaluation protocols differ.

Progress in large language models is often summarized using a single scalar measure, such as a time horizon, a latent ability estimate, or an aggregate benchmark score. These summaries capture the overall performance, but they do not test whether progress is distributed differently across task difficulty. We find that most of the apparent shift in gains toward harder tasks does not reflect a change in the shape of the difficulty-response curve. On METR time-horizon data, a single Rasch model with rising ability reproduces this pattern, so it is largely explained by ceiling effects rather than a qualitative change in capability. This echoes how the choice of metric can make claimed emergent abilities look like a property of the models themselves. We then identify a smaller hard-task effect that survives this control. Isolating it is difficult on agentic benchmarks, because newer models are usually run with newer agentic harnesses, so a gain on hard tasks cannot be assigned to the model or its scaffold. We break the confound with LiveCodeBench, a public competitive programming benchmark that runs no agentic scaffold while pairing dated models with an exogenous difficulty ordering. After accounting for the rise in overall ability, models released after September 2024 still gain on the hardest problems beyond what their easy and medium performance predicts, by about +0.40 logits under our most conservative assumption, raising the hard-problem solve rate from roughly 18% to 25%. The effect is led by the strongest reasoning models and holds for hard tasks that need only short reasoning, not autonomy over long horizons. We present this as a result specific to competitive programming, since our clean identification rests on a single coding benchmark. We release the LiveCodeBench Difficulty Panel (66 dated models x 1,055 problems) and our analysis code.

Tool calling allows large language models (LLMs) to invoke external computation during problem solving, a useful capability in various fields including AI for mathematics. We study this setting through weighted sum-of-squares (SOS) decomposition, a machine-checkable route to proving polynomial nonnegativity and hence polynomial inequalities. A candidate decomposition can be checked exactly, but finding one requires choosing among non-unique regroupings and coordinating multiple symbolic transformations. We develop an agent that combines algebraic task training, symbolic tools, and verifier-grounded optimization for this task. Rather than training only on the composite SOS task, we construct 1.35 million synthetic examples covering eight supporting polynomial tasks together with weighted-SOS decomposition. We first apply supervised fine-tuning (SFT) to direct algebra problems and simulated symbolic traces, and then use Group Relative Policy Optimization (GRPO) with task-specific symbolic rewards. The SFT corpus contains no native tool-calling messages; at evaluation, the agent uses native SymPy calls for expansion, collection, reordering, and factorization. Every final SOS answer is checked by exact expansion and coefficient comparison. On held-out, same-generator synthetic problems, the full SFT+GRPO+tools system is the strongest of four evaluated configurations, reaching 78.96% verified success on weighted SOS, compared with 44.73% for the base model with the same tools, and 91.75% macro accuracy across nine polynomial tasks. Within this controlled setting, our work provides a case study of combining domain-specific skill training, executable tools, and verifier feedback, and may inform the design of tool-calling agents in other domains with exactly checkable outputs.

For reinforcement learning (RL) in self-contained environments, a policy can get rewards by exploiting environment-specific regularities (tool schemas, grader parsing, task templates) rather than by acquiring transferable skill, and an in-distribution holdout shares those regularities. We argue that the discriminating question is behavioral, namely how a trained agent acts, and that cross-benchmark transfer is the right place to look for it. We post-train an open-weight mixture-of-experts model (Qwen3.5-122B-A10B) on 363 long-horizon Model Context Protocol (MCP) tasks across 27 categories, using a two-stage SFT-then-RL pipeline. Toolathlon performance informed the initial base-family and SFT-teacher choices, but no external-benchmark task or grader entered training and no external score informed the reward, training hyperparameters, trained-checkpoint selection, or stopping. At greedy pass@1, the trained model improves over the base on five reported external evaluations: Toolathlon (+9.6 pp), $τ^2$-Bench (+5.3 pp), BFCL-V4 (+3.5 pp), SWE-Bench Pro (+5.8 pp), and Terminal-Bench 2 (+2.8 pp). Both software-engineering benchmarks improve despite the training collection containing no software-engineering tasks. An exploratory paired-trajectory analysis identifies four recurring behavioral differences (more careful local-goal formation, building goal-relevant working state, keeping parent goals stable through local repairs, and verifying completion) that appear in analogous forms across office workflows and code. These results provide descriptive evidence that long-horizon multi-tool post-training can change ways of working that transfer beyond its training domain.

Large language model (LLM) judges are increasingly organized as multi-agent panels under the assumption that exchanging critiques improves judgment quality. We test this assumption for groundedness verification, where a judge must determine whether a claim is supported by the supplied evidence. We evaluate a homogeneous three-agent panel on six public fact-verification and hallucination-detection benchmarks. Relative to a fixed single-agent reference, the panel's system-level accuracy difference ranges from $+8.5$ to $-4.4$ percentage points: two datasets show reliable gains, one shows a reliable loss, and three are statistically inconclusive. Because the reference and panel use different model variants, these differences characterize the complete systems rather than isolate a causal debate effect.

Large language model (LLM) agents can retrieve memory, call tools, ask clarifying questions, and vary response style, yet adapting these execution decisions to an individual user remains difficult. Fine-tuning a separate LLM is costly or impossible for proprietary systems, while prompts and memory primarily expose user information to the agent rather than adapt its execution decisions from feedback. We formulate personalization of a frozen agent as online learning of a per-user execution policy from scalar feedback observed only for the executed action. We propose FABLE (Factorized Adaptive Bandit Layer for Execution), a lightweight policy layer outside a potentially black-box host agent. FABLE factorizes memory, information-acquisition, and response decisions so feedback updates related choices; filters actions through an externally specified feasible set before exploration; and learns user-specific residual preferences relative to a fixed default-and-cost score via Bayesian contextual Thompson sampling. Under a linear residual-reward model, a calibrated variant inherits an expected-regret bound against the best feasible action. We also characterize preferences unidentifiable under persistent feasibility constraints and provide anytime-valid false-promotion control. Across personalized-reasoning, controlled-feedback, and executable tool-use evaluations, FABLE improves several preference-sensitive behaviors relative to rule-only control while remaining competitive on end-to-end task performance.

Vision foundation models, such as DINOv2, learn highly expressive representations but rely on massive, opaque architectures that demand substantial computational power and memory. To provide an interpretable-guided and efficient solution to this issue, we first propose a spectral analysis and new visualization technique for individual attention heads based on the Laplacian eigenvectors of their attention maps. Building upon recent observations regarding the block structure of Vision Transformers, we perform semantic clustering of attention heads and identify functional redundancies. Leveraging these insights, we introduce SAPER (Soft Attention PrunER), an end-to-end differentiable pruning framework based on the LapSum Soft Top-K approach. Extensive experiments on ImageNet-1K demonstrate that SAPER achieves a highly favorable accuracy-efficiency trade-off, outperforming the competitive RAPTOR baseline in FLOPs reduction while preserving strong classification performance.

Surgical augmented reality (AR) can provide contextual guidance by overlaying virtual annotations, tool cues, and procedural information onto the surgical workspace. However, the virtual content may obstruct task-relevant real-world information, such as surgical instruments, and interfere with users' perception during time-sensitive surgical tasks. In this paper, we investigate visual obstruction detection for surgical AR and present a latency-aware pipeline that combines vision-language model (VLM)-based surgical-object recognition with segmentation-based obstruction reasoning. To reduce inference overhead, the system adopts a cascaded small-to-large VLM architecture with segmentation-guided early exiting and attention-based visual token pruning. The small VLM handles easy frames when its key-object prediction is supported by segmentation consistency, while difficult frames are forwarded to a large VLM with pruned visual tokens. We construct a pseudo-AR surgical obstruction detection benchmark by overlaying virtual content onto surgical-tool images and labeling whether the virtual content obstructs task-relevant instruments. Evaluation results show that the proposed system achieves 87.43% obstruction detection accuracy with an average end-to-end latency of 479 ms, reducing latency by 62.90% compared with a cloud large-model baseline. These results demonstrate the feasibility of latency-aware obstruction detection for surgical AR and motivate future work on dynamic surgical videos, multi-object scenes, and clinically grounded AR guidance content.

Memory for self-evolving large language model (LLM) agents is often provisioned as if its byte budget only grows. Cloud platforms, however, adjust quotas with load and cost, and we show that capability does not follow the budget back up: after a squeeze-and-recover cycle, the agent settles below its pre-squeeze level, a gap we call memory hysteresis. The cause is structural. Deletion and one-way compression discard the material needed for later rebuilding, and we prove that any policy that only keeps or drops entries carries a residual-deficit floor. We propose CrystalMem (Crystallized Memory), an elastic memory sidecar that demotes entries across four fidelity states under a crystallization-energy schedule, orders demotions by advantage-weighted influence with dependency coupling, and recovers capability through verified recrystallization under explicit compute and byte caps. Across seven environments, seventeen methods, and six backbones, with multi-tenant serving and a physical edge-cloud deployment, CrystalMem achieves the highest restored capability in every setting and closes the loop left open by every baseline. From a 50% byte budget, CrystalMem matches the strongest budgeted baseline at full provision on every environment; at equal budgets, it leads by +4.6 pp on average.

The limited-memory BFGS (L-BFGS) Hessian update scheme is the critical kernel in many quasi-Newton optimization algorithms. The most common approach to implementing L-BFGS uses $2m$ sequential rank-1 updates as part of solving a linear system when there are $m$ history steps. The performance of this approach suffers when the latency of synchronization is significant, and its poor temporal locality increases the memory traffic when vectors do not fit in cache. The compact dense representation of L-BFGS results in an approach that has minimal synchronization latency and better temporal locality, but it requires an additional pass over the basis vectors and an additional basis that must be recomputed when the $B_0$ matrix changes as in variable-metric methods. In the Portable Extensible Toolkit for Scientific Computation and the Toolkit for Advanced Optimization (PETSc/TAO), we have implemented an intermediate dense formulation of BFGS that retains most of the good characteristics of both the recursive and compact dense approaches. We report single-node performance tests of these implementations on the U.S. Department of Energy's Polaris and Frontier machines, testing both GPU-based and CPU-based computations.

Deploying robot foundation models at scale is the next step towards realizing the potential of general-purpose robots. However, Vision-Language-Action (VLA) and other foundation models are computationally demanding, and on-device compute is constrained by power and space. In this paper, we introduce the problem of serving a robot policy to multiple robots from a remote GPU and formulate it as a scheduling problem. We build Armory, a serving system validated on fleets of both simulated and real robots. Our experiments show that naive scheduling heuristics perform well when all robots are the same, but fall short when robots consume action chunks at different rates, uncovering a mismatch between conventional batching methods and the closed-loop requirements of robot policy execution. To address this, we propose a scheduling algorithm that accounts for this heterogeneity and improves overall system throughput by up to $18%$ in real-world experiments. Additional details are available at https://gatech-rl2.github.io/actionchunkscheduling.

A flat BM25 index over one billion documents occupies about 400 GB. Holding it in memory requires DRAM proportional to corpus size. Serving it from disk takes 4-12 seconds per query. Exact top-k lexical retrieval at this scale is therefore impractical within an interactive latency budget. Hierarchical BM25 gives up exact ranking in exchange for fixed bounds on memory and latency. A resident coarse index selects which of ~1K topical, size-balanced document groups a query visits, using two signals: the total frequency of each query term within a group, and, for informative terms spread too thinly across groups for frequency totals to reflect, whether several of them appear together in one document. Selected groups are then searched exhaustively and scored against ~100 KB of global statistics. Every returned score therefore equals the flat index's score, and the approximation is confined to selection alone. The resident footprint is ~4.4 GB, independent of corpus size. Sixteen-term queries over one billion documents return in ~300 ms (4.7x to 5.6x the throughput of a flat multi-threaded index), and a warmed cache sustains ~32 queries per second versus under 3 for flat indexing. At a 500K-document configuration, visiting 5-10% of clusters recovers 0.83-0.92 of the exhaustive result score. Billion-scale recall and a direct comparison against document-reordered BlockMax-WAND remain open.