Papers

Reinforcement learning (RL) post-training has emerged as a powerful framework for enhancing the capabilities of large language models (LLMs), enabling impressive reasoning, math, and coding capabilities. Yet for many researchers and practitioners, the principles behind classical RL remain a "black box". In this work, we deconstruct the RL post-training algorithm, investigating each step to clarify what is actually happening beneath the surface. By isolating the mechanics of RL with Verifiable Rewards in a controlled and simplified environment, we examine how RL outcomes are shaped by the base model's prior distribution, the granularity of the reward signal, the diversity of the prompt distribution, and model scale. We use the entropy of the policy's output distribution as a lens to compare the distributions learned through pretraining, SFT, and RL post-training, revealing how each stage shapes model certainty. Our investigation sheds light on how these choices interact to affect post-training success. For example, we show that the effect of so-called 'spurious rewards' depends on the prompt distribution used for post-training. We also provide insight into why the success of RL post-training depends on whether the base model already places sufficient probability mass on the desired behavior, linking it to the classical concept of exploration in RL. Ultimately, we provide this primer as a resource to those in the NLP community wishing to incorporate RL as a tool in their toolbox.

End-to-end training of multimodal neural networks often exhibits unstable neural dynamics characterized by three coupled failure modes that degrade learning: (i) modality imbalance, where one branch dominates gradient-based optimization; (ii) unstable gating, where noisy confidence cues induce erratic modality selection; and (iii) fusion interference, where modality-specific gradients conflict at the shared fusion layer. We propose CAT-GS (Calibrated, Adaptive, Thresholded Gating with Fusion Surgery), a neural dynamics-based optimization controller for intelligent computing applications. CAT-GS operates during backpropagation without modifying model architectures, fusion modules, or task losses. Through calibration of teacher-derived reliability via temperature scaling and EMA smoothing, CAT-GS stabilizes neural dynamics using a margin-thresholded policy to switch between warm-up dropout, weak-modality prioritization, and weak-biased blending, stabilizes gradient magnitudes under aggressive gating via capped gradient-budget renormalization, and applies fusion-only PCGrad to reduce destructive cross-modal interference at the primary shared bottleneck. We evaluate CAT-GS on audio–visual multimodal pattern recognition benchmarks (CREMA-D, AV-MNIST, and VGGSound), a tri-modal setting (UR-FUNNY), controlled synthetic data (CG-MNIST), and additional cross-domain benchmarks (AVE and CMU-MOSI). CAT-GS improves or matches fused multimodal accuracy against strong imbalance-aware baselines (including OGM-GE, G$^2$D, and UMT) across settings, and yields smoother gating behavior with fewer conflicting fusion gradients.

Recent years have witnessed remarkable achievements of Large Language Models (LLMs) in multiple domains, while the excessive resource requirements of LLMs hinder the deployment on resource-constrained devices. Although model quantization stands out as an effective approach, conventional quantization approaches typically incur severe performance degradation due to uniform bit-width or simple heuristic sensitivity evaluation. In this paper, we propose a novel Fisher information-based Adaptive Mixed Precision Weight Quantization approach, i.e., FAMPWQ, which performs layer-adaptive weight quantization for effective LLM inference on commodity GPUs. First, we propose a system model with a novel Fisher information metric to measure the layer-wise sensitivity to quantization. Second, we propose a reinforcement learning-based bit-width allocator in FAMPWQ, which generates an adaptive bit-width allocation strategy based on the Fisher information sensitivity metric. Extensive experiments on 7 models and 5 benchmarks demonstrate that FAMPWQ significantly outperforms 7 baseline approaches in terms of PPL (up to 3.39 smaller), accuracy (up to 6.87% higher), and LLM-as-a-judge comparison (up to 76% win rate).

Partial differential equations (PDEs) often have high-frequency and multi-scale features that neural networks struggle to approximate. Physics-Informed Neural Networks (PINNs) build the governing equations directly into training, but suffer from spectral bias: they learn low-frequency components faster than high-frequency ones. Techniques such as Fourier feature embeddings and sinusoidal activations address this, but most studies assume they help across the board without checking which spectral regimes actually benefit. We introduce a dual-branch, spectrally-gated architecture (DBSG-PINN) that splits low- and high-frequency components into separate subnetworks joined by an adaptive gate, and use it to run a partially controlled ablation of frequency decomposition and spectral routing. We test this on five one-dimensional benchmark PDEs, ranging from smooth, single-scale problems to oscillatory, multi-scale ones. Frequency decomposition helps most on the spectrally complex benchmarks, cutting relative $L_2$ error by up to $59.2\%$ on a multimodal wave problem, but gives little benefit on smoother PDEs. On one benchmark (1D Wave), it performs substantially worse than a simpler fixed-combination variant. The gate's benefit scales with how spectrally rich the target solution is: the full model's advantage over the ablations is largest on multi-scale benchmarks and smallest (or negative) on single-scale ones, consistent with the gate exploiting frequency structure rather than acting as noise,though we do not directly visualize or quantify its spatial activations in this study. All results come from a single training seed across five 1D benchmarks, so we present this as an exploratory study meant to raise questions rather than answer them, and outline the additional seeds and benchmarks needed to test whether the pattern holds.

Mixture-of-Experts (MoE) models scale capacity for strong quality while keeping per-token compute bounded through sparse expert activation. Yet low-latency MoE serving is increasingly challenging, because it spans two inference phases with fundamentally different bottlenecks: prefill is dominated by token-wise expert computation, whereas decode is constrained by memory traffic from the batch-wise activated expert set. However, existing training-free acceleration methods optimize only a single resource proxy, either the experts each token executes or the experts a batch activates, and either discard the excluded experts' contribution or leave it only implicitly approximated. In this paper, we propose ExFold, a unified training-free expert-folding framework for jointly accelerating MoE prefill and decode. ExFold casts both prefill and decode as one budgeted output-approximation problem: execute only a phase-specific constrained expert set while projecting the contribution of budget-excluded experts onto retained experts using calibrated scalar projectors. Motivated by the observation that many expert outputs are directionally aligned but differ in magnitude, ExFold calibrates a pairwise scalar-projector matrix on unlabeled data and uses it at inference time to fold excluded expert contributions into retained experts. Under this view, prefill acceleration becomes token-level Top-K folding, and decode acceleration becomes batch-level expert-pool folding. The two phases differ only in how retained experts are selected, while excluded contributions are recovered by one shared folding mechanism. We implement ExFold as a plug-and-play plugin in vLLM, with a lightweight expert-folding CUDA kernel, delivering up to 1.41x TTFT and 2.45x TPOT speedups while retaining about 99% of the original average quality.

Multi-Modal Anomaly Detection (MMAD) detects rare abnormal events from heterogeneous data sources and is increasingly used in safety- and reliability-critical applications such as industrial inspection and cybersecurity. Yet the literature is fragmented across domains and modality combinations, and existing surveys usually group methods by architecture rather than by how abnormality is defined and separated in multi-modal settings. We survey MMAD from an assumption-driven perspective. We formalize the problem, identify five intrinsic characteristics underlying its core challenges, and organize prior work into two complementary paradigms. The first, normality-assumption methods, models regularity via representation learning, cross-modal alignment, and knowledge enhancement. The second, anomaly-assumption methods, sharpens decision boundaries through coarse-grained, structural, and semantic anomaly injection. We also investigate how foundation models are reshaping MMAD through scalable pretraining, flexible cross-modal transfer, and emerging reasoning capabilities. Finally, we compile representative benchmarks and evaluation protocols across domains and highlight open problems and future directions for robust, adaptive, and interpretable MMAD systems.

Data mixing is a central design problem in large language model pretraining: given a fixed token budget, practitioners must decide how much data to allocate to each domain. Recent proxy-based methods address this problem by training small models on candidate mixtures, fitting a response model, and using the response to select mixtures for larger-scale training. We show that this workflow has the structure of a classical mixture experiment. Under this view, data domains are mixture components, token shares are component proportions, proxy-training runs are experimental design points, and validation loss defines a response surface over the probability simplex. We develop this formulation using sparse second-order Scheffé response-surface models and construct model-robust $\mathcal{I}$-optimal designs for proxy data-mixing experiments. Using RegMix as an empirical case study, we demonstrate how the framework can both interpret observed mixture responses and design more efficient proxy experiments. The Scheffé analysis shows that domain value is strongly relational: several domains that are weak under additive effects become favourable through pairwise interactions, especially through combinations with web-derived text. The sparse Scheffé model preserves mixture rankings across model scales and remains competitive with a flexible machine-learning predictor while providing an explicit decomposition of additive and interaction effects. In a simulation study calibrated to observed proxy-training responses, model-robust $\mathcal{I}$-optimal designs recover the relevant mixture ordering after removing about 25% of the original proxy runs. These results suggest that LLM data mixing should be treated not only as a prediction problem, but also as an experimental-design problem in which the proxy mixtures themselves can be chosen to improve statistical efficiency.

Large Language Models excel at code generation, yet competitive programming exposes a persistent failure mode: existing multi-agent pipelines distribute work over generic planner, coder, and debugger roles and delegate the choice of algorithmic technique to the backbone alone. We present MARS (Multi-Agent Relay of Specialized LLMs), a prompt-only framework in which each agent is a topic specialist—dynamic programming, graphs, strings, geometry, and so on—grounded by retrieval-augmented generation over an algorithm-theory corpus. Given a problem, retrieval selects a small team of relevant specialists; a starter writes an initial C++17 solution, and each subsequent turn runs the candidate against public examples in a sandbox, lets the active specialist keep, repair, or hand off the draft, and forwards a structured packet to the next specialist. A single infrastructure-fixer pass normalizes boilerplate at the end. On the CodeContests test split with Gemma 4, MARS reaches $0.624 \pm 0.006$ pass rate at $2.3$ recorded pipeline stages per task ($+14.4$ percentage points over direct prompting), closing most of the gap to CodeSIM ($0.731$) at $3.3{\times}$ lower wall-clock cost and substantially smaller variance in per-task token spend. The source code is available on GitHub: https://github.com/fckand/mars.

Supervised fine-tuning on teacher-generated trajectories is the standard first stage for distilling tool-calling capabilities into deployable models. Post-training pipelines that drive shipped tool-calling agents re-run this stage on a daily or weekly cadence, paying the frontier-teacher cost each cycle, yet the mechanism is generate-and-filter (keep the teacher's passing trajectories, discard the rest) and each cycle leaves behind the same hard scenarios because failures supply no signal. On {\tau}2-bench, 57% of teacher trials fail, two-thirds of them near-misses (most tool calls correct, undone by one decisive error). We introduce PROOF-Gen (Per-scenario Reflective Optimization to Overcome FailedGeneration), which recovers golden trajectories from these failures via per-scenario prompt optimization. For each failed task, a reflector analyzes the execution trace and evaluation feedback, then writes corrective guidance that steers the teacher to a passing trajectory. The guidance is stripped before training, so the student learns from clean demonstrations with no task-specific scaffold. On {\tau}2-bench, per-scenario optimization recovers 93% of failed scenarios. Fine-tuned on the combined data, Qwen3-4B-Instruct-2507 improves from Pass^1=0.132 to 0.529 and Gemma 4 E4B-it gains +7.2pp on BFCL v4 multi-turn. In a deployed pipeline, the method lifts trajectory quality by +6.3pp goal completion and transfers to a deployed on-device model (+1.5pp goal completion; +1.7 to +5.0pp across response-quality metrics), with positive transfer in every locale (non-English average +1.48pp).

Tax-loss harvesting demonstrates consistent benefits to long-term portfolio growth; yet implementing it efficiently often involves complex considerations that are specific to the holdings within that portfolio and the individual who owns it. We introduce a custom capital gains calculation engine and a RAG-retrieved vector store of market advisory reports to provide context for a multi-agent trade recommendation system. We investigate the effects of each context provider on the quality of recommendations, measured by relative capital gains incurred during portfolio liquidation. A 2x2 repeated-measures ANOVA revealed a significant main effect of the tax optimization engine ($F(1,29) = 9.17$, $p = .005$, $\eta^2_p = .240$): enabling the engine reduced tax savings by approximately 55 percentage points relative to the no-engine conditions. The RAG main effect was not significant ($p = .841$), nor was the interaction ($p = .553$). The RAG-only condition achieved the highest descriptive mean tax savings (47.7%), and the baseline condition performed second-best (30.6%), suggesting that the pre-trained language model's internalized financial knowledge may be sufficient for competent tax-loss harvesting recommendations without explicit tooling. These results indicate that augmenting LLM agents with domain-specific computation engines does not guarantee improved performance and may introduce conflicting optimization signals.

Artificial Intelligence (AI) is increasingly integrated into complex sociotechnical systems, including Critical National Infrastructure (CNI), where harms emerge from interactions between technical, human, and organisational elements. Yet current AI evaluation remains model-centric, offering little insight into how observed behaviours might translate into system-level risk. We propose a framework that links structured hazard analysis, component-level testing, and probabilistic system modelling to bridge this gap. By providing a traceable pathway from model behaviour to system-level outcomes, the framework enables practitioners to answer the "so what?" of AI failures, quantify their systemic impact, and move toward evidence-based and anticipatory governance of AI in complex systems. Applied to the UK's Real Time Gross Settlement (RTGS) system as an illustrative worked example, we derive AI-driven loss scenarios using Systems Theoretic Process Analysis (STPA) and examine adversarial manipulation of LLM-based trading as one such loss scenario. Component-level experiments show that simple adversarial inputs induce measurable behavioural shifts where AI recommendations are followed. Under the component-to-system mapping used here for a financial contagion model, these shifts alter system resilience, increasing bank failures and lowering the threshold at which shocks lead to cascading disruption, particularly under widespread or monopolistic AI adoption.

We introduce BenchBench-Protocol, a benchmark for large language models of 149 protocol-modification tasks recovered from modifications that scientists made to published protocols during real experimental work. Adapting a published protocol to a new experiment is a routine task for a wet-lab scientist, and a correct modification requires accounting for prior choices and downstream steps. Recent life-science benchmarks have moved toward open-ended, rubric-graded tasks, but tasks are typically elicited from experts rather than reconstructed from real-world modifications. BenchBench-Protocol tasks are derived from differences between a published protocol and a version a scientist modified, which provides the basis for the query and the weighted rubric elements for a correct response. The benchmark draws from 96 source protocols across nine domains of wet-lab biology and only includes tasks rated highly after review by domain experts. We evaluate nine closed and open models; Claude Opus 5 scores highest at 59.2% normalized rubric score, with other models between 34.1% and 47.1%, and the benchmark remains unsaturated when taking the best of ten attempts. As models are increasingly helpful in life-sciences research, evaluating them on routine wet-lab tasks becomes correspondingly important. We present BenchBench-Protocol as both a grounded assessment of wet-lab reasoning and evidence for the utility of real-world experiments to construct benchmark tasks.

When a code generating language model fabricates a Python package name, an adversary who has pre-registered that name on PyPI can convert that hallucination into a supply chain compromise. This event has been termed as 'slopsquatting'. We propose a two layer detector to counter this issue. The first layer performs a deterministic PyPI existence check. The second is a Random Forest classifier trained on ten features derived from the package name and its PyPI metadata. An import name reconciler bridges the two, resolving cases such as 'import cv2' versus 'pip install opencv-python' without a security bypass. The detector is embedded in a LangGraph state machine that retries at escalating temperatures and, on repeated failure, routes to a stronger fallback model. Across 300 curated prompts, the pipeline produces hallucination free code on 76% of runs. The primary exhausts its retry budget on 28.7%; intra model retries recover roughly a quarter of those, and cross model fallback recovers a further 16.5% of the remainder. Four findings have been observed. First, half of the flagged hallucinations are packages already registered on PyPI, as low quality lookalikes of well known projects, caught by the classifier rather than the deterministic layer (e.g., pil, faiss, tabula, haystack). Second, hallucination rate scales almost linearly with prompt adversariality, from 0 to 10% on routine coding to 40 to 73% on slopsquat baits. Third, the weaker primary refused 6 of 10 direct baits unaided, suggesting recent instruction tuning provides a baseline defense. Fourth, when primary and fallback share a model family, approximately 84% of primary failures recur on the fallback, motivating cross family pairing. A user study (n = 24) reports mean satisfaction 4.4 out of 5 and 21 of 24 stated adoption intent.

Kohn–Sham density functional theory (DFT) underpins electronic-structure simulations, but repeated orbital diagonalizations lead to cubic scaling, restricting quantum calculations to modest scales only. Eliminating these auxiliary orbitals while retaining Kohn–Sham accuracy is the central goal of orbital-free DFT, but both analytical and machine-learning methods have so far fallen short. Prior learning approaches either try to learn the variational kinetic-energy functionals, which are ill-conditioned, or directly predict the ground state, which extrapolate poorly to larger systems. Instead, we identify the Kohn–Sham map as the right learning target for orbital-free DFT. It maps a Kohn–Sham potential directly to the corresponding density and noninteracting kinetic energy, quantities otherwise obtained through an orbital diagonalization. Focusing on the density component in this work, a domain-invariant $\mathrm{SE}(3)$-equivariant Fourier neural operator learns to predict it from the potential as input on real-space grids, enabling stable quasi-linear scaling SCFs. Trained jointly on 8,504 molecules and solids, a single model generalizes to out-of-distribution organic molecules, insulators, and metals. For the first time, the same method converges SCFs across these systems without explicitly constructing Kohn–Sham orbitals, while reproducing densities, electronic spectra, and structural observables at Kohn–Sham DFT accuracy. Linear-scaling SCFs additionally allow converging magnesium dislocation densities containing up to 82,500 valence electrons on a single GPU.

Learning systems deployed over long periods must adapt not only to statistical changes in incoming data, but also to revisions of the definitions that generate their prediction targets. Conventional concept-drift methods typically infer such changes from observations or prediction errors, even when the underlying policy, rule, or query has been explicitly modified. This paper studies rule-induced concept shift, where the target-defining concept is revised directly, causing previously stored instances to acquire different semantic labels without requiring any change in their observed data. We introduce a provenance-guided incremental learning framework that compiles consecutive concept definitions into a structured rule delta, traces the changed components through historical provenance, certifies records whose previous labels remain valid, and restricts reevaluation to a localized candidate region. Executable revisions are relabeled automatically, ambiguous cases are handled through selective supervision, and the resulting changes are used for incremental predictor repair. A versioned concept memory further supports recurring definitions. We also introduce RuleShift-Bench, spanning financial, demographic, cybersecurity, and graph-structured data with threshold, predicate, logical, relational, recurring, and mixed concept revisions. Across the benchmark, provenance-guided repair attains 92.3% accuracy and 90.2% Macro-F1 while reprocessing 14.7% of the historical collection and retaining 94.6% of affected records. Its average update latency is 179s compared with 993s for complete relabeling and retraining. The results demonstrate that an explicit concept revision can be exploited as a data-maintenance signal, allowing learning systems to update the supervision and predictive state that depend on the change while preserving knowledge that remains valid.

This article presents the abridged core of A Mathematical Theory of Interpretation (MTI), which treats interpretation as observer-relative spectral measurement under an access structure. MTI makes interpretation a method-design problem: access, query, utility, and medium determine what an observer can select, identify, communicate, or refuse. On a learning-invariant Hilbert realization, Rational Entropy measures residual uncertainty across knowledge, utility, and medium. In the finite-effective regime, we classify its zero set. Pairwise confusability is equivalent to uniform atomic collapse, while a unique utility maximum can select one atom even when other zero-cost states remain non-atomic. This reverses the usual zero-error role of confusability: agreement in at least one observer direction excludes unresolved multi-atom readings, while the joint label preserves identification. The corresponding free-design capacity is the product of all but the smallest direction budget. A four-condition certificate characterizes sharp, decodable, medium-faithful, and order-independent readout on a finite commuting code sector and returns typed obstructions when those guarantees fail. Together, these results establish MTI as a theoretical basis for constructing interpretation methods with explicit access assumptions, guarantees, and failure modes.

Real-time optimization (RTO) relies on process models to locate economically optimal operating conditions. Because developing first-principles models requires significant process knowledge, data-driven alternatives are increasingly attractive. Modern machine-learning models can fit historical plant data accurately and often pass standard validation tests. Whether such models can be trusted for economic optimization, however, remains unclear. We investigate this question using a vinyl acetate monomer benchmark process with a unique, well-conditioned economic optimum. We train a structured hybrid model that combines known mass balances and thermodynamics with a neural-network closure for unknown kinetics, and a fully data-driven neural ordinary differential equation (ODE) model. Both models reproduce plant measurements accurately and exhibit little variation in predictions across random initializations. Yet their economic optima differ substantially from that of the plant. Where the plant returns a single optimum on multistart search, the trained models return many phantom optima. We further show that the training optimizer alone can be yet another source of error. Even with noise-free data and initialization at weights that recover the plant optimum, stochastic gradient training can drift to weights that yield substantially worse RTO solutions. The identified model is thus an artifact of the training optimizer as well as the data. These results demonstrate that a good predictive fit of all available measurements does not guarantee reliable economic performance. A data-driven model for RTO should at least be required to recover the optimum on a decision-oriented benchmark like the one developed here before being considered for plant testing and application.

Artificial Intelligence (AI) algorithms frequently learn creative and unexpected solutions, surprising even expert researchers who develop and study them. They often astonish practitioners by discovering unanticipated behavior, exploiting loopholes in reward signals, or spontaneously uncovering previously unknown scientific phenomena. However, accounts of such unconventional behavior across machine learning are seldom formally documented. This work presents 26 curated firsthand anecdotes from various machine learning subfields representing the work of over 100 researchers. These anecdotes showcase the capability of modern AI systems to circumvent human-imposed design limitations and discover unexpected solutions to the tasks we train them on. Furthermore, these accounts are particularly important for the safety of future AI systems. They illustrate the fundamental challenge of aligning models with human values without diminishing their creativity, so they can make surprising discoveries without producing surprising, potentially harmful outcomes. The paper first details AI achieving superhuman success through reinforcement learning across many challenging domains. However, reward-driven optimization can fail when the model learns to hack an underspecified reward or unarticulated constraint. We then present case studies suggesting that harnessing internet-scale foundation models (FMs) has not resolved these fundamental challenges and, in fact, can supercharge them. Nevertheless, we argue that these same learning dynamics can be harnessed to accelerate scientific discovery. Finally, we hope this work provides a consolidated resource to inform future research and demonstrates that the tendency toward unexpected behaviors is commonplace in modern AI, highlighting the need to anticipate and manage AI's capacity for innovative, yet unpredictable, solutions. (abstract abridged)

Everything a language model sees is tokens. The serving stack knows what each span is – user input, tool output, instructions – but the model must keep track of that itself, and can lose track or be confused: text can be written to read like anything. Prompt injection is a natural exploit of this phenomenon. By scrambling the model's understanding of span identity, an attacker can induce unwanted and dangerous actions. Adding a non-textual channel to the model's input – a way to communicate span identity beyond text – mitigates this class of attack. We thus introduce a general steering technique called Semantic Overlays: small learned adapters applied at chosen prefill positions to a frozen model's residual stream. Laying an overlay over a span creates an out-of-band annotation channel that cannot be replicated by tokens. Unlike steering vectors, Semantic Overlays are trained, adaptable, and selectively applied. An overlay can encode complex semantics that reshape how the model perceives the marked span: asked to copy a code snippet under an overlay asserting a different programming language, the model rewrites the snippet in the asserted language. Overlays compose, allow transparent reading of underlying content, and can carry complex payloads – including imperatives the model will follow. An overlay which marks a span as "non-executable" defends against the broad class of prompt injections that add instructions in untrusted context. We report strong results on five prompt injection benchmarks: SEP separation rises from 24.3% to 99.0% with utility unchanged (our scoring rule; we correct a defect in the published grader), TensorTrust attack success falls from 34.8% to 6.2%, AlpacaFarm from 99.0% to 0%, and the overlay beats every published PIArena defense that leaves the model able to answer – while marked spans stay readable, all at >95% character similarity to the original.

When it comes to safety policies for generative AI, one size does not fit all. Each organization and use case needs to mitigate different risks depending on the application context, regulatory environment, organizational values, and user personas. Yet, existing policy specification approaches are designed for traditional access control and fail to capture the nuances of GenAI application: the enforcement of content-based constraints. We present two contributions to address this gap: (1) the Actionable Policy schema, a YAML-based format for specifying what model responses can and cannot contain. The schema enables exception-based policy governance, proposing exceptions to track policy violations; (2) synthetic data generation pipeline that produces policy-aligned training data for model alignment and testing, and a set of tools to help define the schema and enforce policy. Together, these enable organizations to specify policies once and enforce them throughout the GenAI application lifecycle: from model alignment to runtime monitoring. The Actionable Policy schema, example policies, and tools are available as open source: https://github.com/ibm-granite/granite.trust.policy-tools We welcome new ideas, contributions and feedback.

Revelation Control is the problem of choosing priced interventions that reveal hidden state only insofar as the revealed distinctions can change a consequential decision, while accounting separately for any useful progress created by the intervention itself. We develop this theory for learning systems, where states equivalent under declared current information can respond differently to future training and favor different actions. The framework defines decision-sufficient revelation and revelation depth, separates pure information value from productive reuse, embeds static Bayes refinement into state-dependent continuation value, and gives an exact cost-adjusted factorization criterion: an additional shallow coordinate is decision-nonredundant only when states sharing a scalar summary lie on opposite sides of the priced Stop/Continue boundary. We also give a target-independent protocol for model-specific instantiation and prove that bounded stop-flip risk alone cannot certify positive expected utility under unrestricted severity. Across Qwen2.5-7B and Mistral-7B-v0.3, deeper future-learning probes have positive decision value and productive reuse yields strict equal-compute utility advantages. Qwen additionally provides evidence for a decision-nonredundant shallow revealability regime; in Mistral, a scalar continuation architecture fit only on an independent development panel retains positive familywise-adjusted lower bounds on a disjoint target panel, consistent with scalar decision sufficiency within the tested architecture family and resolution. The evidence supports structural rather than numerical transfer: the decision theory, cost accounting, continuation logic, and evaluation protocol transport, while empirical proxies, coefficients, thresholds, and even the required shallow state dimension may be system-specific.

The Agent Payments Protocol (AP2), introduced by Google, enables large language model (LLM)-driven shopping agents to authorize and execute payments on behalf of users. Its signed Checkout and Payment Mandates protect the integrity of transaction data after signing. Agent interactions and external inputs that shape a transaction before authorization remain outside that protection, including Agent-to-Agent Protocol (A2A) messages and Model Context Protocol (MCP) tool calls. Prior work identified replay and prompt-injection attacks in AP2 v0.1. AP2 v0.2 addresses some of these issues but adds capabilities and deployment assumptions that require renewed analysis. We present a systematic security analysis of AP2 v0.2 based on its roles, transaction lifecycle, deployment architectures, and trust boundaries. We divide the lifecycle into five phases and identify five deployment architectures. Using MAESTRO (Multi-Agent Environment, Security, Threat, Risk, Outcome), we model four threat actors, eleven attack surfaces, eighteen adversary capabilities, and six attacker goals. The resulting catalog contains 48 threats spanning five attack families. We score these threats with the Artificial Intelligence Vulnerability Scoring System (AIVSS), identifying eight that reach the High band in at least one architecture. Because no complete public AP2 deployment was available, we build a testbed spanning all five architectures and develop five proof-of-concept demonstrations covering all eight High-risk threats and their mitigations. We also develop a deployment-aware scanner that maps applicable threats to static, cross-role consistency, and adversarial checks. Our analysis shows that valid mandate signatures alone do not ensure that an agent-mediated transaction reflects the user's intent when its pre-authorization context is manipulated.

We propose ICI-Time, a novel framework that reframes time series forecasting as a visual inpainting task, leveraging the generalisation power of large vision models (LVMs). Unlike methods that require specialised temporal architectures and extensive domain-specific training, ICI-Time transforms time series into structured visual representations (area charts) and applies visual in-context learning, reformulating forecasting as pattern completion within a grid-structured prompt that pre-trained vision transformers can solve without fine-tuning or architectural modification. Temporal dependencies are represented through spatial layout, with a consistent, invertible mapping between numerical and visual domains. Extensive experiments across epidemiology, meteorology, and power systems demonstrate that ICI-Time performs competitively against deep learning baselines and shows promising adaptability under limited-data settings, introducing a new paradigm that bridges temporal and visual domains.

Budget-constrained agentic search arises when an LLM agent must refine candidates under a small evaluation budget, because validation is expensive, generation requires multiple model calls, or both. In this regime, standard MCTS allocates budget poorly: exploration bonuses dominate at low visit counts, unpromising siblings are expanded before promising chains can deepen, and branching is independent of node quality. We introduce ExTS, a tree-search policy that treats expansion itself as a value-of-information decision. ExTS combines three mechanisms: discriminative reward shaping to separate candidates under narrow score distributions, a stochastic virtual child that estimates the value of creating a new branch from the parent's reward history, and quality-conditioned branching that expands only when a node's score justifies the budget cost. Across prompt optimization, code generation, molecular structure elucidation, and agentic workflow optimization, ExTS is competitive with or improves over task-specific tree-search baselines, with an average relative gain of +5.5% using a single fixed configuration. We further introduce pilot-run diagnostics that characterize what makes budget-constrained agentic search problems structurally different from one another, providing both understanding of the problem space and practical guidance for adaptation.

This paper proposes the Coronavirus Optimization Algorithm (COA), a SARS-CoV-2-inspired success-history adaptive evolutionary optimizer for box-constrained continuous global optimization. COA does not model disease transmission; instead, it maps selected coronavirus mechanisms to explicit search operators, including elite-guided attraction, trial-vector generation, adaptive parameter variation, stagnation recovery, and population-size scheduling. The algorithm combines opposition-based initialization, current-to-pbest mutation, binomial crossover, an external archive, success-history adaptation, population reduction, and partial restart. COA is evaluated on 29 CEC 2017 benchmark functions at 10, 30, and 50 dimensions against 15 competitive optimizers. Results show that COA achieves the best overall Friedman rank across all dimensions, with particularly strong performance on composition functions. The findings demonstrate that COA is a compact, transparent, and competitive adaptive evolutionary optimizer, while also highlighting limitations on some hybrid functions and the need for further high-dimensional validation.

DevOps programming (e.g., using CLI/API scripts or IaC frameworks) is key to cloud infrastructure management. Unlike traditional programming tasks, DevOps program testing needs provisioning and execution against actual cloud resources, which is often time-consuming, unsafe, and costly. Cloud emulators have gained popularity for easing DevOps program testing; they are generally API-level mocks that can execute DevOps programs in a local environment. Still, building these emulators remains challenging: developers must manually interpret extensive cloud documentation and handcraft logic for each service, API, and their interaction. This does not scale to the complexity of the cloud, which is further a moving target as the services and APIs evolve. CloudEmu is an automated approach that constructs emulators based on cloud documentation via neurosymbolic code synthesis. The key idea is to combine LLMs' general strengths in documentation understanding and code generation with cloud-specific symbolic abstractions that suppress hallucinations and enforce precision at scale, while using the real cloud as an oracle for automated testing, repair, and alignment. Our evaluation shows the effectiveness of CloudEmu on major cloud provider (AWS and GCP) services in both coverage and accuracy. CloudEmu outperforms the existing leading tool LocalStack, which was manually developed by a large team of engineers over a decade.

Training large-scale AI models often outgrows a single data center, demanding sharded, multi-cluster, and decentralized training. However, the huge space of resource allocations makes exhaustive benchmarking and manual tuning impractical, while performance depends on tightly coupled factors like model size, GPU memory, batch size, bandwidth, and sharding strategy. We introduce ShardMeter, a lightweight analytical performance model that predicts the end-to-end runtime of transformer-based workloads across arbitrary sharded, distributed, and even decentralized training. Given a model's characteristics and a target hardware topology, ShardMeter estimates per-GPU and per-island throughput, training cost, total wall-clock time, and identifies performance bottlenecks. Our analysis reveals diminishing-return regimes as island size increases, quantifies transitions between compute- and communication-bound scaling, evaluates hyperparameter trade-offs, and models cost-throughput for large-scale decentralized training. ShardMeter exposes these insights to quickly explore the configuration space, choose near-optimal deployment plans, and avoid costly trial and error.

Embodied Agents System (EAS) are increasingly deployed in open-world physical domains, where reliability directly dictates deployment quality and human-agent trust. However, existing evaluations rely on outcome-centric metrics as success rate or safety scores that collapse diverse execution trajectories into coarse scores, obscuring the dynamic processes underlying agent behavior. Therefore, they ignore a critical property of EAS – which we define as the Resilience – that reflects how EASs recover, stabilize, and extend under perturbations and across iterative updates. The lack of resilience is particularly critical in open-world environments due to continuous unexpected disruptions, thus directly affecting the quality of EAS deployment. To address this problem, we gain insight from the resilience-engineering concepts to EAS groundings and propose a novel resilience evaluation framework that can be flexibly applied to any EAS. Specifically, we define the first comprehensive resilience metrics suite for EASs system that exposes Rebound, Stability, and Graceful Extensibility across embodied tasks execution, providing a practical grounding for EAS resilience analysis. We further implement the resilience evaluation layer that transforms execution process into assessments for diagnosis and optimization. Across 400 household tasks with 10 EAS, we reveal the process-level distinction hidden by outcome metrics, including recovery cost differences among successful episodes ($\Delta C_{rec}=25.2$), increased instability and task-family degradation. Metrics-guided optimizations reduce recovery cost and increase stability, graceful extensibility completion, showing the diagnostic effect of resilience evaluation. Our results reveal a trade-off among resilience characteristics, suggesting that a resilient EAS construction should be configured according to deployment-specific requirements.

Healthcare documentation in the neonatal intensive care unit (NICU) presents significant challenges, with nurses spending approximately 25% of their time on record-keeping, while up to 60% of interventions remain undocumented. Motivated by the need to detect interventions from video automatically, we present the Infant Care Video Dataset (ICVD), a collection of 4,144 videos spanning 12 simulated intervention classes designed for developing automated documentation systems. Our manikin-based approach systematically varies conditions, such as camera angle and clinician skin tone, while ensuring privacy compliance. Using video transformer architectures (TimeSformer and MotionFormer), we establish strong baseline performance (93.97% and 93.17% top-1 accuracy) among the 12 infant care classes. Our ablation study comparing temporal models with a framewise approach (23.17% accuracy) demonstrates a 70.80% performance gap, validating the need for temporal modeling. The ICVD provides a foundation for developing automated documentation systems to reduce clinical burden in neonatal care environments and improve existing practices.

Large language models (LLMs) are known to exhibit social sycophancy, often validating or agreeing with users in socially sensitive contexts. Existing evaluations typically measure sycophancy under a fixed prompt formulation, leaving unclear whether such behavior is stable when the same underlying situation is presented with different sycophancy-relevant prompt variants. In this work, we study sycophancy prompt sensitivity: the extent to which changes in user confidence, emotional framing, social consensus, or validation-seeking language alter a model's sycophantic behavior. We refer to our evaluation framework as SyPS, short for Sycophancy Prompt Sensitivity. Building on existing social sycophancy evaluation settings, SyPS constructs controlled prompt variants that preserve the same underlying user situation while varying sycophancy-relevant social cues. We introduce the Sycophancy Prompt Sensitivity Score (SPSS), an instance-level measure of sycophancy variation across paired prompt variants. Unlike aggregate sycophancy rates, SPSS separates baseline sycophancy from prompt-induced shifts, enabling model-level comparisons of robustness to sycophancy-relevant social cues. Empirically, we find that sycophancy prompt sensitivity is socially structured: validation-seeking and emotional-pressure cues often increase sycophancy, whereas counter-framing and anti-sycophancy prompts tend to reduce it. Our framework highlights whether LLMs maintain stable social judgments while adapting appropriately in tone.