Papers

Generative music systems can now produce impressive audio from text prompts, but audio outputs are difficult to inspect, edit, and diagnose as musical structure. We introduce Libretto, an agent-facing framework for symbolic music generation and revision. Libretto uses an LLM-native grammar with explicit onset slots, voices, and bar-level organization, then evaluates each piece in a corpus-calibrated statistical space over rhythm, harmony, melody, texture, form, and variation. The same structural axes support retrieval, diagnosis, copy-risk control, and iterative self-revision. Across gap filling, reference-guided full-piece generation, gradual morphing, and educational music generation, Libretto turns symbolic music from a raw token sequence into a measurable and editable object for language-model agents.

Existing driver intervention systems rely on auditory alerts and fixed templates, failing to leverage multi-task recognition outputs. General-purpose metrics such as BLEU and BERTScore cannot capture intervention-specific quality dimensions including risk-urgency alignment, cognitive load, and driver acceptability. In this paper, we propose the Driver Safety-Aware Intervention Score (DSAIS), a domain-specific metric evaluating five dimensions through a hybrid architecture combining lightweight rule-based computation with LLM Judge evaluation, together with an end-to-end framework integrating four-task recognition outputs into an LLM through risk fusion, state history management, and dynamic prompt construction. Experiments on the AIDE dataset with five models and seven conditions demonstrate that DSAIS achieves ICC 0.798-0.840 across three architecturally distinct judges and Cohen's d > 1.5 across all control conditions. Multi-dimensional sub-score analysis quantifies the contextual adaptability gap between rule-based and LLM-based systems, revealing that multi-task integration improves contextual relevance by 9.1% over rule-based baselines. Ablation experiments demonstrate that each framework component contributes to contextual relevance, with sub-score decomposition revealing gains that aggregate scoring masks. Driver emotion recognition is identified as the most critical upstream factor, and compact local LLMs (7B–9B parameters) achieve quality superior to API-based models, providing practical design guidelines for in-vehicle deployment.

Federated Learning (FL) enables collaborative model training without sharing raw data, making it a promising paradigm for privacy-sensitive applications. However, its decentralized nature makes it inherently vulnerable to backdoor attacks, where malicious clients embed hidden triggers into local training data to manipulate model predictions. Existing defenses mainly operate during before and during aggregation cannot fully eliminate backdoor behaviors that persist in the converged global model. Moreover, the effectiveness of post-training sanitization is often limited by the server's lack of knowledge of trigger patterns or poisoned clients after convergence, resulting in residual backdoor behaviors or accuracy degradation due to neuron entanglement. To address this limitation, we propose SCRUB-FL (Sanitizing and Cleansing Representations via Unlearning of Backdoors), a two-phase solution for post-training backdoor removal in FL. During training, clients identify suspicious samples using spectral analysis and activation clustering, then train lightweight Wasserstein Generative Adversarial Network with Gradient Penalty (WGAN-GP) models to capture trigger-related distributions. The generator parameters are aggregated server-side to construct a global representation of suspicious patterns without exposing raw data. After convergence, the server synthesizes trigger-approximating samples and applies machine unlearning to erase the trigger-target association by redistributing predictions toward a uniform distribution. Experimental evaluations on CIFAR-10 and GTSRB across three attack types and up to 40% malicious participation demonstrate that SCRUB-FL reduces the backdoor attack success rate to as low as 3.88% while maintaining over 91% normal task accuracy, outperforming state-of-the-art defenses without requiring prior trigger knowledge or a large clean proxy dataset at the server.

We introduce VISTA Architect, a database-oriented AI architecture for integrating large language models (LLMs) with longitudinal electronic health records (EHRs). At ingestion, it transforms complex clinical documentation into a persistent, provenance-linked knowledge graph, eliminating repeated reprocessing of raw records at query time. The architecture has two layers: a source-faithful MEDS Graph preserving granular EHR structure with full provenance, and a clinically abstracted Timeline Object Architecture (TOA) that uses graph-guided LLM extraction to synthesize a concise timeline of deduplicated, temporally coherent clinical events. This addresses key limitations of direct long-context prompting and retrieval-augmented generation (RAG), which often miss temporal relationships and incur high cost and latency from repeated raw-text processing. By precomputing clinical synthesis once, downstream queries access an organized patient state and traverse to source documentation only when detailed verification is needed. We demonstrate the system in multidisciplinary thoracic oncology tumor boards at Stanford Medicine, where precise reconstruction of patient histories is critical. Across 1,180 patients, VISTA Architect achieved 96.4% accuracy (mean 9.75/10) on 15 tumor board-salient variables (17,700 evaluations; 95% CI 96.1-96.7%), surpassing a matched BM25 RAG baseline and recent benchmarks for LLM-based clinical extraction. An agentic interface reduced preparation for a 30-patient held-out cohort to about 2.2 minutes without sacrificing accuracy. While configured here for thoracic oncology, the modular design adapts to other specialties through customizable event definitions, episode structures, and agentic tools; validation beyond thoracic oncology remains future work.

Modern Large Language Models (LLMs) rely on extensive safety alignment, yet the mechanistic basis of refusal remains opaque. In this work, we investigate whether safety compliance is a deep semantic decision or a manipulable linear feature. We introduce Contrastive Logit Steering (CLS), a zero-optimization framework that isolates the "refusal direction" by contrasting hidden states derived from safe and unrestricted system prompts. Unlike representation engineering methods that intervene on internal activations, CLS operates directly on the output distribution, serving as a diagnostic probe for alignment fragility. When coupled with prefix injection to bypass initial refusal reflexes, this method induces a phase transition where guardrails collapse. Our experiments on 7 model families reveal that safety implementation is architecturally deterministic. While models like Llama-3.1 exhibit a "Late Decision" topology that is easily bypassed by CLS (reaching 95% ASR in approximately one second), others like Qwen-2.5 demonstrate "Early Divergence" by integrating safety mid-computation. Direct comparison with established activation-level steering methods shows that CLS achieves substantially higher attack success rates on Llama 2 (73% vs. 22.6%) and Qwen 7B (91% vs. 79.2%), demonstrating that logit-level intervention exposes alignment vulnerabilities that hidden-state methods underestimate. Beyond attacks, we show that this linearity enables bidirectional control: inverting the steering vector "hardens" models against jailbreaks without retraining. Our findings suggest that current alignment techniques create a steerable "safety axis" that serves as both a critical vulnerability and a precise primitive for defense.

Multi-hop question answering remains challenging for Retrieval-Augmented Generation (RAG) because existing approaches either propagate errors across iterative retrieval rounds or over-generate reasoning steps, increasing cost without improving accuracy. We propose Grounded Delta Planning RAG (GDP-RAG), a plan-based framework that targets only the information delta based on three simple design choices: (1) preliminary retrieval to ground planning before execution, (2) a gap-conditioned planning prompt that asks only for missing information, and (3) a skeletal trajectory that pairs each subquery with a Thought capturing evidence from preliminary retrieval and carrying it through to the final answer. GDP-RAG focuses computation on unresolved gaps, yielding concise, reliable reasoning trajectories. Extensive experiments on HotpotQA, 2WikiMultiHopQA, and MuSiQue show that GDP-RAG achieves the highest accuracy (60.63%) among all compared systems while maintaining a cost-of-pass of 0.51, 22% lower than PAR-RAG (0.65) and 68% lower than KnowTrace (1.57), with no method achieving both higher accuracy and lower cost.

Agentic coding harnesses - such as Agent-Skills, Superpowers, and Agent-Rigor - are increasingly deployed to augment underlying LLMs for real-world software engineering tasks. Existing benchmarks evaluate these agents almost exclusively on outcome correctness: whether generated code passes tests or resolves issues. We argue that this outcome-only lens is insufficient: an agent that arrives at a correct solution through reckless trial-and-error, without planning, verification, or graceful recovery, is fundamentally less reliable than one that follows sound engineering discipline. We introduce RigorBench, the first benchmark designed to measure process discipline in AI coding agents. RigorBench evaluates these harnesses across five pillars: Planning Fidelity, Verification Coverage, Recovery Efficiency, Abstention Quality, and Atomic Transition Integrity. A composite RigorScore aggregates these dimensions into a single metric via a weighted sum. We curate a suite of 30 tasks spanning five categories - Plan-Then-Build, Verify-Or-Die, Doom Loop Gauntlet, Know When to Fold, and Don't Break the Build-and evaluate leading harnesses in a controlled with/without experimental design against baseline coding assistants. Our results show that structured process discipline not only improves process quality scores by an average of 41% but also raises downstream outcome correctness by 17%, providing the first quantitative evidence that how agents code matters as much as what they produce. We release the full benchmark, scoring rubrics, and trajectory analysis tools as open-source artifacts.

Refusal on a safety benchmark does not reveal how stable that behavior will remain after model updates. Benign downstream fine-tuning can weaken refusal, yet behavioral evaluations typically expose this fragility only after an intervention. We introduce SKIN-DEEP, a geometric diagnostic that examines the unmodified model's residual-stream activations. It compares aligned and base checkpoints to identify safety-separating directions, tests their behavioral relevance through ablation, and summarizes the layer-wise pattern in the Geometric Fragility Score (GFS). Across twenty-one instruction-tuned models, harmful requests and benign instructions exhibit a recurring low-rank separation pattern. Selected direction ablations weaken refusal, with the effective direction varying across models. In benign low-rank fine-tuning experiments, the initially safe model with the lowest score before fine-tuning has the lowest harmful-compliance rate when trained on the largest tested set of harmless examples. These findings connect representation geometry to subsequent behavioral susceptibility and support activation-based diagnostics as a complement to refusal tests. Our code is available at https://github.com/js-lee-AI/skin-deep.

This paper introduces Wittgenstein's Rule Following (WRF) data evolution, a framework in philomatics for evolving or generating a new dataset from a sequence of previously observed datasets. The method is inspired by Ludwig Wittgenstein's rule-following considerations and his notion of family resemblance in Philosophical Investigations. Unlike standard synthetic data generation, where the goal is usually to sample from or augment a fixed distribution, WRF aims to continue the implicit rule expressed by a historical sequence of datasets while preserving resemblance to the previous datasets. WRF represents each dataset by structural descriptors rather than pointwise correspondences. These descriptors summarize geometric, distributional, clustering, and, in the supervised case, label-based properties of the data. The method predicts a rule-following target by extrapolating descriptor trajectories and a family-resemblance target by averaging historical descriptors. Candidate datasets are then generated from the observed history through balanced or bounded mixture recombination, scored according to these targets, and optionally refined through differentiable optimization in descriptor space. The proposed framework allows both sample size and feature dimension to vary over time and does not assume that the next dataset is a direct transformation of the last one. Simulations on synthetic and image datasets show that WRF can generate meaningful continuations of evolving datasets in both unsupervised and supervised settings.

Command-Line Interface (CLI) agents based on large language models (LLMs) demonstrate remarkable autonomous capabilities, but they also introduce significant safety and misuse risks during multi-turn interactions with external environments. Existing safety mechanisms mainly rely on external guardrails, which have a limited ability to perform fine-grained behavioral control during execution. Meanwhile, recent mechanistic interpretability methods for LLM safety are mostly confined to single-turn or jailbreak-style QA settings, limiting their ability to capture the evolving risk dynamics of multi-turn agent execution. In this paper, we investigate the safety of multi-turn CLI agents from an internal perspective. We propose Agent MechSuits (Mechanistic Subspace Intervention and Steering), a white-box defense framework that performs runtime safety detection and representation-level mitigation for CLI agents. Unlike conventional agent guardrails, Agent MechSuits detects harmful execution states from step-level hidden representations and mitigates unsafe behavior by intervening in a 10-dimensional subspace within a single layer. To support this research, we introduce the Mechanistic Agent Safety (MAS) benchmark, comprising comprehensively annotated multi-turn execution trajectories across 194 tasks using LLaMA-3.1-8B, Qwen-2.5-7B, and Gemma-2-9B. Extensive experiments show that Agent MechSuits achieves strong safety detection performance, provides preliminary evidence for lookahead risk anticipation, and substantially reduces harmful actions of the CLI agent, establishing a foundation for applying mechanistic interpretability to dynamic LLM agent safety.

Prompt-injection detectors are deployed as guards: a model scores an input and a downstream system trusts or blocks it on that score. I study the confidence of these scores, not only their accuracy, when the attack distribution shifts away from the clean benchmark on which the operating point was chosen. I evaluate three released detectors, ProtectAI-v2 and two Prompt-Guard-2 checkpoints, at a single source-calibrated threshold that I freeze and transport across five shifts. I report a severity metric S, how confident a detector is on the attacks it misses, alongside the false-negative rate and discrimination. Across every shift and every detector, severity on the missed attacks stays between 0.99 and 1.00 while the false-negative rate ranges from 0.01 to 0.97: when these detectors miss, they miss with near-certainty. All three confidently pass indirect behavior-hijack injection, a blind spot unanimous across two vendors and a fourfold size range. Standard pooled calibration error does not register this; one detector it rates well-calibrated, at 0.06, is miscalibrated at 0.91 on the attacks alone. Run against live models, the missed injections leak the majority of working exploits, passing them at the rate they catch others. A controlled experiment traces the cause to content-keying rather than injection structure, an instruction-tuned model used as a judge shows the same hijack blind spot, and a black-box rewriter exploits the content-keying to manufacture working confident misses, most effectively on the most dangerous attack category. Code and data are public.

Accurate identification of the epileptogenic zone (EZ) is essential for seizure freedom after resective surgery in drug-resistant epilepsy, yet seizure freedom rates remain below 50%. We developed EpiiSLM, a dual foundation model system for EZ identification with stereo-electroencephalography (sEEG), by training a signal foundation model on 104,990 minutes of sEEG recordings from the Montreal Neurological Institute & Hospital, while leveraging all recordings regardless of surgical outcome and anchoring EZ biomarker extraction on non-epileptic signals. A language foundation model then integrates sEEG-derived outputs with multimodal clinical information to produce interpretable predictions. Under leave-one-patient-out evaluation, EpiiSLM achieved 0.978 contact-level positive predictive value (PPV), outperforming the seizure onset zone(SOZ)-as-EZ baseline by 15.1% (p < 0.05), and 100% region-level accuracy; on an external dataset, EpiiSLM achieved 0.857 contact-level PPV. EpiiSLM requires only one night of interictal sleep data, suggesting potential to reduce invasive sEEG monitoring duration and improve surgical outcomes.

Large language models (LLMs) frequently encounter inputs that disagree with their prior outputs, through user pushback, retrieved documents, or web search results. While the way they resolve such conflicts – a process we frame as cognitive dissonance resolution – has been characterized behaviorally, its connection to internal model uncertainty is not well understood. To study this systematically, we vary persuasion attempts along two dimensions, source authority and evidence quality, across 12 health-science claims of stratified epistemic status. Dissonance can be resolved through persuasion, backfire, or immunity. We introduce Trust Elasticity (TE), an econometrics-inspired measure of how readily a model is persuaded toward conflicting evidence. Across four LLMs, TE varies substantially, while clearly false claims elicit near-zero TE across all models. On two open-weight models, we further find that this variation is associated with two complementary internal uncertainty indicators, Confidence Miscalibration in Qwen and Internal Uncertainty Change in Llama. These results link cross-model behavioral variation to a measurable internal property and point to interventions targeting internal uncertainty as future work.

Knowledge editing aims to efficiently update factual information in Large Language Models (LLMs) without full retraining. However, existing methods still suffer from performance degradation in batch knowledge editing. We identify that semantic representation entanglement, such as overlapping concepts and shared syntactic patterns, accumulates interference in the representation space and reduces editing precision. To bridge this gap, in this paper, we propose Orthogonal Representation Editing (ORE), which performs edits in the hidden representation space of LLMs by constructing a general semantic subspace and enforcing orthogonal constraints on edit vectors, effectively decoupling semantic entanglement. Furthermore, we introduce a gated non-linear representation head to enable adaptive learning of editing locations and precise control over knowledge injection. Extensive experiments show that ORE outperforms existing methods and achieves superior performance in cross-lingual knowledge editing scenarios. We release our code at https://github.com/YVVH/ORE.

Climate change, primarily driven by carbon dioxide (CO2) emissions, requires accurate forecasting tools to support effective mitigation policies and sustainable development strategies. Existing forecasting approaches typically rely on centralized data collection, which is often restricted by privacy regulations and the distributed nature of emission data across countries and industrial sectors. This paper proposes a novel federated hybrid forecasting framework that integrates ARIMA-based trend modeling, GARCH-based volatility modeling, LSTM-Attention temporal representation learning, and XGBoost prediction within a privacy-preserving federated learning environment. The proposed framework enables collaborative learning among distributed clients without requiring the exchange of raw data. Experimental evaluation across 14 clients demonstrates strong forecasting performance, achieving client R2 values between 0.50 and 0.97 with an average of 0.73, RMSE values ranging from 0.06 to 2.35 with an average of 1.21, and MAPE values between 1.5% and 11.3% with an average of 6.5%. The results indicate that the proposed framework provides an accurate, scalable, and regulation-compliant solution for collaborative carbon-emission forecasting.

Agent skills have become a practical way to extend large language model agents, but the growing skill ecosystem still lacks a reliable way to judge whether a skill is worth deploying. Existing evaluation methods remain largely anchored to fixed task suites, assessing skills through performance on predefined tasks and environments. As skill marketplaces expand, this paradigm becomes inadequate: fixed suites can conflate a skill's marginal contribution with backbone strength and miss its value when tasks fall outside the skill's intended scope. We introduce SkillAudit, an end-to-end framework for skill-centered assessment that takes an arbitrary agent skill as input and automatically generates a comprehensive, multi-dimensional evaluation report spanning utility, efficiency/cost, and safety. SkillAudit focuses on the skill artifact itself and constructs capability-aligned evaluation tasks directly from the skill package. The generated tasks are conducted in isolated sandbox environments to collect execution evidence, followed by automated checks with LLM-based judging to produce auditable results. To dissect the agent skills, we propose the baseline comparison principle to measure utility and efficiency/cost, and introduce a two-stage detection paradigm combining static semantic analysis with dynamic runtime verification to assess safety risks. After scanning top-ranked real-world skill packages spanning 23 occupational categories, we found that over 7% of skills are at risky status.

Large language models have become capable reasoners and tool users that write and run code and search the literature, which makes automating the research process itself a realistic goal. We present PAPERCLAW, a harnessed multi-agent system that carries a project autonomously, from a field of study to a finished paper. PAPERCLAW curates a domain from a field's live literature, datasets, and code; brainstorms it into an idea with a pre-registered main-result contract; and drives a stoppable hypothesis map through an iterative propose, test, reflect loop that grows only from measured verdicts and halts once the evidence supports the idea, at which point it writes a venue-compliant paper. A full-lifecycle memory keeps each stage in a single living record, so a long run can be paused, inspected, and resumed without losing context. At the centre is an in-cycle research assistant with research tools and skills: it can drive the whole pipeline on its own, while the same interface lets a person step in at any stage, turning a first autonomous draft into a stronger paper through human-in-the-loop refinement. Throughout, PAPERCLAW keeps its output grounded and checkable, citing only references validated against open scholarly indexes and reporting results that genuinely ran. An evaluation with an LLM judge finds that PAPERCLAW produces strong papers both fully autonomously and with human-in-the-loop refinement.

On-Policy Distillation (OPD) improves the learning efficiency of standard reinforcement learning through dense, token-level supervision from teachers. In the standard KL objective of OPD, token-level losses are uniformly averaged, implying equal weights for all tokens. However, we discover that not all tokens are created equal: as student rollouts grow longer, they deviate further from the teacher's distribution, leading to degraded supervision quality at later positions. As a result, OPD using only the first 30% of tokens can perform comparably to using all tokens, whereas OPD using only the last 30% of tokens barely learns anything. In this work, we provide a principled understanding of this issue through the lens of constrained optimization. Based on these insights, we derive Importance-Weighted On-Policy Distillation (IW-OPD), in which the weight assigned to each token depends on the accumulated discrepancy between the student's and teacher's distributions, naturally upweighting earlier tokens and downweighting later ones with larger deviations. We show that IW-OPD converges significantly faster than OPD, with better learning efficiency, and achieves better final performance than standard OPD in both same-size and cross-scale settings, improving performance up to 6.9 points on AIME-2025.

Domain-specific languages (DSLs) are widely used for managing operating system security policies, yet manually authoring rules in such languages demands high expertise and is error-prone. This paper formalises the task of automatic DSL code generation from natural language descriptions - Text2DSL - as a distinct problem class, separate from Text-to-SQL and general-purpose code generation. We introduce the PolkitBench dataset comprising 4,204 verified natural-language-to-Polkit-rule pairs, each validated through a three-level AST-based pipeline. Controlled prompt experiments on two MoE models of different scale and provenance - GigaChat-10B-A1.8B (1.8B active parameters) and Nemotron-3-Nano-30B-A3B (3B active) - demonstrate the critical role of structured context (BNF grammar, API specification, permitted identifier vocabulary) for LLM-based DSL code generation. Across both models, supplying context raises syntactic validity to 98.6-99.4%, structural validity by +9.7 to +35.5 pp, and the CodeBLEU score by +60% to +95%. The consistency of the effect across models of different scale and provenance indicates that, for the Text2DSL class of problems, injecting a formal target-language specification into the prompt context is a robust enabling factor for high-quality generation without model fine-tuning.

Host intrusion detection systems (HIDS) based on system-call traces are typically trained and evaluated against individual Common Vulnerabilities and Exposures (CVE) instances. In operational settings, however, defenders need to recognise new exploits of an already known type of weakness. We empirically examine whether a one-class anomaly detector trained on the normal behaviour of a set of CVEs that share a Common Weakness Enumeration (CWE) class generalises to a different, unseen CVE inside the same class. Using six scenarios drawn from LID-DS-2021 and grouped into three CWE families (CWE-307 broken authentication, CWE-89 SQL injection, CWE-434 unrestricted file upload), we extract a 66-dimensional Peng-Guo-style feature vector per sliding window and train Isolation Forest and SGD One-Class SVM detectors with normal-only thresholds calibrated to fixed target false positive rates. We define and answer four research questions covering self-detection, asymmetric cross-CVE transfer, the value of a combined CWE-level normal profile, and the effect of feature filtering on transferability. The combined CWE-307 detector reaches F1 = 0.6976 at calibration target FPR = 0.05 (precision = 0.8994, recall = 0.5698), whereas CWE-89 and CWE-434 collapse to F1 <= 0.21 under the same protocol. Cross-CVE transfer turns out to be strongly direction-dependent and dominated by the breadth of the source normal profile rather than by the CWE label. We conclude that CWE-level generalisation in HIDS is empirically attainable for some but not all weakness families with current syscall features, and we argue that calibrated FPR is a methodological prerequisite for honest reporting in this setting.

We extend our prior work on Text2DSL automatic generation of domain-specific language (DSL) code from natural language descriptions along two complementary axes. First, we replace prompt-only synthetic generation with context-aware distillation, in which a teacher large language model (DeepSeek-V4-Flash) operates under an explicitly defined structured context comprising a BNF grammar, an API specification, and a closed identifier vocabulary; the resulting corpus is verified by a two-tier pipeline combining AST validation through esprima and runtime acceptance through the production polkitd daemon and the pkcheck client. This scales the verified PolkitBench corpus from 4,204 to 10,073 natural-language-to-Polkit-rule pairs at 100.0% AST validity and 99.7% runtime pass rate. Second, we conduct the per-component factorial ablation of structured context that was identified as future work in the precursor study: eight conditions C0-C7 are evaluated on GigaChat-10B-A1.8B with the new corpus. Three findings emerge. (i) The new harder corpus collapses the baseline mode (Syntax Valid 97.6% -> 58.5%, Combined Score 0.482 -> 0.252), whereas the context-enhanced mode degrades only marginally (Syntax 98.6% -> 97.4%, Combined 0.801 -> 0.750), confirming that structured context is not a cosmetic improvement but a load-bearing mechanism. (ii) The best absolute condition is the full context C7 across all metrics, while the strongest partial conditions (C5 = BNF + Vocabulary, C6 = API + Vocabulary) both contain the vocabulary. (iii) A Shapley-style decomposition assigns the largest semantic-quality effect to the vocabulary (Combined +0.198), the largest structural-validity effects to API (+24.7 pp) and BNF (+22.3 pp).

While synthetic data generation resolves the manual labeling bottleneck in computer vision, minimizing the syn-to-real domain gap requires optimizing rendering variables. This paper presents a systematic study analyzing the impact of lighting configurations and background complexity on object detection performance. We introduce SmartSDG, an automated, reproducible pipeline built on NVIDIA Isaac Sim using Physically-Based Shading (PBS), alongside ILLUM\_INTRUCK, a new multi-object industrial benchmark dataset. Through 18 controlled experiments utilizing a state-of-the-art YOLOv12 framework, we demonstrate that complex, indirect lighting configurations paired with domain-relevant background variability significantly increase visual cue richness. Our quantitative findings show that avoiding direct specular peaks preserves crucial surface textures, mitigates the domain gap, reduces false positives, and accelerates model convergence compared to using conventional direct-light synthetic data. Ultimately, we provide actionable virtual scene design guidelines to maximize object detection robustness in industrial automation.

Few-shot prompt learning is an effective strategy for adapting CLIP to downstream tasks, but class-only prompt optimization can overfit base-class supervision and weaken transfer to unseen classes. We propose Concept-Constrained Prompt Learning (CCPL), a lightweight regularization framework that anchors learnable class prompts to frozen concept-level text prototypes without updating CLIP encoders. CCPL learns a set of shared context tokens, instantiates class prompts by appending class names, and constructs frozen concept prototypes from a class-level concept bank. During training, a text-space cosine consistency objective aligns learnable class-prompt embeddings with frozen concept prototypes; concept dropout provides additional regularization against over-reliance on fixed concept lists. At inference, CCPL optionally fuses class-prompt logits with concept-prototype logits using a controllable ensemble weight alpha. Our default configuration uses text-space concept regularization lambda = 0.5, concept dropout p = 0.3 and weak concept-guided fusion (alpha = 0.1), with no KL-based prediction consistency term. Experiments under identical automatically-generated fallback splits show that CCPL improves the base-to-new harmonic mean on DTD (+0.6) and EuroSAT (+2.9) compared with CoOp, while remaining near-neutral on OxfordPets (-0.1). Ablations indicate that text-space concept regularization is consistently beneficial, while the best concept-guided inference strength is dataset- and protocol-sensitive. These results suggest concept constraints are most effective when concept prototypes align naturally with dataset semantics, and identify fine-grained categories as a current boundary condition. The code is released at: https://github.com/richael-sang/concept-constrained-prompt-learning.

Chain-of-Thought (CoT) has become a standard method for improving reasoning capabilities in large language models (LLMs) by eliciting step-by-step thinking, but its effectiveness in multimodal tasks remains unclear. In this paper, we aim to systematically investigate the key question: What can multimodal Chain-of-Thought reasoning do, and where and why does it fall short? To this end, we evaluate 12 multimodal tasks across perception and reasoning categories using both 14 non-reasoning models and 8 reasoning models. Our analysis reveals several important findings: (1) CoT is not a free lunch and should be used selectively depending on the specific requirements of each task. For perception tasks, CoT can lead to undesirable side effects, such as reduced performance in visual grounding and object counting. In contrast, it proves effective for reasoning tasks involving mathematical, scientific, and multi-image reasoning; (2) Compared to original models, existing open-source multimodal reasoning models often yield only marginal overall improvements, possibly due to an overemphasis on mathematical reasoning at the expense of broader capabilities; (3) Visual reasoning remains a key bottleneck for current multimodal CoT, as models exhibit a Look Light, Think Heavy pattern where verbal reflection rises and falls during reasoning, whereas visual reflection consistently diminishes. These findings suggest that while multimodal CoT handles verbal reflection relatively well, it lacks the ability to maintain deep visual introspection throughout the reasoning process.

Computer use agents (CUAs) have advanced rapidly in desktop automation, and a growing number of users deploy CUAs such as OpenClaw on Mac Mini for always-on automation. However, existing benchmarks, including those for macOS, evaluate agents without framework augmentation and rely on binary evaluation. As a result, they fail to capture both the framework capabilities leveraged by modern CUAs and the partial progress on long-horizon, multi-application tasks. We present MacAgentBench, a comprehensive macOS agent benchmark comprising 676 tasks across 25 applications, with nearly 60% involving both GUI and CLI interaction. The benchmark adopts deterministic rule-based evaluation and introduces fine-grained multi-checkpoint scoring with capability annotations for multi-application tasks. Experiments across three frameworks and 16 models show that the best configuration, Claude Opus 4.6 on OpenClaw, attains 73.7% Pass@1, while this advantage is primarily driven by the skill library rather than by framework design. Fine-grained metrics further reveal that models with similar Pass@1 can differ substantially in sub-goal completion. Our code and data are publicly available at https://github.com/JetAstra/MacAgentBench.

Autoregressive visual models (AVMs) based on next-scale prediction have emerged as a prominent paradigm for image and video synthesis. However, decomposing the generation process into discrete scales with varying granularities in AVM makes semantic errors difficult to identify and correct, thereby undermining the quality of the final output. Prior efforts to enhance AVM can be categorized into training-based and training-free approaches. Although training-based efforts to enhance AVM generation quality come at substantial computational cost, existing training-free methods neglect intermediate generation states, leaving semantic errors undiagnosed and allowing them to accumulate into the final output. In this paper, we focus on training-free paradigms and propose Gazer, a framework that integrates multimodal large language model feedback into the AVM sampling loop for in-generation semantic correction. Concretely, Gazer operates via two cooperating stages: the Reflective Diagnosis stage diagnoses semantic errors from intermediate states, while the Semantic Correction stage rewinds and rectifies the generation trajectory to realign with the target prompt. Experiments on compositional image and video benchmarks demonstrate that Gazer improves semantic alignment and compositional accuracy across multiple AVMs without additional training.

Classical uncertainty sets for robust optimisation impose fixed geometric shapes that cannot represent the complex dependencies present in real-world data. We propose Generative Robust Optimisation (GRO), a framework in which a deep generative model defines the uncertainty set as the image of a neural network decoder over a calibrated latent set, naturally accommodating nonlinear correlations, asymmetry, and multimodality. A five-point evaluation framework (reconstruction fidelity, distribution matching, latent regularity, robust relevance, and computational tractability) provides systematic, model-agnostic criteria for assessing any neural network-based uncertainty set. We instantiate this framework with a Wasserstein Adversarial Autoencoder employing Gaussian mixture model-guided training for latent regularity and constraint-consistency regularisation for robust relevance. Restricting the decoder to ReLU activations enables exact worst-case verification through mixed-integer programming embedding. Extensive experiments on a production planning problem across six uncertainty distributions and six generative architectures, together with a multi-period facility location study, validate the framework and demonstrate that systematic attention to all five criteria yields uncertainty sets that are simultaneously expressive, well-calibrated, and optimisation-tractable.

Modern LLM agents increasingly rely on context compaction, summarization, or eviction to keep long-running sessions within a token budget. We show that this context-management layer is a safety-critical failure surface: in-context governance constraints that agents reliably obey while visible can be silently removed by compaction, causing the same agent to perform prohibited tool actions later in the session. We call this failure mode Governance Decay. We introduce ConstraintRot, a benchmark of long-horizon agent scenarios with deterministic tool-call grading, and measure compaction-induced violations across seven model families. Across 1,323 episodes, violation rises from 0% with the policy in full context to 30% after compaction, reaching 59% for some models; when the constraint survives the summary, violation remains 0%, but when it is dropped, violation reaches 38%. We further study a Compaction-Eviction Attack, in which adversarial in-context content biases the summarizer to omit a legitimate policy, and show that optimized injections defeat every evaluated model. Finally, we propose Constraint Pinning, a simple training-free mitigation that quarantines governance constraints from lossy compaction and restores violation to 0% in our benchmark. These results identify context management as a first-class governance surface for deployed LLM agents.

While federated learning enables collaborative modelling on decentralised data, standard methods merely fit historical observations. This purely observational approach is fundamentally insufficient for interventional inference and policy evaluation, as sequential actions dynamically alter future states. We propose Fed-CausalDiff, a federated causal diffusion framework for do-simulation. The architecture decomposes the evolution of the latent state into a global causal score function and a local confounding score function. This design enables decoupled synchronisation (DSS), where clients aggregate only the shared causal mechanism while retaining site-specific confounders locally to handle heterogeneity. Experiments on four datasets demonstrate that Fed-CausalDiff achieves better ATE and policy-value estimation accuracy, offering a favorable trade-off between communication cost and inference fidelity.

This work investigates the safe exploration problem in reinforcement learning, where an agent must maximize cumulative performance while simultaneously satisfying safety constraints. This challenge becomes even more pronounced in long-horizon tasks, where existing safe methods face fundamental limitations due to compounding estimation errors and restricted exploration capabilities. To address this problem, we propose a method that combines a learnable world model with two complementary policies a high-level policy and a low-level policy to promote safety at both hierarchical levels. The high-level policy generates intermediate subgoals that bias exploration toward safe regions, while the low-level policy uses imagined rollouts in the learned world model to reduce unsafe behaviors when reaching these subgoals. The proposed method was evaluated on challenging long-horizon navigation and manipulation tasks with high-dimensional action spaces, where it significantly outperforms existing Safe RL baselines in both success rate and strong empirical constraint satisfaction, consistently meeting the prescribed safety budget across seeds, while prior approaches fail to effectively solve these complex long-horizon scenarios.