The era of Big Science has long been defined by increasingly large and specialized research teams pushing the frontiers of knowledge. However, recent advances in artificial intelligence (AI), particularly large language models (LLMs), are beginning to reshape academic writing and scientific research, potentially disrupting the longstanding trend toward ever-larger teams and transforming other dimensions of research team structure. Drawing on 147,074 full-text publications from the PLoS family and the Nature portfolio since 2020, we examined whether and how AI-assisted writing influences team structure and team outcomes in science. Using multiple methods, including ordinary least square, quantile regression, Poisson regression, logistic regression and propensity score matching, we found that research teams using AI-assisted writing tend to be younger and smaller. Importantly, this shift toward more compact, junior-leaning teams does not come at the expense of scientific impact. On the contrary, we observed a higher probability of research teams that employed AI-assisted writing producing highly impactful publications. These results highlight the significant role of AI-assisted writing in reshaping not only how research is produced, but also how research teams are formed and assembled. Our findings call for policy improvements in research evaluation, funding, and training to address this emerging trend.

Written reflection assignments give students valuable opportunities for critical self-assessment, meaning making, and learning processing. Additionally, such reflections provide rich data for qualitative education research. However, qualitative data can be time-consuming to analyze. It is even more time-intensive to qualitatively compare findings between different groups of participants, usually limiting comparison to, at most, one variable (e.g., binary gender). Large language models (LLMs) have recently begun to be critically evaluated for use as qualitative research assistants. Using a longitudinal case of written student reflections (n=151) from a study abroad program, we investigate how LLM-assisted sentiment analysis can enable longitudinal mixed-methods research combining computational and thematic analyses. First, statistical testing is used to quantitatively compare sentiment differences according to seven different student identity/lived experience variables. Then, these results inform qualitative data analysis to investigate the reasons underlying these differences. For the case of undergraduate students studying abroad, we found that prior experience living abroad was the only personal variable impacting students' sentiments of their verbal language and communication behaviors. This workflow has implications for how qualitative researchers can more easily probe multiple variables when comparing participants from different demographic groups.

Open-ended grading is central to equitable and personalized education, yet manual grading remains time-consuming and costly, underscoring the need for automated grading systems. Although recent neural and large language model (LLM) based systems have demonstrated superior performance, they are typically black-box models whose scoring processes and rationales are difficult for educators to verify and trust. Concept bottleneck models (CBMs) have emerged as a promising approach by routing predictions through human-interpretable concepts, providing a mechanistic guarantee of transparency. However, standard CBMs are not tailored to open-ended grading: they do not explicitly model fine-grained rubric dimensions, inadequately capture the ordinal semantics of scoring scales, and neglect inherent reliability issues in human concept annotations. To address these limitations, we propose REC-CBM, a rubric-aware error-correction concept bottleneck model for trustworthy open-ended grading. REC-CBM introduces a rubric-aware concept encoder that learns concept-specific representations over responses and an ordinal pairwise calibration objective that preserves ranking structure among rubric dimensions. It further incorporates a latent concept error-correction module that denoises concept predictions before final grade prediction while preserving interpretability. Comprehensive experiments on publicly available datasets show that REC-CBM consistently improves grading performance and produces more faithful concept-level reasoning than both state-of-the-art baselines. Further analyses validate the contribution of each component and demonstrate the applicability in realistic educational settings. Overall, this work provides a practical, interpretable grading solution that enables educators to inspect, intervene in, and trust automated decisions, advancing more transparent and trustworthy education.

There is a growing interest in utilizing synthetic populations for a diverse range of applications. At the same time, we are witnessing a tremendous growth in artificial intelligence in all walks of life. This paper evaluates whether zero-shot large language model (LLM)-generated health survey data can serve as inputs to a conventional iterative proportional fitting (IPF) workflow for geographically explicit population synthesis. Using the 2023 Behavioral Risk Factor Surveillance System (BRFSS), we generate synthetic survey records for the U.S. states of Colorado and Mississippi with GPT-4.1 and Gemini-2.5-Pro. We use the generated data in an IPF-based synthesis pipeline and evaluate the resulting census tract-level synthetic populations against external benchmarks. Results show both LLMs capture several major state-level contrasts, indicating zero-shot generation produces geographically differentiated survey data. However, performance is strongly variable-dependent. Downstream effects in population synthesis are mixed, as IPF sometimes amplifies or reduces errors in the generated data. Spatial validation shows that LLM-based populations reproduce census tract-level patterns reasonably well, especially for variables that were more aligned with the ground truth data. Overall, the LLM-generated survey data shows promise as supplementary input, but not yet as a replacement for real survey data.

The rapid uptake of generative artificial intelligence (AI) in higher education is reshaping assessment practices and intensifying concerns around academic integrity, fairness, and learning quality. While institutional responses increasingly emphasise policy guidance and ethical principles, there remains limited formal understanding of how collective norms of responsible or opportunistic AI use emerge and stabilise within student cohorts. This paper reframes student AI use in assessment as a coordination problem shaped by peer expectations and assessment design rather than individual compliance alone. We develop a coordination-based evolutionary game-theoretic framework that captures learning value, effort, perceived fairness, and transparency, with institutional AI governance modelled implicitly through reflective assessment incentives. We use analytical results and finite-population simulations to reveal threshold-driven behavioural transitions in student AI use: small, well-calibrated changes in reflective assessment incentives can trigger rapid shifts towards responsible, learning-oriented AI-use norms, whereas weak or misaligned incentives allow opportunistic practices to persist. These non-linear dynamics explain why policy statements alone often fail to change behaviour, while modest assessment redesigns can have disproportionate effects. By providing a mechanism-level account of how assessment structures shape collective AI-use practices, this work offers higher education institutions an analytically grounded tool for Future Facing Learning, supporting proportionate, pedagogy-led AI governance without reliance on surveillance or punitive enforcement.

What looks like acceleration can be a quiet transfer of burden from the present to the future. Attempts to replace human labor with AI systems are often presented as rational responses to technological progress, but that view is often structurally short-sighted. Across software development and adjacent knowledge industries, AI is increasingly attractive because it appears to reduce labor costs, speed output, and improve short-term metrics. Yet those gains may be achieved by drawing down human capabilities that are slow to build and difficult to restore. This paper develops a mechanism of capability masking and capability erosion under AI labor substitution. AI-generated output can create the appearance that organizational capability has been replaced, even when dependence on skilled human labor remains. That appearance can support hiring restraint while slower costs accumulate in the background. Evidence from AI-assisted coding shows that generated output still requires substantial human verification and remains uneven in correctness, maintainability, and security. Repository-level studies also suggest limits in handling broader codebase context. More broadly, labor-market, political-economy, and industrial-strategy evidence suggests that substitution pressures are being driven by managerial cost incentives and national competition while increasing risks of concentration and platform control. The result is a system that may look more efficient in the short term while becoming more fragile over time.

Autonomous AI agents now plan, decide, and act on behalf of users across healthcare, financial services, and workplace contexts, often without step-by-step human approval. Existing AI literacy frameworks were built for a world in which humans evaluate AI outputs and decide whether to act; they have no vocabulary for the user who has delegated decision-making authority to an agent whose actions may not be observable, reversible, or controllable. This paper names the resulting problem agentic literacy debt: the accumulating societal deficit that grows when agentic AI systems are deployed at scale without corresponding literacy infrastructure. The debt compounds through three reinforcing channels (normalization of opaque delegation, multi-agent ecosystem complexity, and institutional path dependence), and it is incurred by the organizations that deploy agents but paid by the users, patients, and citizens on whose behalf the agents act. Evidence from healthcare, financial fraud, and global equity contexts suggests the gap is already consequential. The problem is structural, not a temporary lag that curriculum reform will close. It demands a reframing of AI literacy as a governance capability, not an evaluative one.

As the rapid proliferation of AI systems and harms spurs efforts in AI governance around the world, prioritizing among competing policy options has become increasingly challenging for policymakers and researchers. We introduce a methodology for identifying viable policy options to mitigate specified AI harms, helping policymakers and researchers target areas that warrant greater time and resource investment. This method combines participatory evaluation of policies, expert assessment of implementation costs, and an LLM-based assessment of perceived harm mitigation under each policy option. We leverage a genetic algorithm-based simulation study to explore a vast solution space of potential policy combinations, and examine how outcomes change under different weightings of cost, participatory input, and harm mitigation. We find that this method enables exploration of different balances between participatory and expert components, allowing policymakers and researchers to assess how much weight to assign to each. We argue that the diversity of viable policy combinations found by the genetic algorithm could be a useful starting point for deliberation. This method operationalizes existing work on participatory AI by integrating it directly into practical policy development pipelines.

Determining whether published scientific findings can successfully be replicated is a long-standing challenge in the empirical sciences. Existing approaches for replicability assessment typically rely either on human judgment, i.e., creative assembly of human experts, or on machine learning models trained on paper content metadata. While both approaches have demonstrated value, each also has important limitations. Human forecasts can be influenced by cognitive biases and narrow exposure to the research literature, while automated assessments often struggle to capture contextual cues and subtle signals of credibility. In this paper, we examine a hybrid approach. Specifically, we introduce a hybrid prediction market in which algorithmic agents trade alongside human participants to jointly estimate the likelihood that a published scientific finding will be corroborated via the outcome of a controlled replication study. Agents are trained on outcomes from hundreds of prior replication studies while human participants contribute domain knowledge through real-time trading. We evaluate this hybrid approach through multiple live experiments involving participants from different academic disciplines and compare its performance to artificial-only and human-only baselines. Our results show that, except for a few cases, hybrid markets match or outperform artificial prediction markets, producing more accurate and reliable replication forecasts.

This paper examines whether students are entering the generative AI era with sufficiently strong educational foundations, focusing on the relationship between learning environments and changes in ICT related career aspirations across countries. The analysis uses country-level data from PISA 2018 and 2022, combining indicators of student autonomy, digital skills and teacher support. A mixed-method approach is applied, including descriptive statistics, regression analysis, clustering, latent representation learning (using Variational Autoencoder-VAE), discriminant analysis and probabilistic modeling to capture both observable and latent dimensions of educational readiness. Unlike prior research that treats learning loss, digital skills and career expectations separately, our analysis integrates them within a comparative longitudinal framework. It shifts the focus from short-term post-pandemic effects to the structural capacity of education systems to prepare students for digital and AI-driven labor markets. Results show a global but uneven increase in ICT career aspirations. Digital skills emerge as the strongest and most consistent predictor, while teacher support plays a complementary role. Autonomy shows weaker, context-dependent effects. Educational readiness is multidimensional, and ICT aspirations evolve relatively independently from other career domains.

We study how conditioning context shapes personalization behavior in a teacher-facing educational recommender system. We compare contextual conditioning based on the current student question with memory-based conditioning using persistent learner information. Using deviation correlation and paired statistical tests, we find that contextual recommendations exhibit stronger question-level responsiveness, while memory-based recommendations exhibit history-dependent behaviors, including learner-specific differentiation under identical input. Teacher-facing evaluation signals suggest these recommendations are interpretable and actionable. These results indicate that embedding-based similarity metrics capture responsiveness to the current question but do not characterize personalization grounded in learner history, motivating behavior-level diagnostics for studying conditioning effects.

This position paper reflects empirical data collected during my PhD from a large-scale within-subjects study (N = 90). The study compared a highly human-like, spoken embodied conversational agent (ECA) against a low human-like text base agent (no embodiment, text bubble only) within a mobile, Unity-developed game about pre-decimal UK currency. The game included two agents with different roles-an Instructor (Alex) and a Shopkeeper/Collaborator. Users interacted using voice and mouse input. The quantitative data I collected included a usability questionnaire (CCIR MINERVA) and the Agent Persona Instrument. Data was analyzed using paired t-test, repeated measures ANOVA and multiple linear regression to identify correlations between the persona and usability. The results showed a statistically significant preference for the version of highly human-like agents, with a large effect size. This is further discussed alongside qualitative findings from observations and exit interviews. The results are framed for Human-Agent collaboration, especially for how roles, mixed-initiative dialogue, and breakdowns/repairs become apparent in goal-oriented tasks. I conclude with questions on timing, user expectations, and role-specific interactions. This submission does not propose new frameworks; it reports empirical findings and questions I hope to workshop with the community.

Claims about recursive self-improvement in AI often slide from repeated internal revision to the possibility of qualitatively stronger capability without clearly distinguishing the underlying computational regimes. This paper gives a formal separation result in classical computability theory that blocks that move under a precise modeling assumption. For an oracle $A$, let $\mathcal{C}(A)=\{B : B \leq_T A\}$ be the corresponding computational layer. We prove that finite internal self-modification remains inside $\mathcal{C}(A)$, while stabilized revision is governed instead by the jump $A'$ via the relativized limit lemma. Together with a local closure versus escape theorem, this yields a clean formal separation between within-layer iteration and ascent to a stronger relative level. The point is not that stronger layers never arise, but that they are not explained by finite repetition inside one already settled layer. The resulting separation gives a computability-theoretic limit on a broad class of recursive-improvement narratives in which repeated internal updating is treated as sufficient for qualitative capability ascent.

MerLean-Prover is an end-to-end Lean4 theorem prover that replaces sorry declarations with kernel-checkable proofs. It is built from three agent types (Planning, Check, and Lean) composed by a recursive outer loop whose unit of revision is the proof plan itself, and uses no fine-tuning, no custom RL objective, and no theorem-specific scaffolding. On FormalQualBench, a benchmark of 23 PhD-qualifying-exam theorems, MerLean-Prover solves 10/23, surpassing the strongest published open-source baseline (OpenGauss, 8/23). On Putnam2025, the same harness closes 12/12 with substantially lower total wall-clock than the next-best system that closes the full set. The harness also transfers to smaller models: Sonnet closes all four tested FormalQualBench problems, and Haiku closes the two short ones. These results suggest that harness design is a central factor in end-to-end Lean4 theorem proving, alongside raw model capability, and that a relatively simple harness can already be effective.

Generative retrieval (GR) maps queries directly to document identifiers (docids) using parametric knowledge, However, this design makes corpus expansion costly: adding new documents requires updating model parameters to encode new document-docid associations incurs repeated training and catastrophic forgetting of previously indexed documents. In this work, we revisit incremental GR as an in-context retrieval problem, where newly added documents are supplied as inference-time document-docid evidence. We propose ICICLE, an in-context indexing framework that performs source-aware docid generation over both parametric memory and context-provided document-docid pairs. ICICLE combines a `[COPY]`-based routing mechanism, preference-based calibration, and large context adaptation to distinguish context-grounded retrieval from parametric retrieval. Experiments on MS MARCO and NQ320K show that ICICLE improves retrieval of newly introduced documents while preserving seen-document retention without corpus-specific retraining. Our analysis further shows that high-shot degradation is mainly caused by routing failure, highlighting source-selection calibration as a key bottleneck for scaling in-context generative retrieval.

Practical garment design spans two modes: intuitive creation from high-level intent, such as a reference image or text description, and complex low-level editing across 2D sewing patterns and 3D draped geometry, which requires professional training to navigate their complex interdependencies. Yet existing frameworks address only part of this challenge, offering either garment generation from casual inputs or direct editing on sewing patterns. To support both ends of the spectrum, we propose Garment Particles, a 5D point-cloud representation that jointly encodes 2D sewing patterns and 3D geometry. This representation enables Garment Particles Flow (GPF), a rectified flow framework that supports intuitive generation from high-level inputs (text, images, sketches) and various editing operations on 2D sewing patterns and 3D geometries via diffusion posterior sampling. Finally, we introduce Particles-to-Pattern Flow that converts generated garment particles into curved-based patterns for simulation. We validate our model's generation ability on multiple datasets, achieving state-of-the-art garment generation results against competitive baselines. Our model also enables many garment editing scenarios, including garment interpolation, sewing pattern editing, point-cloud- and silhouette-conditioned garment generation. Our project website is at https://garment-particles.github.io .

Functionality-correct repository setup aims to configure execution environments (e.g., dependencies, build scripts) to successfully execute a repository's documented features. It presents significant challenges due to diverse, repository-specific failures, including dependency incompatibilities, missing toolchains, incomplete installations, and verification-strategy mismatches. Existing LLM agents struggle to robustly resolve these issues, specifically failing to support (1) cross-repository experience transfer, (2) multi-step trial-and-repair under non-invertible state changes, and (3) robust verification of setup outcomes to distinguish setup-induced failures from repository bugs. To address this, we introduce SetupX, an experiential learning-based setup framework. First, we construct a Self-Evolving Experience Representation (XPU), a dual-modality knowledge unit encoding setup signals, textual guidance, executable actions to dynamically transfer verified environment fixes to unseen repositories. Second, we employ Experience-Augmented Speculative Execution backed by a LIFO Docker snapshot stack, enabling the agent to proactively trial fixes and safely roll back to known-good states. Third, we introduce a Prosecutor-Judge Verification Protocol that separates evidence collection from final judgment, enabling more reliable setup verification beyond superficial build-time metrics. Evaluation results on carefully-crafted benchmarks show SetupX achieves highest performance (e.g., 92% pass rate) and outperforms the strongest baseline by over 19%. Crucially, SetupX excels in complex multi-repository setup requiring coordinating multiple interconnected services across different containers. The code repository is available at https://github.com/OpenDataBox/SetupX.

We study denoising score matching (DSM) when the latent distribution is supported on a smooth embedded manifold $M \subset \mathbb{R}^D$. Under ambient Gaussian corruption, the tangent denoising target contains a singular normal-fiber noise channel whose variance diverges as $d/σ^2$ as $σ\to 0^+$. We show that conditioning on the nearest-point projection $π(X)$ canonically removes this singularity: the resulting conditional expectation is the unique $L^2$-optimal Rao-Blackwellized predictor of the tangent DSM target among all estimators depending only on the projected observation $π(X)$. We then compute the small-noise expansion of this canonical target and show that it equals the intrinsic Riemannian score up to an explicit order-$σ^2$ correction that decomposes into an intrinsic Tweedie term and an extrinsic curvature term involving the Weingarten and Ricci operators. In the flat case, the construction reduces exactly to ordinary lower-dimensional Gaussian DSM, while on $S^d$ the extrinsic correction simplifies to the scalar factor $(1-d/2)\nabla_M \log q$; this extrinsic $σ^2$ correction cancels identically on $S^2$, though the intrinsic Tweedie term remains.

Electroencephalography (EEG) visual decoding remains challenging due to the modality gap between low-SNR neural signals and highly structured vision--language spaces, making direct cross-modal alignment unstable. To address this, we propose STAMBRIDGE, a versatile two-stage framework that sequentially tackles feature conditioning and cross-modal alignment. First, we introduce a Spectral-Temporal Amplitude-aware Modulation (STAM) to extract well-conditioned EEG representations. By replacing hard frequency masking with amplitude-derived soft channel weighting and multi-scale temporal convolutions, STAM explicitly preserves frequency-aware transients while reducing the risk of time-domain ringing artifacts. Building upon these robust neural features, we further introduce a model-agnostic Mid-Feature Semantic Bridge (MFSB) that constructs a regularized intermediate space through directed cross-modal interactions, enabling staged distillation and more stable semantic alignment. Experiments on the THINGS-EEG benchmark show competitive 200-way zero-shot retrieval performance, with 34.50\% Top-1 and 65.95\% Top-5 accuracy. In addition, embeddings learned by STAMBRIDGE produce semantically coherent image reconstructions with a diffusion model, demonstrating robust EEG-to-vision semantic alignment. The code is available at: https://github.com/thabeatmjh/STAMBRIDGE.

In a landscape of high-performance distributed ML systems, JAX has emerged as a framework of choice. However, JAX's modular design philosophy leaves it without a standardized checkpointing solution. In this paper, we introduce Orbax, a modular, JAX-native checkpointing library that abstracts the complexities of distributed accelerator systems while also providing flexibility for user-friendly checkpoint manipulations throughout the ML model lifecycle. We demonstrate performance exceeding comparable PyTorch competitors by up to 3.5$\times$ for saving and 2$\times$ for loading. The library is available at https://github.com/google/orbax.

Computer-aided design (CAD) is the backbone of modern industrial design, yet learned CAD generators still fall short of real engineering pipelines: they neither iterate like engineers nor evaluate what engineering requires. Prior work has treated CAD generation as two disjoint steps, part synthesis and assembly, where the former is graded by proximity to a gold reference and the latter, when handled at all, is reduced to a separate constraint solving step. In this work, we introduce a more industry-native task formulation that requires a model to produce a fully assembled multi-part STEP file from a free-form engineering brief, which is then validated via finite element analysis (FEA). FEA validation reveals that Codex (GPT-5.5) and Claude Code (Opus-4.7) agents do not produce a single strict-passing artifact in the main first-attempt sweep, with the best configuration meeting only about 20% of typed requirements on average. Moreover, we introduce two additional supervision signals, a novel text-only blueprint schema and a 21-view image renderer that aids the agent's visual inspection, that better align the generation loop with how engineers iterate in practice. On S2O and Fusion360, the same feedback tools improve geometric reconstruction, with GPT-5.5/xhigh rising from 0.444 to 0.592 Box-IoU on S2O and from 0.397 to 0.505 on Fusion360. Together these signals move CAD programs toward artifacts that are not only visually plausible but also checked against physical and structural requirements.

High-quality training datasets are essential for the performance of neural networks. However, the audio domain still lacks a large-scale, strongly-labeled, and single-source sound event dataset. The FSD50K dataset, despite being relatively large and open, contains a considerable fraction of multi-source samples where background interference or overlapping events could limit the usefulness of the data. To address this challenge, we introduce a data curation framework designed for large-scale open audio corpora. Our approach leverages a generative diffusion model to synthesize clean single-class events to construct controlled noisy mixtures for supervision. We subsequently employ a pre-trained audio encoder coupled with a discriminative classifier to automatically identify and filter out multi-source samples. Experiments show that our framework achieves strong performance on a human expert-curated test set. Finally, we release FSD50K-Solo, a model-curated subset of FSD50K containing single-source audio samples identified by our method. Beyond FSD50K, our method establishes a scalable paradigm for curating open source audio corpora.

Artificial intelligence has advanced rapidly in biomedicine through large-scale multimodal data integration, enabling increasingly accurate prediction of clinical outcomes and patient stratification. These systems, however, remain fundamentally observational: they learn statistical associations from historical data and operate within previously observed biological and clinical states, limiting their ability to generalize to novel therapies or unobserved interventions. We argue that AI in biomedicine is undergoing a structural transition. As biomedical decision-making increasingly depends on reasoning about intervention rather than extrapolation from past observations, predictive architectures become structurally insufficient. Systems that learn from historical data cannot, by construction, represent how biological systems evolve under perturbation, and therefore cannot reliably support decision-making in the presence of novel interventions. We introduce a conceptual framework distinguishing observational and interventional intelligence and define disease-level models as systems that explicitly represent the state, dynamics, and intervention response of biological processes. These models enable a shift from inference to simulation -- reasoning about what will happen under intervention rather than what is likely based on the past. This transition also implies a shift in where value is created: from data processing and prediction toward systems that support and define decision-making under intervention. It follows directly from the structure of biomedical decision-making and defines the next stage of AI in medicine. Systems that cannot model intervention will be structurally excluded from decision-making.

Score-based diffusion models demonstrate superior performance in generative tasks but encounter fundamental bottlenecks in inverse problems due to the analytical intractability of the time-dependent likelihood score. To bridge this gap, we propose a novel proximal-based generative modeling (PGM) framework that rigorously circumvents explicit likelihood evaluation. Our framework is built upon a theoretical equivalence between Gaussian convolution in diffusion processes and Moreau-Yosida regularization in nonsmooth optimization. This enables a new sampling mechanism driven by the proposed Moreau score, which admits a closed-form expression via proximal operators. Moreover, we introduce Moreau score matching to learn the proximal operators that rely solely on samples drawn from the prior distribution. Theoretically, PGM eliminates the early-stopping bias inherent in the score-based diffusion model and achieves non-asymptotic convergence. Experiments demonstrate that PGM significantly surpasses state-of-the-art methods in reconstruction quality and sampling time.

Reusable skills are becoming a common interface for extending large language model agents, packaging procedural guidance with access to files, tools, memory, and execution environments. However, this modularity introduces attack surfaces that are largely missed by existing safety evaluations: even when the user request is benign, unsafe influence may reside in skill guidance, local artifacts, or execution-environment files that steer the agent toward unsafe actions. We present SkillSafetyBench, a runnable benchmark for evaluating such skill-mediated safety failures. SkillSafetyBench includes 155 adversarial cases across 47 tasks, 6 risk domains, and 30 safety categories, each evaluated with a case-specific rule-based verifier. Experiments with multiple CLI agents and model backends show that non-user attacks can consistently induce unsafe behavior, with distinct failure patterns across domains, attack methods, and scaffold-model pairings. Our findings suggest that agent safety depends not only on model-level alignment, but also on how agents interpret skills, trust workflow context, and act through executable environments.

Every major benchmark for LLM memory systems, LoCoMo foremost, measures whether a model answered correctly, not whether the memory system retrieved correctly. A system returning its entire belief store achieves recall of 1.0 and passes answer-quality evaluation. This is the difference between a unit test and an integration test: retrieval quality must be measured in isolation from the generative model it feeds into, and no existing benchmark does this. We demonstrate that this failure persists even when entity extraction is entirely faithful. Memory baselines achieve mean retrieval precision of just 0.05 to 0.08 on cases referencing their own extractions. The failure is structural: cosine similarity over a domain-specific corpus cannot discriminate relevant beliefs from semantically proximate ones, an invariance confirmed across a 20x range in embedding model scale. Multi-turn evaluation surfaces a compounding failure; after topic drift, comparison systems allow semantic mass to bleed across turns, yielding high drift scores on re-entry. Single-turn metrics conceal this cost: Hindsight reports sub-700ms single-turn latency but exceeds 2,700ms mean per session turn, with p95 above 6,000ms. Under LLM-as-a-Judge evaluation, these failures remain invisible. We present two contributions: PrecisionMemBench, an 89-case benchmark measuring retrieval precision independently of generative models across diverse scope, mutation, and isolation assertions; and Tenure, a local-first structured belief store using multi-path BM25 with analyzer asymmetry, differential boosting, and hard scope isolation. Tenure passes 89/89 cases with mean precision 1.0 and sub-15ms retrieval latency. Comparison providers perform worse than the raw vector baseline they are built on, with zero active retrieval passes and ingestion costs of 98 to 897 seconds, failures that answer-quality benchmarks cannot detect.

Modern astrophysical studies rely heavily on complex data analysis pipelines; however, published descriptions often lack the detail required for computational reproducibility. In this work, we present an information-theoretic framework to quantify how effectively a method can be reconstructed from its written description. By treating algorithmic reconstruction as a probability distribution generated by Large Language Models (LLMs), we utilize Shannon entropy and Jensen-Shannon divergence to measure how strongly text constrains the hypothesis space of valid implementations. We demonstrate this approach through a case study of Trans-Neptunian Object (TNO) spectral reconstruction from sparse photometry. By prompting frontier LLMs with varying levels of manuscript text (Title, Abstract, and Methods), we find that while increasing text successfully clarifies the overall algorithmic structure, it fails to eliminate variance at the implementation level. This persistent variance establishes an "entropy floor," demonstrating that multiple divergent implementations remain consistent with explicit instructions. To evaluate practical reproducibility, we convert these reconstructed algorithms into executable pipelines. Our results reveal that, while LLMs easily recover core functional methodologies, they systematically fail to infer the tacit expert knowledge required for strict scientific calibration. This pilot study demonstrates that LLMs can be repurposed as a zero-shot diagnostic tool to audit methodological transparency, helping authors identify missing structural constraints and preserve scientific integrity in an era of automated research.

Change of measure inequalities translate divergences between probability measures into explicit bounds on event probabilities, and play an important role in deriving probabilistic guarantees in learning theory, information theory, and statistics. We propose novel change of measure inequalities via a unified framework based on the data processing inequality, which is surprisingly elementary yet powerful enough to yield novel, tighter inequalities. We provide change of measure inequalities in terms of a broad family of information measures, including $f$-divergences (with Kullback-Leibler divergence and $χ^2$-divergence as special cases), Rényi divergence, and $α$-mutual information (with maximal leakage as a special case). We apply these results to generalization error analysis, PAC-Bayesian theory, differential privacy, and data memorization, obtaining stronger guarantees while recovering best-known results through simplified analyses.

While the ecosystem of Lean and Mathlib has enjoyed celebrated success in formal mathematical reasoning with the help of large language models (LLMs), the absence of many folklore lemmas in Mathlib remains a persistent barrier that limits Lean's usability as an everyday tool for mathematicians like \LaTeX{} or Maple. To address this, we introduce MathlibLemma, a modular LLM-based pipeline for automated folklore-lemma mining: the discovery, formalization, and proving of reusable intermediate facts that mathematicians often take for granted but that are not always present in formal libraries. At its core, MathlibLemma proactively mines the missing connective tissue of mathematics. The pipeline produces a verified library of folklore-style lemmas, including 1,506 Lean-checked proofs that pass a proof-bypass screen; a small curated pilot subset has also been merged into Mathlib, providing external evidence that selected outputs can meet expert library standards. Leveraging this pipeline, we further construct the MathlibLemma benchmark, a suite of 4,028 non-trivial type-checked Lean statements spanning a broad range of mathematical domains. By transforming the role of LLMs from passive consumers to active contributors, this work takes a step toward AI-assisted expansion of formal mathematical libraries.

Training a language model on data scattered across bandwidth-limited nodes that cannot be centralized is a setting that arises in clinical networks, enterprise knowledge bases, and scientific consortia. We study the regime in which data must remain distributed across nodes, and ask what statistical guarantees are in principle achievable under explicit bandwidth budgets; we aim to characterize what is provably possible, not to demonstrate a deployment-ready system. Existing theory treats either training-time consistency or inference-time calibration in isolation, and no prior work makes bandwidth a first-class statistical parameter. We analyze two protocols, Federated Probe-Logit Distillation (FPLD) for training and Federated Conformal RAG (FC-RAG) for inference, as the analytical vehicles for our results. Our first main result is an explicit high-probability KL-consistency rate for FPLD with simultaneous dependence on node count $K$, per-node sample size $n$, quantization budget $B$, probe-set size $m$, and vocabulary size $V$; bandwidth enters only through an exponentially vanishing quantization term. Our second main result is a distribution-free marginal-coverage bound for FC-RAG, whose novel retrieval-bandwidth slack $Δ_{\mathrm{RAG}} = f_{\max}\sqrt{K^{-2}\sum_i v(B_i)}$ makes per-node retrieval bandwidth a first-class statistical parameter, with arithmetic aggregation across $K$ nodes shrinking the slack as $K^{-1/2}$ in the per-node-uniform regime. A Pinsker-type corollary composes the two bounds into an end-to-end coverage guarantee. Synthetic experiments verify the predicted scaling along the bounds' parameters; small-scale experiments on a GPT-2 testbed illustrate that the qualitative bandwidth-accuracy tradeoff survives on a real language model. A deployment-scale empirical evaluation is out of scope.