PyTorchKR Papers

최신 논문

총 568,830편
태그·기간 등 상세 필터

카테고리 탐색

arXiv:2611.03721v2cs.AI

Counterexample-driven autonomous research with application to excess radiated energy in a Z-pinch

Recent breakthroughs in artificial-intelligence-assisted mathematics show that open-ended machine search can outperform human mathematicians, yet transferring this paradigm to physics requires refactoring the research workflow. We introduce Simjecture, an open-source framework in which language-model agents pursue human-defined physical conjectures. The framework tackles two central challenges in this transfer: first, because physics is empirical and no model is final, it structures inquiry around counterexample-driven falsification and prospective hypothesis repair; second, because physical theories must interact with the real world, it uses computational physics as an executable testbed while separating the qualification of numerical instruments from scientific claim review so solvers and diagnostics can be adapted within a recorded investigation. We apply the framework in a 136.3 h investigation to tackle the long-standing problem of radiated energy exceeding the imploding-shell kinetic energy in a Z-pinch. The investigation reveals competing effects of energy conversion, retention, and release timing. Stronger coherent helical deformation increases resistive conversion while reducing radiation and increasing retained energy. In a matched fine-grid test, broadening the trailing-plasma profile delays the later radiation maximum by 1.80 ns and recovers an early 153 J bolometric deficit to within 0.08% by 44 ns, with a spectral response that shifts during the pulse. These results provide new insight into energy conversion in high-energy-density plasma physics and demonstrate a new paradigm for physics research.

arXiv:2611.03698v2cs.AI

Training LLM Agents to Automate LLM Post-Training with RSI-Dojo

Recursive Self-Improvement (RSI) aims to enable AI systems to autonomously build and improve AI systems themselves, including models, harnesses, and underlying infrastructure. Within this broader agenda, a central capability is enabling agents to train and improve base models. However, existing studies on automated model training primarily focused on workflow- or harness-level optimizations, rather than internalizing the ability to conduct post-training into the model itself. Developing this capability requires executable model training environments that can produce high-quality trajectories demonstrating how to select data, train models, diagnose failures, and iteratively improve results. In this paper, we introduce RSI-Dojo, a collection of executable model training environments designed to teach LLM agents to conduct post-training. To construct RSI-Dojo, we develop RSI-Craft, an automatic pipeline that converts human research artifacts into executable model training environments. RSI-Craft automatically collects research papers from top venues, instantiates datasets and metrics into a unified interface, and reproduces evaluation dependencies. It then filters tasks by data size, execution reliability, and discriminability, yielding RSI-Dojo with 281 seed model training tasks. Building on RSI-Dojo, we employ a trajectory collection method with an improved harness to collect high-quality model-training trajectories from frontier models, then retain examples through rubric-based and outcome-based filtering. We fine-tune Qwen3.5-9B and Qwen3.6-35B-A3B on these trajectories. The resulting RSI-Dojo-9B and RSI-Dojo-35B-A3B improve over their base models by 8.74 and 16.87 PostTrainBench points, respectively. RSI-Dojo-35B-A3B reaches 29.95 on PostTrainBench and 53.29 on PostTrainBench-Lite, matching the performance of frontier models such as GLM-5.2 and GPT 5.5.

arXiv:2611.03601v2cs.AI

DiabAgent: Toward Agentic Physiological Control

As pretraining scales, medical language models have advanced rapidly in clinical question answering. Their ability to achieve effective physiological control through sustained intervention still lacks systematic quantitative evaluation. Type 1 diabetes provides a concrete setting: impaired insulin secretion creates dependence on exogenous insulin, with dosing continually adjusted using information about meals, exercise, illness, and glucose measurements. We introduce DiabAgent, a research suite for agentic physiological control comprising the Agentic Physiological Control (APC) framework, the DiabetSim simulator, and the SweetBench benchmark. APC controls glucose through medical-device tools, using ongoing observations to infer individual states, accumulate personal experience, and choose treatment actions. We build DiabetSim to simulate different patients, events, and device conditions for scalable virtual testing and training. Using DiabetSim, we construct SweetBench by instantiating 32 scenario designs as 212 glucose-control tasks. With Claude Opus 5.5 and a fixed decision interval of one hour, APC achieves a COGI of 73.5, compared with 71.2 for the open-source closed-loop baseline OpenAPS. Ablations further show that individual adaptation and contextual information improve glucose control. Combining high-frequency rule-based control with lower-frequency agent decisions shows potential for balancing control performance and computational cost. Our work provides a framework, simulator, and evaluation foundation for agents to participate in human physiological control.

arXiv:2611.03555v2cs.CR

MCP Tool Poisoning in Agentic AI Frameworks: A Cross-Framework Measurement Study

The Model Context Protocol (MCP) lets agentic AI frameworks connect language models to external tool providers, but it carries attacker-controlled content (tool descriptions, schema fields, and response data) into model context without validation. Prior benchmarks measure model compliance with poisoned instructions without confirming real execution. We present the first cross-framework measurement of MCP tool poisoning with real tool execution. A 25-class root-cause taxonomy unifies prior classifications, and the open-source tool mcpvenom drives a 9,720-cell susceptibility matrix across 12 frameworks, 18 models, and 45 attack scenarios, with a dual oracle that confirms exfiltration via canary HTTP receipt. The overall attack success rate is 22.9% (2,113 vulnerable cells). Of these, 64.9% dispatch an attacker tool call, 23.0% contact the attacker endpoint, and 12.1% verifiably exfiltrate a secret. Resistance is model-level: nine core frameworks cluster within 23 to 26% (chi-squared = 3.30, p = 0.91), and no framework implements intentional defense. Incidental framework properties mitigate in specific cases, and a regex description scanner rejects 20 of 45 scenarios at a 7.2% false-positive rate, so metadata scanning alone is a partial control. Model ASR ranges from 0.4% to 49.7%. Claude Opus 4.8 is vulnerable in 8 of 530 cells while issuing 223 explicit refusals. A code-level audit finds no intentional defense across twelve frameworks, and all five coordinated disclosures across four vendors were rejected.

arXiv:2611.03523v2cs.SE

DistDebug-Bench: Can LLM Agents Diagnose Bugs in Distributed Systems?

LLM agents show strong potential in software engineering tasks such as competitive programming and code synthesis. Existing benchmarks, however, largely focus on small-scale settings such as self-contained Python modules, leaving distributed systems, an essential class of production software, under-examined. We introduce DistDebug-Bench, a benchmark for evaluating agents' ability to diagnose complex-system failures at the code level. It curates 120 high-quality real-world bugs from 11 widely used distributed systems, spanning 5 programming languages and 10 failure patterns. Through an extensive evaluation of state-of-the-art LLMs and agents, we identify substantial gaps between current agents and real-world debugging tasks. A recurring gap is tunnel vision due to a lack of explicit reasoning structure and nondeterminism from inconsistent tool-use scope. The benchmark and evaluator are available at https://github.com/OrderLab/distdebug-bench, with agent traces and a leaderboard at https://distdebug.com.

arXiv:2611.03486v2cs.LG

A Hybrid Visitor Prediction Framework using Interpretable Temporal features, Integrating IQR and Anomaly Detection in Lakes in Nebraska

For effective resource management, infrastructure planning, and sustainable tourism, it is really important to predict accurate visitors. This paper estimates the daily visitation of Nebraska lakes using spatial, basic temporal, and interpretable temporal features. Moreover, we evaluate the base models with multiple metrics. Using the best three models, our method will ensemble hybrid models and choose the best hybrid technique. Further, we investigate the impact of lag features and how the models work with different outlier detection techniques such as Interquartile Range (IQR) and anomaly detection. The models are evaluated using regression metrics (MAE, MSE, RMSE, MAPE, and R2). Experiment results show that, after removing anomalies, MAE, MSE, RMSE, and MAPE are reduced. Adding lag features also shows some positive effects on the results. Moreover, anomaly detection techniques perform better than IQR in improving performance while preserving data. The reported anomaly-based cross-validation results favored the voting ensemble over the compared base models on the evaluated metrics. In anomaly-based cross-validation, VOTE improved prediction errors compared with MLP (256,128,64); improvements varied in the single-split comparison. Specifically, the hybrid model reduced MAE by 9.82%, MSE by 3.12%, RMSE by 1.76%, and MAPE by 6.77%. In the reported cross-validation comparison, the best anomaly-based hybrid model had lower MAE, MSE, RMSE, and MAPE than the best IQR-based hybrid model by 9.81%, 34.30%, 18.99%, and 0.66%, respectively.

arXiv:2611.03412v3cs.AI

DeliverBench: Benchmarking AI Agents on Delivering Realistic Multi-Stage Professional Projects

LLM agents are judged by what they deliver, and in professional work what gets delivered is a project, not a task. A project pays off only if the whole package holds together, and one weak part can limit the value of the rest. Yet existing benchmarks mostly test agents on one task at a time, leaving a gap between the capability we measure and the outcome that makes agents useful. To address this gap, we introduce DeliverBench, which tests whether agents can carry a project from brief to final delivery. Reconstructed from real materials with 13 domain experts, its 11 projects span settings such as financial due diligence, legal contract drafting, and drug safety reporting. They comprise 150 subtasks, with 2-15 subsequent releases of new materials and client feedback per project. Later subtasks build on the agent's own earlier deliverables, so unresolved errors can carry into downstream work. Expert rubrics assess each subtask submission and the final delivery package. Across seven model-harness systems, the best scores 65.0/100 on final delivery and passes 50.0% of subtasks at a 90-point threshold. Together, these results suggest that task-level capability does not simply compose into delivery; delivery also depends on keeping the work coherent as it evolves.

arXiv:2611.03196v2cs.IR

Dhara Bench: An Evidence-Oriented Retrieval Benchmark for Indian Knowledge Systems

Dhara Bench is a retrieval benchmark for Indian Knowledge Systems (IKS), with 325 questions in 12 categories and 60,121 text chunks from 15 source groups. The categories were developed based on 4,500 questions submitted to the Dhara application; the evaluation set combines corpus-aligned application queries with expert-contributed questions. The annotations comprise 7,578 question-chunk relevance links involving 6,847 distinct corpus chunks. For 34 composite and multi-hop questions, 401 conjugate groups specify alternative sets of passages that jointly support each question. Chunk identifiers and hierarchical references connect the judgments to their source texts. The paper describes corpus preparation, annotation, and validation, and reports a fixed BM25 baseline implemented with Pyserini with $k_1=0.9$ and $b=0.4$. Across all questions, BM25 achieves MRR@20 of 0.315, Recall@10 of 0.182, and nDCG@10 of 0.205. On the group-annotated subset, complete-evidence scores are CE_Recall@20 = 0.294 and CE_MRR@20 = 0.141. The category results expose particularly weak verse retrieval, while group scores reveal gaps between finding individual passages and completing their support. The corpus and evaluation code are available at https://github.com/Avinash2236/Dhara-Bench.

arXiv:2611.02686v3cs.CV

Discovering Semantic Granularity in Images through Clustering of Pretrained Visual Representations

Semantic granularity, which refers to the levels at which visual representations organize image content into meaningful entities, is fundamental to image understanding. However, whether pretrained visual representations inherently encode semantic structures at multiple granularities remains poorly understood. This paper presents a representation-driven approach for discovering semantically meaningful image regions at different granularities through clustering of pretrained visual representations without task-specific training. Specifically, we concatenate feature maps from appropriately deep layers of a pretrained visual model and apply a singular value decomposition (SVD)-based projection to derive a pixel-wise principal representation map (PRM) tensor. The resulting PRM tensor captures latent semantic structures embedded in pretrained visual representations, making them accessible for subsequent semantic clustering. We then introduce the semantic continuity ratio (SCR) and analyze the signal-to-noise ratio (SNR) of SCR profiles to assess the suitability of PRM tensors for semantic granularity discovery. Our results indicate that PRM tensors exhibiting a higher proportion of relatively high SNR values over an initial range of cluster numbers tend to yield more semantically meaningful regions across multiple granularities. Furthermore, once a suitable PRM tensor is identified through SNR-profile analysis, pixel-wise hierarchical clustering enables the level of semantic granularity to be controlled by varying the number of clusters. We evaluate the proposed approach on the COCO-2017 and VOC-2012 validation datasets using Hungarian-matched intersection over union (HIoU). The results suggest competitive segmentation performance among the compared methods, supporting the feasibility of the proposed approach.

arXiv:2611.02324v2cs.CV

The DynaCell Evaluation Suite for Dynamic 3D Virtual Staining of Live Cells

Understanding how cells respond to perturbations over time is central to cell biology and drug discovery. Live-cell imaging captures the dynamics of perturbation-induced phenotypes in individual cells. While fluorescence imaging can visualize specific molecular phenotypes, labeling and repeated illumination often affect cell health. Virtual staining offers an alternative by predicting these fluorescence images from images acquired without fluorescent labels. Although many virtual staining models have been developed, their performance on volumetric time-lapse data remains poorly characterized. We address this gap with an evaluation suite for 3D time-lapse virtual staining of live cells. The suite combines paired 3D time-lapse images acquired with and without fluorescent labels, a set of baseline models, and three metric tiers measuring pixel-level fidelity, organelle segmentation performance, and single-cell phenotype similarity. Using this evaluation suite, we study metric and model behavior across 3 cell types, 2 microscopes, 4 organelles, and 3 perturbation states. We find that regression baselines better predict the spatial localization of target organelles, whereas the generative baseline better captures the distribution of phenotypes across cells. These differences expose trade-offs between structural and phenotypic fidelity that are obscured by single-metric evaluation and suggest how virtual staining models can be developed for diverse downstream tasks such as localization, segmentation, tracking, and phenotypic profiling. We release the curated datasets, baseline checkpoints, and evaluation code as part of DynaCell, a broader framework for image-based profiling of cell-state dynamics.

arXiv:2611.02267v2cs.LG

Overcoming Demand Censoring through Randomization for Risk-Sensitive Newsvendor Learning

We study online newsvendor learning with bounded demand, censored observations, and convex nonlinear shortage costs. Because stockouts can conceal the shortage magnitudes needed by gradient-based algorithms, we use randomized ordering to construct unbiased subgradient estimates. We characterize admissible randomization distributions within this framework and develop an online inventory policy that combines the resulting subgradient estimates with gradient-based updates. Our regret analysis quantifies the cost of randomization via the estimation variance and an additional loss term. We characterize the randomization distributions that minimize this objective. For piecewise-linear shortage costs, the optimal distribution is computed through a linear-time recursion; for costs with bounded curvature density, it is characterized by a scalar ordinary differential equation. The policy achieves expected regret of $O(T^{2/3})$ against the best fixed order in hindsight over $T$ periods. Under independent demand and suitable curvature of expected losses, the bound improves to $O(\sqrt{T})$. Matching minimax lower bounds establish rate optimality over the corresponding classes with a fixed marginal-cost bound. Both rates extend to nonperishable inventory under uniformly positive conditional expected demand. Numerical experiments on synthetic and retail data show improvements over alternative policies using censored observations.

arXiv:2611.02076v2cs.CV

Scene-Adaptive Readout and Decoding in Diffusion Driving VLAs

Diffusion driving vision-language-action models repeatedly execute the backbone at every denoising stage, incurring substantial inference cost, while it remains unclear whether all scenes require the same decoding computation. To distinguish when a trajectory becomes readable from how much further refinement it requires, we analyze trajectory readout across network depth and decoding progress. We identify a depth-decoding dissociation in which late-layer representations already support useful trajectory readout after the first forward sweep, whereas early-layer representations benefit more from continued decoding. We further find that both the preferred readout layer and decoding depth vary across scenes. Based on these observations, we propose a scene-adaptive readout and decoding framework. A scene-conditioned readout MoE adaptively weights and combines trajectory predictions from layer-specific late-layer representations after a single backbone sweep, while an adaptive decoding policy applies additional native refinement only when required by the scene. We evaluate our framework on two diffusion driving VLAs across the Waymo E2E and nuScenes benchmarks. Our method improves validation RFS by up to 0.56 while also improving test performance, and achieves an end-to-end speedup of up to $6.0\times$ on Waymo E2E, while consistently reducing trajectory error on nuScenes. These results demonstrate that scene-level readout and decoding allocation improves planning quality while avoiding unnecessary decoding. Our code is available at https://github.com/sjyu001/SARD

arXiv:2611.01835v2cs.LG

ReFrame: Silence-Aware Compute Reclamation for Full-Duplex Speech Models

Full-duplex speech models execute dense model inference for every audio frame throughout a live interaction. This uniform framewise schedule creates sustained GPU load across dialogue states whose computation has strong, reusable temporal structure. ReFrame formulates full-duplex inference as a silence-aware computation problem. Mutual Silence Speculation (MSS) operates when both participants are silent: it evaluates multi-frame Temporal continuations, verifies text and hidden-state stability, and applies batched Mimi audio-token correction. Listening Frame Folding (LFF) operates while the user speaks and the agent listens: it processes already-arrived user frames through parallel Temporal execution and frame-batched audio-token generation while accepting only the verified prefix. Frame-Budget Reinvestment (FBR) assigns the released frame intervals to resumable external retrieval and language-model workers, then injects timely references through Moshi's conditioning interface. We instantiate ReFrame on five full-duplex models spanning Moshi-family, phase-structured, and flattened autoregressive generation. Across the complete 1,225-conversation FDBench, ReFrame reduces measured GPU execution time by 18.80-31.01% and retains 96.92-100% non-PAD text agreement on five different models. On HaluEval tests, FBR increases the number of correct generated spoken answers from 86 to 205 under the same auxiliary-service budget. ReFrame turns recurring dialogue structure into lower inference cost and higher end-to-end service capability.

arXiv:2611.01728v2cs.MA

MAC-Bench: Benchmarking Multi-Agent Collaboration via Interconnected Tasks

Large language model (LLM)-based multi-agent systems (MAS) tackle challenging tasks by dividing work among agents and coordinating their contributions toward shared objectives. Evaluating these systems therefore requires benchmarks that assess not only task-solving performance but also the ability to handle interdependent work. However, existing benchmarks often emphasize task difficulty without substantial collaboration requirements or evaluate only a limited range of predefined collaboration patterns. In this paper, we introduce MAC-Bench, a comprehensive benchmark for evaluating multi-agent collaboration through interconnected tasks. Rather than imposing predefined agent roles or communication protocols, MAC-Bench embeds collaboration requirements directly into task dependencies and shared-resource constraints. Specifically, we mine reusable collaboration structures from selected benchmark tasks at two levels. Local collaboration patterns capture recurring interactions among subtasks, while global task structures characterize how these interactions contribute to complete workflows. Through coarse-to-fine composition, these structures are recombined into task graphs combining different collaboration requirements. Tasks are instantiated in executable environments with automated verification of partial progress and complete task resolution.

arXiv:2611.01563v2cs.AI

CHATDIFY: Automating End-to-End Dify Workflow Synthesis from Natural Language via Multi-Agent Collaboration

The proliferation of Large Language Models has catalyzed powerful application development platforms like Dify, which enable users to build sophisticated LLM-powered workflows through modular node orchestration. However, manually configuring these workflows requires deep platform expertise and remains time-consuming and error-prone, creating a significant barrier for non-expert users. To address this challenge, we propose \textsc{ChatDify}, a novel multi-agent framework that automates end-to-end Dify workflow synthesis from natural language descriptions. \textsc{ChatDify} employs an Actor-Critic architecture comprising three specialized agents: the \textit{Planner Agent} that decomposes user requests into structured execution plans, the \textit{Executor Agent} that performs context-aware parameter synthesis and schema-augmented DSL compilation, and the \textit{Reviewer Agent} that ensures reliability through hybrid verification with closed-loop refinement. We evaluate \textsc{ChatDify} on a curated benchmark of 100 production-inspired tasks across three state-of-the-art LLM backends, achieving 0.96, 0.91, and 0.75 success rates with Qwen-3.7, GLM-5.1, and DeepSeek-v4, respectively. These results demonstrate its robustness and practical applicability for lowering the technical barrier to creating sophisticated LLM applications.

arXiv:2611.01472v2cs.LG

Causal Optimal-Transport Attention: Positional Priors and Length Extrapolation

Causal masking obstructs the column balancing used by optimal-transport (OT) attention. We study an unbalanced OT operator that uses only current and earlier rows and characterize its induced positional prior. For bounded scores and any fixed balancing strength below one, its fixed point assigns vanishing mass to every fixed window as context grows. With constant scores, the prior has an exact scale-invariant form. We classify positional encodings by whether they prevent this dilution. We also derive finite-length window-mass bounds for log-distance priors with exponents near one. Preregistered experiments use 25M- and 100M-parameter language models. The OT operator improves loss extrapolation over softmax, while exploratory comparisons favor an explicit distance bias at a fraction of the cost. Changing only the bias beyond the training length moves long-context loss in the predicted direction on all 30 tested checkpoints. During training, learned exponents approach one even when constrained above it. A margin chosen from prior calculations controls distant-mass growth in the tested setting.

arXiv:2611.01370v2cs.CL

XDecideBench: Evaluating Typed Decision Models and Confidence Policies Across Languages

Typed Decision Models such as Jev, Clef and pplx-decider return a probability for each answer an application provides. Applications use them in many languages, but are compared mostly in English. We introduce XDecideBench, a multilingual benchmark of nine suites in 122 languages. On our twelve-language suite, three open models fine-tuned for typed decisions beat their Qwen base models in every language. On new test data with local base models, their accuracy gain is larger in languages with less web text than in those with more. They fix a similar share of base-model errors in both, so the larger gain may reflect weaker base models. The best model on the twelve-language suite is best on only two other suites. The best English model on each suite, chosen in hindsight, is significantly beaten in about a quarter of suite and language combinations. On Belebele, Decision-1 is best in English but falls far behind Jev in Burmese. A confidence threshold fitted on English to allow 5% errors gives a higher error rate on XNLI, even in German and French. XDecideBench shows that results from one suite or from English are a poor guide to how a model and its threshold behave in other languages.

arXiv:2611.01077v2cs.LG

Multi-Turn Semantic Caching for Low-Cost LLM Serving

Semantic caching lowers the cost of serving large language models (LLMs) by reusing stored responses for similar requests. In multi-turn conversations, whether a cached response is acceptable depends on both conversations' histories, which embedding similarity of requests and histories predicts poorly (AUC 0.62--0.74 on five workloads). We therefore learn the \emph{reuse risk}, the probability that a cached response is unacceptable, from judge feedback, and choose a budgeted set of conversation states to cache by weighing the generation cost a reuse avoids against a price for unacceptable reuse. We propose CUCB-SC-MT, which learns a cache from logged conversations with a suboptimality guarantee, and CLCB-LS-MT, which learns online and has $\tilde O(\sqrt N)$ regret over $N$ requests when conversation states recur. On math-tutoring (SocraTeach) and movie-ticketing (Taskmaster-3) conversations, both achieve lower loss than prior semantic caching approaches: ContextCache, multi-turn variants of ElastiCache and vCache, and single-turn semantic caching at every cache budget. With 1,000 cache entries, CLCB-LS-MT avoids 35--41\% of the generation cost through acceptable reuses, against at most 22--25\% for the baselines, and lowers estimated serving latency by 19--35\%.

arXiv:2611.00683v2cs.CV

Content-Preserving Principal Feature Blending for Training-Free Style Personalization

Reference-based style personalization should reproduce the appearance of a style reference while preserving the subject specified by a content prompt. In visual autoregressive models, dominant reference features can also carry object and spatial information, causing content leakage. We propose Content-Preserving Principal Feature Blending (CoPFB), a training-free method that estimates a low-rank content basis from a clean generation branch and removes the content-aligned component of the style residual. CoPFB applies rank-one blending at the first two autoregressive scales, followed by weaker content-projected refinement at the next three scales. On 1,900 combinations of Parti content prompts and FineStyle reference images, CoPFB improves text alignment from 0.293 to 0.297 and the harmonic score from 0.391 to 0.393 over PFB+SAC, while style similarity decreases from 0.590 to 0.577. CoPFB requires no fine-tuning or learnable parameters.

arXiv:2611.00627v2cs.AI

Autonomous Analysis of Thermoelectric Materials Big Data by Multi-Agentic AI and Construction of a Knowledge Base and Specialized AI

Large databases of paper-derived experimental materials data hide essential information on materials. Their analysis is demanding, large-scale and complex, making manual knowledge extraction very difficult. To address this challenge, we developed Multi-Agentic AI (MAA) for automated, autonomous analysis of materials experimental big data and a specialized AI built on the obtained knowledge. MAA's nine role-specific AI agents, under predefined instructions and quality checks, autonomously perform data extraction, compositional feature generation, machine-learning modeling, feature-importance analysis and knowledge-base recording, iterating while inheriting previous knowledge. Delegating numerical processing to code makes large-scale data tractable; recording every choice and result enables later tracing to sources. In four thermoelectric-data experiments varying theme and cycle counts, the knowledge graph grew: knowledge clustered by theme and vocabulary was shared across themes. The specialized AI, searching and reading these records, answered materials questions citing sources. Retrospective verification also revealed errors, traced through the records to their sources.