Review
-
QuranicMMLU: A Cognitively-Aware Benchmark for Evaluating Generative AI Solutions on Quranic Linguistic Knowledge
2026-09-18 17:30:27Rawan El Ghali, Umm Kulsoom, Anas Madkoor, Dima Faris Alsaudi, Roaa Abdelmagid, Roaa Ibrahim, Raghad Mousa, Hamza Aljaji, Abdullah Khanafer, Abdallah Alkanani, Salah Feras Alali, Rawan Khaled Mohamed, Ehsaneddin AsgariAbstract
We introduce QuranicMMLU, a benchmark for evaluating generative AI on Quranic Arabic across multiple dimensions of linguistic complexity. Existing Quranic benchmarks center on general question answering and semantic retrieval, without probing specific linguistic competencies or stratifying by cognitive demand and verse difficulty. We construct a five-pillar Quranic taxonomy spanning Phonology, Morphology, Syntax, Semantics, and Pragmatics, with 31 leaves covering phenomena from tajwīd and root-and-pattern morphology to occasions of revelation and inter-surah coherence. For each leaf we generate questions stratified by Bloom's cognitive level and verse perplexity, then have LLM as a judge to independently answer and score every item and route the annotations to manual review. The resulting dataset comprises 980 human-reviewed questions, each issued in both open-ended and multiple-choice form. We benchmark 12 systems on these items and find that the Islamic-specialized model leads, yet every system scores higher on multiple-choice accuracy (average 84%) than open-ended answer quality (average 60%): the two rankings agree closely (Kendall's {\tau}=0.73), but multiple-choice scoring hides failures that surface only once answer choices are removed. QuranicMMLU thus offers a rigorous, linguistically grounded framework for evaluating Arabic NLP in the Quranic domain.
Abstract (translated)
URL
https://arxiv.org/abs/2609.22038
PDF
-
RecreationWorld: Scalable and Verifiable Environments for Hybrid Computer-Use Agents
2026-09-18 17:00:56Shuai Bai, Jiayong Deng, Yikun Fu, Chang Gao, Xuhao Hu, Mianqiu Huang, Yizhen Jiang, Yuheng Jing, Dehui Kong, Keliang Li, Ning Li, Wanli Li, Dayiheng Liu, Dunjie Lu, Changwei Luo, Que Shen, Zheyuan Wang, Zijian Wang, Jie Wu, Gao Wu, Zhihui Xie, Rui Xie, Haiyang Xu, An Yang, Jiakang Yuan, Yanming Zhang, Jiajun Zhang, Xi Zhang, Zhenru Zhang, Zhuo Zhen, Mingkang Zhu, Bowen ZhouAbstract
Computer-use agents (CUAs) have advanced along two separate lines: graphical interaction and software development through code and the command line. Real digital work requires both, interleaved rather than stacked end to end. We study hybrid CUAs that autonomously decide when to explore an interface, implement software, and run and visually verify their artifacts. We introduce RecreationWorld, a five-platform framework built around recreation: given a running reference, an agent must discover its behavior and build a faithful implementation with no prescribed workflow. RecreationWorld provides reproducible environments on Ubuntu, macOS, Windows, Android, and Web, plus a unified harness with native GUI control and coding tools. The running reference serves as an oracle for hidden behavioral tests, providing execution-grounded rewards. We scale trajectory generation with high-quality open-source applications. Models trained on these trajectories improve across five out-of-distribution coding and hybrid computer-use benchmarks and more frequently verify their rendered outputs, providing evidence of transfer beyond recreation. For held-out evaluation, we introduce RecreationBench, comprising 250 diverse tasks across domains and platforms. Reference-grounded programmatic and visual assertions cover action-conditioned outcomes at multiple interaction depths; each is validated on the reference and by human reviewers before the suite is frozen for automatic scoring. GPT-6 Astra leads at 58.1% overall, but passes all programmatic tests on just 2.8% of tasks. Agents reproduce static interface structure more reliably than interactions and computed outputs, while generated applications remain smaller and more monolithic than their references. We release the benchmark, environments, and test suites.
Abstract (translated)
URL
https://arxiv.org/abs/2609.22000
PDF
-
Social Influence and the Allocation of Scientific Attention in AI Populations
Abstract
AI systems are becoming participants in the evaluation and use of scientific research. They encounter citation counts, download statistics and lists of popular articles developed around human readers, but the collective consequences of these signals for artificial readers remain uncertain. This paper adapts the Music Lab design to a market for academic attention. In the first experiment, 1,000 AI agents choose papers from the titles and abstracts of all 114 regular research articles published in the American Economic Review in 2025. The experiment has five independent-choice communities and five social-influence communities, each with 100 sequential agents. Only agents in the social-influence condition observe earlier selections within their community. Agents may select any number of papers. Social-information communities select 17.2 percent fewer papers per agent, concentrate their choices more heavily, and collectively cover 73 papers, compared with 90 independently. Between-community variation is greater under social information. In a second experiment with 200 agents across twenty social communities, randomly assigning papers five initial selections raises their subsequent selection rate by 45.55 percentage points (95% CI: 41.20 to 49.90). Choices have modest correspondence with external citations and little correspondence with download counts. The results show how a simple information rule shapes the volume, breadth and distribution of scientific attention in an artificial population.
Abstract (translated)
URL
https://arxiv.org/abs/2609.22408
PDF
-
Morphology-Aware Ambiguity Learning for Wafer Defect Decision Support
2026-09-18 14:56:09Seungjun Chu, Seokhyun ChungAbstract
Wafer map defect recognition is commonly formulated as a fixed-taxonomy classification problem that assigns each wafer to a single defect class. However, some wafers exhibit morphologies near class boundaries, for which forcing a single prediction may be less informative than providing plausible diagnostic alternatives. This paper proposes a morphology-aware ambiguity learning framework that supports three diagnostic actions: automatic single-class diagnosis, assisted diagnosis with two plausible defect classes, and full review. Using the radial, angular, and geometric characteristics of training wafer maps, the framework constructs a class-level ambiguity matrix representing defect-class pairs with similar morphology and plausible diagnostic alternatives. It guides the model to learn plausible alternative classes rather than treating all incorrect classes equally. During inference, the matrix determines whether an uncertain prediction can be represented by a meaningful two-class diagnostic set or should be escalated for full review. Experiments on WM-811K show that the proposed framework outperforms conventional approaches in defect recognition and diagnostic decision support, providing meaningful two-class alternatives while reserving full review for cases with unresolved ambiguity. Illustrative cost analyses further show the potential cost advantage of the proposed routing strategy. The diagnostic behavior of the framework remains consistent across different backbone architectures.
Abstract (translated)
URL
https://arxiv.org/abs/2609.21866
PDF
-
EnterpriseVal: Quantifying the Efficacy, Reliability and Value of Generative AI in the Enterprise
2026-09-18 14:39:26Abbas Raza Ali, Muhammad Ajmal Siddiqui, Moona ZahidAbstract
Frontier language models now produce professional deliverables that expert graders judge to match human work on a substantial share of economically valuable tasks, yet most enterprise GenAI initiatives fail to show a measurable business effect and a large fraction of agentic projects are expected to be cancelled. We argue that this is substantially a measurement problem: public benchmarks answer "what can the model do?", whereas a deployment decision requires "is this workflow fit, reliable, safe and worth scaling - here, on our data, under our controls?". We present EnterpriseVal, a use-case-level evaluation system that closes this gap. It comprises (i) a formal specification of the use case and of the frozen socio-technical configuration under test, model, prompts, retrieval, tools, guardrails and human oversight, with an autonomy level and consequence tier that jointly set the required evaluation intensity; (ii) a metric catalogue spanning fidelity, utility, efficiency, reliability, assurance and oversight; (iii) a grading protocol that scales blinded expert judgement with calibrated LLM-as-judge scoring through prediction-powered inference; (iv) a two-tier threshold gate, stated as an executable algorithm, that maps metric vectors with confidence bounds to REJECT/CONDITIONAL/SCALE decisions; and (v) a value-and-risk model in which the reviewer catch rate is a measured parameter. We report a pilot across three workflows in a global bank. In credit-memo drafting, human-graded citation precision reached 88% and hallucination rate 1.6% for the best model against gates of 70% and 5%; in procedure transformation, analyst refinement effort fell from an estimated 27.4 to 2.9 hours per document. We separate established results, documented pilot evidence, the proposed system and open hypotheses, and specify the experiments required for full validation
Abstract (translated)
URL
https://arxiv.org/abs/2609.21841
PDF
-
An Agentic Just-in-Time Adaptive Intervention System for Personalized Sleep Support: Proof-of-Concept Study with N of 1 Data
Abstract
Background: Just-in-time adaptive interventions (JITAIs) can use behavioral data to adapt support to changing contexts, but many rely on predefined rules and manual configuration. Objective: We developed a proof-of-concept sleep JITAI using an AI agent to review personal data, evaluate reminders, adapt interventions, and record decisions for human review. Methods: Running in Home Assistant on a configurable schedule, the agent follows a reusable skill file to review 30 days of sleep and behavioral data, including physical activity, smartphone use, and bedtime routines, to identify patterns and create or update automated reminders. Results: Initial runs demonstrated technical feasibility, successfully completing data review and intervention decisions while limiting reminders to three per day and saving decision records. Conclusions: Agentic AI may enable flexible, adaptive sleep JITAIs. The architecture supports future comparison with fixed or rulebased interventions, requires human oversight, and could extend to other health behaviors.
Abstract (translated)
URL
https://arxiv.org/abs/2609.21805
PDF
-
From Code Archival to Knowledge Graph: Bridging Software Heritage, COAR Notify and Wikidata
2026-09-18 12:03:19Camillo Carlo Pellizzari di San Girolamo, Francesco TosoniAbstract
Software is a first-class scientific object, yet validated links between source code and the scholarly record remain largely absent from the Linked Open Data (LOD) cloud, isolating archived artefacts from semantic discovery. This paper presents an end-to-end reconciliation pipeline that harvests, validates, and models publication-to-repository pairs from sources where the link between a paper and its source code is explicit and editorially verified: the software-centric journals JOSS, SoftwareX, and IPOL, together with the reproducibility reports of the SIGMOD Availability and Reproducibility Initiative (ARI). This yields a curated corpus of 4,397 $\langle$DOI, repository-URL$\rangle$ pairs. We design two distinct application profiles grounded in Wikidata classes (one for scholarly articles, one for software instances) aligned with the this http URL and CodeMeta vocabularies. This architectural separation enables rule-based reconciliation at two granularities: lightweight, inline publication references or standalone, first-class Wikidata software nodes equipped with SWHIDs, Software Heritage's content-addressed identifiers. A read-only lookup against Wikidata shows that only 82 of the harvested repositories were already modelled there; human-reviewed batches have since created 4{,}182 new software items cross-linked to their articles. We further show that payloads of the emerging COAR Notify protocol, an external effort we do not develop, map natively onto our input format, so the same backend could later serve a live enrichment stream. Our core contribution is a pair of application profiles that turn Wikidata into a connector between the scholarly record and archived source code; we openly release all code, application profiles, and harvested datasets.
Abstract (translated)
URL
https://arxiv.org/abs/2609.21667
PDF
-
Detection is solved, delineation is not: what governs tooth segmentation on panoramic radiographs
2026-09-18 11:08:02Muhammad Rehan, Moaz Amjad, Syed Danial Ahmed, Mariam Adnan, Haider AliAbstract
Automatic tooth segmentation and FDI numbering on panoramic radiographs underpins computer-assisted dental diagnosis, yet which factors govern performance remains unclear. We assemble a corpus of 1,422 panoramic radiographs containing 42,142 expert-delineated tooth polygons across the 32-class FDI taxonomy, annotated by 30 dental practitioners and independently reviewed by two others, and use it to isolate input resolution, architecture and anatomical priors under a single evaluation protocol. First, resolution dominates: across a controlled 640/1024/1280 ablation, mask mAP50-95 rises 0.656 -> 0.710 -> 0.717 while mAP50 stays flat at ~0.982. Both gains are significant under a paired bootstrap over images (p < 0.001, p = 0.024); neither mAP50 change is distinguishable from zero. Added resolution buys boundary precision, not detection. Second, architecture is nearly irrelevant in-domain: a query-based transformer with 2.1x the parameters is statistically equivalent to a one-stage detector (95% CI [-0.0064, +0.0064]), only marginally better under domain shift, 5.5x slower on CPU and not executable under standard ONNX runtimes. Third, three targeted interventions fail: a LoRA-adapted self-supervised encoder underperforms, a promptable foundation segmenter degrades masks by 39%, and globally optimal anatomical label assignment yields +0.0007 despite correcting a constraint violated in 40% of out-of-domain predictions. Zero-shot transfer to an independent multi-centre cohort, verified overlap-free, costs 62% of mask mAP50-95 but only 18% of mAP50, reproducing the dissociation. Decomposing masks along the tooth axis localises the residual error to the apical third. Boundary precision is therefore the binding constraint, and effort is better directed at resolution and acquisition diversity than at architectural novelty.
Abstract (translated)
URL
https://arxiv.org/abs/2609.21628
PDF
-
Quantifying Overclaiming Propensity in Frontier LLM Agents
2026-09-17 17:59:04Nolan Smyth, Yorguin-Jose Mantilla-Ramos, Pascal Jr Tikeng Notsawo, Saskia Helbling, Alberto Tosato, Mohamed Amine Merzouk, Nouha Dziri, Gauthier Gidel, Tommaso TosatoAbstract
Frontier coding agents are increasingly trusted to work autonomously for long periods, yet an agent's final response is often the only account of that work a user sees. We quantify the propensity of frontier agents to \emph{overclaim} task completion, a misrepresentation that can mislead the user. An agent overclaims when its final response contradicts information in its context. This definition requires no inference about intent and is independent of task success. We introduce \emph{OverclaimBench}, an evaluation suite composed of five file-review scenarios, transcript-based coverage measurements, and registered planted defects. We evaluate eight proprietary frontier models in their own production command-line interfaces, and four open-weight models under a single fixed harness on OverclaimBench and find that 1) agents do not read all the files they were asked to review in 67.9\% of runs; 2) among runs where not all files are read, agents are \emph{misleading} 80.4\% of the time (59--96\% per model), either falsely claiming to have read all files or omitting that coverage is incomplete; 3) requiring delegation to subagents increased reading coverage, but among reviews that remained incomplete, a large majority were still misleading; and 4) agents that falsely claimed a complete review missed planted defects at about 1.8 times the rate of agents that read every file, showing that claims of completion can conceal substantive failures. Together, these results show that agents' final responses are not reliable accounts of their actions.
Abstract (translated)
URL
https://arxiv.org/abs/2609.20812
PDF
-
Generative Artificial Intelligence Chatbots for Motivational Interviewing: A Scoping Review From System Design to Intervention Outcomes
2026-09-17 13:43:09Runze Hu, Jingqi Kong, Yang Yang, Yihang Yang, Jingyao Liu, Haizhou Tang, Shanghang Zhang, Zheng LiuAbstract
Motivational interviewing (MI) is a collaborative approach to elicit autonomous motivation for health behavior change. Generative AI (GenAI) offers new ways to deliver MI via conversational systems, but evidence on their design, assessment, and translation into interventions remains fragmented. This scoping review characterized evidence on GenAI-MI chatbots across system design, safety, MI quality, user perceptions, and intervention outcomes. We conducted a PRISMA-ScR scoping review. Nine datasets were searched for studies published or publicly available from January 1, 2015 to June 2, 2026 that used GenAI to generate MI chatbot responses or counselor utterances. Data were extracted using a predefined framework and synthesized descriptively. Forty-seven reports (48 studies) were included. Twenty (41.7%) focused on system design without direct participant use; 28 (58.3%) involved direct interaction. Most systems were text based and disembodied; 23 (47.9%) incorporated dynamic adaptation. Safety measures were unevenly reported. Among studies with direct use, 21/28 (75.0%) reported informed consent or user education. Thirty (62.5%) assessed MI quality, generally suggesting MI-consistent interactions. User perceptions were favorable, especially empathy, usability, helpfulness, and intention to use, though measures were heterogeneous. Eighteen (37.5%) reported intervention outcomes, mostly after a single session. Positive findings were more consistent for short-term motivation than sustained behavioral or functional change. GenAI-MI chatbots can deliver MI-consistent interactions perceived favorably, but evidence for sustained behavioral or functional change is limited. Future research should strengthen runtime safety monitoring, standardize MI quality assessment, and use longer-term comparative designs with behavioral and functional outcomes.
Abstract (translated)
URL
https://arxiv.org/abs/2609.20902
PDF
-
MTVA-Bench: Evaluating the Language Model Inside Cascaded Voice Agents
2026-09-17 12:44:03Pritish Mishra, Ishaan Kumar, Akshat Mandoli, Sudarshan KamathAbstract
Generally, most voice agents are cascaded systems, i.e., an ASR model transcribes the caller's audio, a language model reads the transcript and decides what to say and which backend tools to call, and a TTS model speaks the reply. Nearly all of the decision making happens in the language model, but existing evaluations measure it either too broadly or too narrowly. End-to-end voice benchmarks score the full pipeline, so recognition errors and model errors mix into a single number. LLM benchmarks isolate the model but they do not evaluate what makes real phone calls hard, such as transcription issues, caller's voice being split across messages and the requirement that replies follow the language and script specified. We introduce the Multi-Turn Voice Agent Benchmark (MTVA-Bench), which evaluates the language model on the same conditions it faces inside a cascaded system. The caller is played by an LLM following a set of rubrics and tool calls are answered by a mock backend which responds to the arguments the model actually sent. The benchmark contains 49 agents working across 490 reviewed scenarios and supports 7 languages. Scoring is a combination of deterministic checks on tool calls with two LLM judges, one that scores scenario specific rules and one that grades conversation quality without access to the task. Both judges must cite specific messages from the transcript. Task and conversation scores are weighted equally, since a call can complete its task and still go badly for the caller. In a seven-model study, six of the models select the correct tool within 6.4 points of one another, but their overall scores span 24.4 points. Most of the gap comes from argument values, action ordering, rule compliance, and what the model says around its tool calls.
Abstract (translated)
URL
https://arxiv.org/abs/2609.20152
PDF
-
GPT-6-Astra in a Navigation Workflow: Behavioral Analysis in Zero-Shot Vision-and-Language Navigation in Continuous Environments
2026-09-17 12:14:43Guangzhao Dai, Qi Wu, Bin ZhuAbstract
We study GPT-6-Astra in a zero-shot Vision-and-Language Navigation in Continuous Environments (VLN-CE) system, where it interprets instructions, assesses its surroundings, and proposes actions. The system uses a common observation--decision--execution workflow with direct model API calls, without a packaged agent harness or navigation-specific fine-tuning. In this workflow, each request receives selected observations, execution feedback, and retained progress records. Evaluation covers the complete system, including context management and action control. We evaluate the system on 50 of the 100 R2R-CE val-unseen episodes used by Open-Nav. It achieves a success rate of 52.0\%, an SPL of 48.9\%, and an nDTW of 70.8\%. Our analysis highlights three findings. First, recorded responses link landmarks and earlier actions to instructions using observations and supplied history. Second, reviews include requests for additional views and revisions of uncertain judgments. Third, the results suggest a gap between task understanding and autonomous completion: an unfinished crossing is recognized while rotation continues. At termination, 36.0\% of episodes succeed with a workflow-accepted STOP, while another 16.0\% meet the distance criterion at the step limit. These results highlight a central challenge: translating correct local judgments into sustained progress and appropriate stopping.
Abstract (translated)
URL
https://arxiv.org/abs/2609.20116
PDF
-
Perception, Layout, and Validation: Calibrated Confidence for Reliable Straight-Through Processing of Financial Documents
2026-09-17 12:09:21Yichao Jin, Yushuo Wang, Yuxuan Han, Kwan Ching Yee Sonia, Weiyang Song, Chiu Jin-Chun Kent, Wong Chong Hwee, Wong Tiong Kiat, Kenneth Zhu Ke, Jingyuan ZhaoAbstract
Straight-through processing (STP) on extracted key-value fields from financial documents without human review requires a calibrated probability together with a bounded guarantee on the residual error of the auto-approved tier. The emergence of modern Vision Language Models (VLMs) provides an out-of-the-box capability for extracting the key-values, but their verbalized confidence signals are unreliable and weakly track field correctness. This paper introduces a decomposed confidence layer along three interpretable channels, including perception, layout, and validation. Together with a final conformal risk control, the score can be used for reliable STP of financial documents. The method is validated on three public datasets covering real invoices, synthetic invoices, and ad-buy forms, using two different VLM families (Qwen3.6-27B and Gemini-3.1-Flash-Lite). Our decomposed score consistently improves the separation of correct from incorrect extractions, substantially raising the AUROC from 0.54-0.74 for VLM verbalized signals to 0.90-0.99 with contributions from all three designed channels. Crucially for industrial deployment, this enables usable STP. The native VLM confidence signals could clear only 0.1%-7.0% of fields under risk control at a target error of <10%. In contrast, the proposed method auto-approves 49-72% of fields while holding the empirical error of the accepted tier at or below the target.
Abstract (translated)
URL
https://arxiv.org/abs/2609.20110
PDF
-
KoNeoBench: A Curated Evaluation Dataset for LLM Understanding of Korean Neologisms
2026-09-17 08:58:49Soha Lee, Soojin Lee, Heesung Yang, Hyunju Song, Hyunji Lee, Jinsan An, Jeongwan Shin, Jin Hyun Park, Jun Lee, Hyeyoung Park, Kilim NamAbstract
Large language models (LLMs) are typically evaluated on static benchmarks, even though natural language constantly evolves through newly emerging words and meanings. Existing Korean benchmarks are centered on established vocabulary and therefore provide limited coverage of such recent lexical change, and their English-oriented design makes it difficult to assess the typological properties of Korean, in which content words combine productively with functional morphemes. In this paper, we introduce KoNeoBench, a benchmark for evaluating LLMs' understanding of Korean neologisms. KoNeoBench is built on 1,785 Korean neologisms attested in online news since 2020 and curated through expert lexicographic review. Each entry provides usage examples, word-formation analyses, and dictionary-style definitions. Based on this resource, we define four tasks and report results on recent models, together with a human baseline. Our experiments show that current LLMs exhibit clear limitations in recovering source components, distinguishing semantic categories, and generating accurate definitions. These results reveal specific aspects of recent Korean lexical change that remain challenging for current LLMs. KoNeoBench is available at this https URL .
Abstract (translated)
URL
https://arxiv.org/abs/2609.19916
PDF
-
PACE: Precise AI Cinematic Expression: A Typed Specification for Script-Grounded Previsualization and Geometric Conformance
2026-09-17 08:05:08Bing Duan, Qiang Guo, Linpu Li, Zhijian Mao, Min Zhu, Zhirui Ren, Yiwei Yan, Xi Chu, Xiaoding LiAbstract
Between a screenplay and a film sits a planning problem that is spatial first: who stands where, and what a camera sees from where it stands. An image diffusion model asked for a shot in free text settles that plan by its own defaults. We present PACE (Precise AI Cinematic Expression), a typed representation for the plan: the screenplay evidence, the characters, props and locations it needs, where each subject stands, and what the camera does. A value is written once at the level it belongs to (script, scene, shot or panel) and inherited below it. A compiler turns the result into both the prompt sent to the diffusion model and a 3D scene built in metres, and a camera solver places the camera so that the declared framing is the framing built. Where a declared value becomes geometry, PACE measures, field by field, how far the compiled camera and the staged render sit from the declaration, rather than asking a model to judge. On the 11-scene Automatic Drive screenplay, every staged single-subject panel places its subject within 1.2% of frame width of its declared position; with two or three subjects one camera pose cannot satisfy every position, and the residual is reported rather than absorbed. On 204 external director-storyboard shots, delivered head height is 1.906 times the staged target from the director's words, 1.733 from the compiled prompt, and 0.955 with the greybox control; the condition that holds framing best draws the described action least. Declaring the pose on 30 shots raises the action drawn from 58.9% to 74.4% without moving the framing. Transitions, fitted motion and human review of the generated panels remain open. Code: this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2609.19853
PDF
-
Reproducing Transparent and Scrutable Recommendations: Exploring Open-Weight Models via Natural-Language User Profiles
2026-09-17 07:36:34Noah Mami\'e, Laurin van den BerghAbstract
In this reproducibility study, we investigate the transparency and scrutability of recommender systems enhanced by incorporating generated natural-language user profiles that represent user preferences. The original paper explores the synthesis of user profiles from raw user-generated review text across domains such as movies and accommodations (Amazon Movies & TV, TripAdvisor). Crucially, these natural-language user profiles enable direct user interaction and intervention, allowing users to customize recommendations by correcting misattributed preferences or addressing cold-start settings. We successfully reproduce the core findings of the original study. Additionally, we extend the evaluation by conducting systematic context ablation experiments, multi-seed stability across five distinct random seeds to establish statistical reliability, and a mechanistic interpretability analysis using the nnsight framework to probe internal model representations under counterfactual profile perturbations. Our findings verify the original paper's claim that User Profile Recommendation (UPR) achieves competitive performance under its test-set reranking protocol and makes recommendations more transparent. Perturbing the natural-language profiles does change predictions, but it shifts predicted ratings uniformly across genres with no detectable genre-selective effect, leaving rankings unchanged even under direct activation steering. We trace this back to the rating-regression objective rather than the profile interface, with ranking-objective models clearly exceeding in this task.
Abstract (translated)
URL
https://arxiv.org/abs/2609.19831
PDF
-
Long-horizon autoformalization of a core theorem underlying MIP* = RE
2026-09-17 07:26:49Sirui Lu, Ruixuan Deng, Yanqiao Zhu, Zhengfeng JiAbstract
Landmark mathematical formalizations have taken specialist teams years to complete. We present FormalFlow, a system that coordinates AI proving agents under human supervision to address statement drift and proof composition in long-horizon formalization. Drawing on software engineering principles and practices, it uses a shared blueprint to guide nested planning, proving and review loops. Agents strengthen verification and review throughout formalization. We completed a machine-checked Lean 4 proof of the quantum soundness of the classical low individual-degree test, a core theorem underlying MIP* = RE. Developing the proof took 63 days; greater parallelism could further reduce this time. The final library contains 126,367 lines of Lean code, all generated by agents. The formalization corrects side conditions and intermediate errors while preserving the published final error bound under corrected assumptions. This work provides a verified foundation for quantum complexity and demonstrates a route to affordable verification of major research proofs by small teams.
Abstract (translated)
URL
https://arxiv.org/abs/2609.19814
PDF
-
ScientistTwo: Pioneering the Human Knowledge Frontier with Autonomous AI
2026-09-17 03:37:42Jaehyun Nam, Jinsung Yoon, Yanzhou Pan, Yubo Wang, Rui Meng, Parthasarathy Ranganathan, Tomas PfisterAbstract
Scientific discovery is defined by the ability to identify the boundaries of existing knowledge and venture into unexplored territory. The ultimate vision for AI in science is problem-driven autonomous research: given a fundamental challenge by a human expert, the AI independently navigates the scientific landscape, uncovers theoretical and empirical bottlenecks, and systematically expands the frontier of knowledge. In this paper, we introduce ScientistTwo, a fully autonomous multi-agent framework designed to realize this vision. Specifically, ScientistTwo takes an initial problem as input, establishes state-of-the-art baselines, formulates novel hypotheses, and coordinates specialized agents to orchestrate an end-to-end discovery cycle without human intervention. Moreover, the framework rigorously conducts experiments using diverse datasets and metrics, refines methodologies through automated ablation studies, and validates research findings via a closed-loop simulated peer-review rebuttal engine. To evaluate ScientistTwo's capabilities against the highest standards of human scientific achievement, we benchmark it across papers accepted at top-tier conferences such as ICLR, ICML, and NeurIPS. As a result, ScientistTwo autonomously generates expert-level, publishable papers and fully verified, executable codebases. Its solutions consistently outperform human state-of-the-art models, and achieve higher average review ratings than human-authored papers under automated AI review agents. These results show that ScientistTwo is not merely an assistive tool but an autonomous scientific pioneer capable of pushing the frontiers of human discovery. Project website: this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2609.19644
PDF
-
From Intent to Action: Benchmarking LLM Safety in Vehicle Voice Command Authorization
2026-09-17 03:20:22Diba Afroze, Xingli Zhang, Yazhou Tu, Xiali HeiAbstract
Large language models (LLMs) are increasingly integrated into vehicle voice assistants. But linking natural-language requests to vehicle functions creates a safety-critical authorization problem. Before executing a command, the system must choose whether to execute, refuse, clarify, require confirmation, defer to manual control, trigger an emergency response, or make no tool call. To our knowledge, prior evaluations do not isolate this pre-action decision across speaker role, authentication status, vehicle state, and tool availability. We introduce a 202-scenario benchmark with Reference Decisions under a seven-class taxonomy. We evaluate two local open-weight models and three API-based LLMs using Decision Alignment and safety-specific error metrics. Alignment ranges from 40.1% for Llama 3.2 3B to 89.1% for Gemini 3.1 Pro Preview. The API-based models score between 83.2% and 89.1%, with no statistically significant differences among them. Even these models produce two to three False Executes among 161 non-execution scenarios, and persistent errors remain in confirmation and manual-control decisions. A controlled Llama 3.2 3B ablation increases alignment to 40.1% under the structured authorization policy, versus 28.2-29.2% under schema-only and generic-safety baselines, but it does not eliminate False Executes. Structured LLM decisions are therefore insufficient as a standalone safety mechanism, and deployment requires an independent enforcement layer that verifies tool permissions and vehicle-state constraints before invoking any vehicle function.
Abstract (translated)
URL
https://arxiv.org/abs/2609.19630
PDF
-
DataCanvas-EDU: An Agentic Framework for Instructor-Guided Synthetic Data Generation in Business Analytics Education
2026-09-17 03:08:26Bang An, Maria Hamdani, Joseph FoxAbstract
Business analytics education requires diverse datasets to support different learning objectives, student backgrounds, and analytical tasks. Real-world data can be difficult to obtain and offer limited flexibility for adapting a case to a particular course. Even when suitable data are available, instructors must investigate the patterns, verify the results, and prepare assignments and reference solutions, requiring substantial time and effort. The use of large language models (LLMs) introduces an additional concern about training data contamination. Widely used public datasets often have extensive tutorials and worked analyses that models may have encountered during training. Students may therefore receive explanations drawn from existing analyses without practicing how to investigate unfamiliar data in collaboration with AI. This paper presents DataCanvas-EDU, an agentic framework for instructor-guided synthetic data generation in business analytics education. Instructors specify teaching goals and intended patterns through conversation, while an AI agent writes generation code, checks the resulting data, and prepares assignments, reference analyses, and rubrics. Four phases, Plan, Create, Verify / Test Analysis, and Evaluate, organize the process and support instructor review and revision. The framework is intended to simplify case preparation while creating opportunities for students to investigate newly designed patterns with AI. We illustrate the approach with WindowDash, a food delivery case containing 15,000 orders and nine designed patterns. DataCanvas-EDU is packaged as a reusable AI Agent Skill for compatible agent environments, with the package and installation instructions available at this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2609.19617
PDF
