Peer Revenue Now accepting positions

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Friday 6 November

Artificial Intelligence

Tracing Psychometric Inference in Large Language Models

Do large language models represent psychological constructs internally, or do they only generate outputs that resemble psychometric structure? We address this question using a cross-persona paradigm across 14 model variants from two families (Llama 3 and Qwen 2.5, 0.5B-14B, base and instruct). Given responses on one psychological scale from real individuals (N = 272 Taiwanese university students), models predict responses on six other scales. At the behavioral level, larger variants reproduce and amplify human cross-scale correlations; amplification persists after correcting the seven-scale human correlations for measurement attenuation. We then examine whether this structure is reflected internally. Contrastive direction analysis finds geometry aligned with the human matrix in larger instruct variants, but little alignment in base variants under this assay. Across models, geometry strength is associated with behavioral amplification (r = 0.68, p = 0.008; partial r = 0.69 controlling for log size). To relate internal structure to output behavior, a matched activation-probing paradigm shows that probe slopes vary less than behavioral slopes (1.38-1.77 vs 0.52-1.53). A synthetic control shows that ridge probing can inflate these slopes; after subtracting a baseline based on a reliability proxy from another analysis, adjusted slopes remain associated with behavior (r = 0.88). We term the measured probe-behavior gap readout attenuation. These results establish decodability and geometric alignment under the tested assays, without showing that the measured directions mediate the cross-persona answers.
Read the full abstract

Artificial Intelligence

LandslideFM: A Foundation Model Towards Unified Landslide Perception and Understanding

Intelligent landslide detection is essential for disaster prevention. However, existing models primarily focus on isolated visual tasks, require extensive labeled data, and lack deep analytical and scene-understanding capabilities. To address these limitations, we propose LandslideFM, the first unified foundation model that connects visual perception tasks with contextual interpretation of landslides. Specifically, LandslideFM adapts to multiple visual tasks via a Source-Agreement Graduated Trust semi-supervised learning framework that leverages limited labeled data alongside abundant unlabeled data. Furthermore, predicted landslide regions provide explicit spatial guidance to a Vision-Language Model adapted through parameter-efficient fine-tuning, enabling comprehensive landslide reasoning. To support training and evaluation in this domain, we also present LandslideVL, the first benchmark for intelligent landslide analysis, combining visual imagery and newly introduced textual prompts. Comprehensive experiments across public datasets and LandslideVL demonstrate that LandslideFM achieves superior quantitative and qualitative performance. Most importantly, LandslideFM surpasses general-purpose VLMs in specialized landslide reasoning and scene analysis. Our code is released at https://github.com/wangshanwen001/LandslideFM.
Read the full abstract

All authors Shanwen Wang, Sirui Wang, Xin Sun, Jungong Han

Artificial Intelligence

Train on prefill, monitor at decode: Using sparse autoencoders for topic monitoring

Topic monitoring is the task of labelling the topic of each token in a model's response during generation. Activation-based monitors that run at generation time have so far been trained to detect harmful content and are limited to two or three classes. We introduce TopicScope, a framework for training causal topic classifiers on SAE features that assign a topic to every token, so that the label can change partway through a response. We train three detector architectures on synthetic multi-topic conversations covering seven topics. Across Gemma-3 1B and 4B, monitors trained on features extracted from prefill hidden states were able to generalise to decoding. We also found that the classifiers are robust to perturbations, losing at most 5 macro-F1 points when off-topic asides are inserted, topics are removed or sections are reordered. Ablations conducted to understand the contribution of the SAE revealed that classifiers reaching baseline performance without it were several times larger. Our findings show that compact topic monitors can be trained inexpensively in prefill and deployed during generation. More broadly, our work shows where SAEs are useful beyond safety classification. Our code, dataset and checkpoints are available at https://github.com/Mercity-AI/SAE-Guardrails and https://huggingface.co/collections/Mercity/online-guardrails-using-sparse-autoencoders-6ac602940468c62fefb2b9e8
Read the full abstract

Artificial Intelligence

EarthRecord: From Pixels to Land Change with Multimodal Agents over Four Decades of Satellite Imagery

While multimodal agents now carry out complex data analysis with tools, their ability to turn decades of satellite imagery into reliable land-change analysis, crucial for Earth observation science, remains unmeasured. Existing benchmarks fall short: they pose questions on curated images or give agents preprocessed inputs, leaving out the raw multi-sensor archives where such analysis begins. We introduce EarthRecord, a benchmark built on the raw record of 56 sites: four decades of Landsat, MODIS, VIIRS, and Sentinel-2 data, with cloud masking and cross-sensor harmonization left to the agent. It poses 238 land-change questions across ten analysis tasks, scored against withheld references from global products, agreement among methods, and visual interpretation. Two controls test what answers rest on: the closed book removes the record, and instrument pairs alter one satellite's data while nothing changes on the ground. Evaluations of twelve multimodal agents reveal that leading agents complete most analyses end to end but falter when an answer requires mapping from imagery. Further analysis shows that the record moves answers from prior knowledge to the pixels but not yet to the real change on the ground: when one satellite's data are altered, half of the right answers follow the alteration. EarthRecord thus provides a principled platform for measuring and advancing agents from pixels to land change.
Read the full abstract

All authors Haoran Zhang, Huizhen Sun, Hanbin Wen, Jie Leng, Jiaxin Zhu, Zhaohua Sun

Artificial Intelligence

Auditable Human-Agent Teams for Autonomous X-ray Experimentation

Closed-loop autonomous experimentation can accelerate scientific discovery, but implementation effort limits accessibility. Agentic systems may lower this barrier, although sustained operation on physical instruments remains largely unexplored. Here, using coding-agent infrastructure grounded in validated scientific methods, we demonstrate autonomous synchrotron X-ray experiments on nanoparticle superlattices under evolving, human-defined objectives. Complete human-agent and agent-agent transcripts enable an audit that reconstructs decision provenance and characterizes the division of cognitive labor. Across three experiments, 61-67% of messages involved interpretation or judgment, while the share of methodological and scientific contributions increased from 46% in the setup-focused first experiment to 56-61% in later experiments. Provenance showed that 59-68% of error corrections involved multiple participants and that about half of all decisions incorporated the input of another participant. These results may guide design principles for multi-agent scientific systems and inform future agent harnesses, model training, and accessible autonomous experimentation.
Read the full abstract

All authors Gunik Luthra, Ronald J. Pandolfi, Benjamin T. Diroll, Ricky Ronquillo, Ruipeng Li, Christopher B. Murray, Marcus M. Noack, Kevin G. Yager, Esther H. R. Tsai

Computer Vision

URL-Sim: Unified and Regulated LiDAR Simulation for High-Fidelity Reconstruction of Dynamic Driving Scenes

Reconstructing high-fidelity LiDAR simulation scenes from real-world driving logs is essential for cost-effective safety validation of autonomous driving systems. While simulators based on Neural Radiance Fields (NeRF) achieve realistic scene reconstruction, their large-scale adoption is limited by slow rendering speeds. Recent 3D Gaussian splatting (3DGS) approaches have improved rendering efficiency through explicit primitive representation and differentiable rasterization. However, existing 3DGS-based methods remain largely restricted to rigid vehicles and produce redundant primitives through unconstrained densification. To address these limitations, we propose URL-Sim, a unified and regulated dynamic LiDAR simulation framework. URL-Sim constructs a dynamic Gaussian scene graph that decouples the modeling of traffic participants, including vehicles, pedestrians, and cyclists, enabling accurate reconstruction of heterogeneous dynamic objects within a single framework. It further introduces a regulated Gaussian densification strategy that ranks primitives with multi-dimensional importance scores and allocates Gaussians to the most critical regions while keeping the total primitive count bounded. Experiments on the Waymo and KITTI-360 datasets demonstrate that URL-Sim achieves higher quantitative accuracy and visual quality than existing baselines under comparable primitive budgets, and accurately recovers the geometry and poses of diverse dynamic traffic participants.
Read the full abstract

All authors Kefei Qian, Xipeng Feng, Jie Li, Long Cao, Yuhan Chen, Xiangfei Huang, Guofa Li

Quantum Physics

Adaptive Protocol for Rényi Entropy Estimation

The second-order Rényi entropy is one of the most important quantities used for characterizing the entanglement structure of states prepared in quantum many-body experiments. With access to qubit projective measurement data, it can be estimated via a protocol in which repeated measurements are taken after each of a set of random local unitary operations is applied to the quantum state. To obtain accurate values for the second Rényi entropy on shot-limited devices, care must be taken when choosing the number of different unitary settings to apply, and the respective number of projective measurements to take. In this paper, we propose an adaptive protocol for estimating the Rényi entropy that iteratively optimizes the ratio of basis settings to measurements, using a gradient-based approach and starting from an initial small set of measurements. Using synthetic data, we demonstrate that for several prototypical states, our method converges to values for this ratio that are comparable to the optimal values obtained by a brute-force grid search, and it exhibits slower growth of the measurement budget required for a fixed target accuracy than fixed-ratio protocols. Our adaptive protocol therefore promotes efficient estimation of the Rényi entropy in finite-size, shot-limited quantum experiments by making effective use of the available measurement budget.
Read the full abstract

All authors Mu Niu, Yi Hong Teoh, Roger G. Melko, Rajibul Islam

Artificial Intelligence

SIRIS: Can LLMs Tell When Simulation Errors Matter in AI-Driven Systems Research

LLM agents increasingly drive systems research and rely on simulators to evaluate candidate designs at scale. Simulators are imperfect by design: their modeling limitations can distort predictions yet leave the recommended design correct, or change which design should be selected. We introduce Siris to test whether LLMs can tell these cases apart, judging from their systems understanding when a limitation invalidates the simulator's recommendation. By perturbing a simulator, Siris constructs workload pairs with the same candidates, limitation, and simulator recommendation. The perturbation changes predictions on both workloads, yet a reference simulator supports the recommendation on only one. Evaluating thirteen frontier LLMs on LLM serving design problems, we find that identifying a limitation does not ensure a correct choice, and even the strongest models sometimes abandon valid recommendations. Self-correction through a review turn helps these models restore valid choices but introduces new errors in several others. These findings motivate evaluating systems understanding through design decisions under imperfect simulators.
Read the full abstract

All authors Jinwoo Hwang, Sukmin Cho, Tao Zhang, Soheil Abbasloo, Jongse Park, Peng Cheng

Geometric Topology

Liftable braid factorizations from line arrangements: Hironaka's relations and a Zariski pair

We realize Hironaka's generalized lantern relations by lifting positive braid factorizations. For a nonpencil real line arrangement with no parallel lines, we construct a family of branched coverings on the regular fibers of a blown-up pencil of spheres. The motion of the branch points gives a factorization of a power of the full twist into least liftable powers of half-twists. The Artal-Carmona-Cogolludo-Marco Zariski pair gives two such factorizations of the same braid for a single simple branched covering of the disk. Their lifts are Hurwitz inequivalent even up to simultaneous conjugation by page homeomorphisms. The corresponding Milnor fibers fill the same planar contact three-manifold and have the same incidence matrix. They are homeomorphic and lie in distinct strong symplectic deformation classes.
Read the full abstract

Computation and Language

Policy Over Purpose: Auditing LLM Content Moderation of Reclaimed Slur Usage

As Perspective API is deprecated, platforms are turning to large language models for content moderation, encoding policy as natural language that the model must interpret at inference time. Reclaimed slurs test this design directly. The same term can express in-group pride or target the community that reclaimed it, and misreading it silences the speech moderation is meant to protect. We audit four general-purpose LLMs and four moderation-specific systems, with Perspective API as a legacy baseline, on IYKYK, a corpus of reclaimed and derogatory usage of three slurs annotated by community members. Across the policy, decision, reasoning, disparity, and steering layers of this audit, we find that whether a system was purpose-built for moderation explains little about how it handles reclaimed language. No system reliably agrees with community judgment, and all fail most on language reclaimed within Black communities. What separates systems is instead how their policy is represented. Only systems whose policy is expressed in the prompt can be steered at all, and even then, explicit definitions change how models justify a decision more than how they make it. We argue that policy transparency determines whether moderation failures can be corrected, but that correcting them requires contextual judgment that definitions supplied at inference time cannot provide.
Read the full abstract

High Energy Physics (Theory)

An RG perspective on the (un)naturalness of Starobinsky inflation

Starobinsky inflation is often regarded as a successful minimal model of the early universe, yet it relies on a large $R^2$-coupling, in apparent tension with the expectation that Wilson coefficients in a gravitational effective field theory are of the same order. We test this expectation with a first-principles renormalization group calculation. We use asymptotically safe gravity as a testing ground to define ultraviolet-complete renormalization group trajectories and to uniquely predict higher-order couplings in terms of the free parameters of the theory—the cosmological constant and the $R^2$-coupling in Planck units. Including higher-curvature operators up to the fourth order, we find that indeed the flow generates dimensionless couplings of comparable size in the ultraviolet, with the irrelevant couplings tied to the relevant ones by linear relations. However, at sub-Planckian energies a canonical-like scaling takes over, and the irrelevant couplings get suppressed by powers of the ratio between the inflationary and Planck scales. This hierarchy yields the suppression of higher-order operators needed for consistency with cosmological data, so that the $R^2$-dominated dynamics emerges as a natural low-energy approximation. At the same time, the higher-order terms required by fundamental physics induce small but calculable deviations from the Starobinsky predictions, which may be worth considering in light of the latest and future DESI and ACT results.
Read the full abstract

All authors Francesca Campani, Francesco Ferrarin, Serena Giardino, Alessia Platania

Machine Learning

Update Slope Is Bounded by Prior Calibration: A Structural Bound and Its Contingent Empirical Signature

Evaluations of LLM forecasters increasingly measure how forecasts improve as richer evidence is supplied. We identify two confounds in this update-slope metric. First, outcome-aware evidence construction substantially inflates measured responsiveness: on the same models and questions, slopes are roughly four times larger than with resolution-blinded evidence ladders. Second, update slope is a change score with a prior-dependent structural ceiling, so raw comparisons can reflect differences in starting calibration rather than updating behavior. The ceiling defines only a feasible region; whether slope and prior calibration are empirically coupled depends on headroom utilization and varies across our Manifold, ForecastBench, and ACLED pools. Despite this variability, conditioning on prior calibration roughly triples the association between intermediate updating and final Brier score under both blinded and outcome-aware construction. Question identity also explains substantially more variance than model identity across the evaluated pools. We recommend building evidence ladders blind to outcomes, conditioning on prior calibration, and reporting headroom-normalized measures alongside raw slopes.
Read the full abstract

Software Engineering

Caleta: An Evidence-Oriented Trust Layer for Reproducible Local Large Language Model Inference

Local large language model (LLM) inference is often described as private, verified, or reproducible without separating the technical claims behind those terms. This paper presents Caleta, an Apache-2.0 process supervisor and evidence layer around llama.cpp rather than a new inference engine. Caleta combines versioned manifests, exact-size and SHA-256 verification, controlled backend construction, a restricted OpenAI-compatible proxy, prompt-free execution records, benchmark reports, and a small conformance suite. Its contribution is the composition of these mechanisms into explicit contracts that distinguish integrity from authenticated provenance and privacy controls from operating-system isolation. We evaluate the implementation through static checks, 22 unit tests, deterministic end-to-end tests on Ubuntu, macOS, and Windows, and a pinned real-model experiment on Windows CPU using Qwen2.5-0.5B-Instruct Q4_K_M and llama.cpp b11981. The real experiment validates artifact verification, startup, model listing, non-streaming and server-sent-event chat, and conformance. Three independent benchmark invocations produced fifteen successful warm requests with no failures. In this single configuration, median time to first token was 18.066 ms and median decode throughput was 54.394 tokens/s. These descriptive measurements do not establish cross-engine superiority, output quality, energy efficiency, or hardware-general performance. Raw JSON, commands, and environment metadata are published. Remaining challenges include signed manifests, time-of-check/time-of-use replacement, operating-system containment, proxy-overhead measurement, and broader experimental matrices.
Read the full abstract

Functional Analysis

Characterizations of minimal Hermitian completions via moments and subdifferentials

Given a Hermitian matrix $A_0$ and a real subspace $W$ of Hermitian matrices, we study the matrices $A_0+W_0$, $W_0\in W$, of minimal spectral norm, which we call $W$-minimal. The main motivation is the Hermitian matrix completion problem, where $W$ is spanned by the Hermitian matrix units corresponding to the unspecified entries. Our main tool is the moment of a subspace $S\subseteq\mathbb{C}^n$ relative to $W$, a compact convex subset of $\mathbb{R}^r$, $r=\dim(W)$, which we relate to the joint numerical range and to the $k$-dimensional field of a family of matrices. We compute the subdifferential of $\|A_0+W(x)\|$ in terms of the moments of the eigenspaces of the extreme eigenvalues of $A_0+W(x)$, and we obtain several equivalent characterizations of $W$-minimality: in terms of eigenvectors, of moments, of subdifferentials and of norm one functionals that annihilate $W$. When the identity belongs to $W$, every $W$-minimal matrix has balanced spectrum, and minimality is equivalent to the moments of the eigenspaces of the maximum and minimum eigenvalues having nonempty intersection; we show that this characterization fails if $I\notin W$. We illustrate the results with several examples, and we prove that the adjacency matrix of a connected graph is minimal for its completion problem if and only if the graph is complete bipartite.
Read the full abstract

Computers and Society

Police Behavior Patterns Through Radio Analysis

In the United States, most police departments communicate using police radio, which allows for officer-to-officer and officer-to-dispatch communication without depending on cellular network or WiFi availability. While some departments use encrypted digital radio, many continue to transmit in the clear, and are legally monitored by public citizens with `police scanners.' Because police radio is intended for internal police use and represents real-time communication, it may offer valuable insight into patterns of law enforcement activities without the potential bias and filtering introduced by a press release after the fact. In this study, we collect and examine over 14 million radio transmissions representing over 100,000 hours of speech by officers in 34 cities across the United States from July 2025 through July 2026; to our knowledge the largest scale analysis of Broadcast Police Communication (BPC) to date. We highlight both typical patterns of police communication and categorize anomalous activity that represents incidents including protests and major weather events. We present this dataset as a starting point for an array of police communication studies, conduct our own preliminary analysis, and propose several further areas of inquiry.
Read the full abstract

All authors Milo Z. Trujillo, Adam Perhala, Elaine Zou, Tyler Mckenzie, Juniper Lovato, Brooke Foucault Welles, Alyssa Hasegawa Smith

Computer Vision

LoCo: Selective Local Competition for Discriminative Open-Vocabulary Multi-Label Recognition

Open-Vocabulary Multi-Label Recognition (OV-MLR) aims to identify all queried semantic concepts present in an image. Existing methods mainly improve global image-text matching or text-side representations, leaving local prediction largely under-explored. In this paper, we revisit local prediction and reveal that it can serve as a strong discriminative branch once two properties are properly handled: spatial consistency of local visual features and compatibility-aware local competition. This perspective is motivated by local semantic sparsity: each image region is related to only a small subset of queried concepts, while compatible concepts of different granularities or types may still coexist locally. We instantiate this perspective with LoCo, a selective Local Competition framework for OV-MLR. LoCo adopts a spatially consistent visual encoder to reduce local feature entanglement, and introduces Selective Softmax to impose competition only among locally incompatible categories while preserving compatible responses. Multi-scale inference further improves local prediction at different scales. LoCo can also be ensembled with an existing global predictor to provide complementary image-level context. Experiments on four multi-label benchmarks show that standalone LoCo improves average F1-score and mAP over TagCLIP by 3.4% and 3.8%, respectively. LoCo + DualPrompt achieves the highest averages among the compared methods, with gains of 4.8% and 5.2% over TagCLIP. Our results suggest that local semantic sparsity offers a useful basis for developing more discriminative open-vocabulary multi-label recognition methods. The code is available at https://github.com/lee-plus-plus/LoCo.
Read the full abstract

All authors Zhuoming Li, Bo Han, Xiaoyu Wang, Yuheng Jia

General Relativity and Quantum Cosmology

Relativistic Simulations of Self-Gravitating Radiation Fluids

We present an approach for numerical simulations of pure radiation fluids within the general relativistic hydrodynamics framework. The approach models radiation as a self-gravitating perfect fluid with vanishing rest mass, enabling simulations of radiation fluids in strong-gravity spacetimes. Building on the existing relativistic hydrodynamics code Whisky, within the Einstein Toolkit framework, we introduce modifications to incorporate radiation fluids, including updates to the source terms, recovery scheme, and Riemann solvers. Modifications are validated using standard shock tube tests. We present results for radiation accretion onto a single black hole, as well as a photon star solution for radiation in unstable hydrostatic equilibrium, demonstrating both collapse and dispersion under perturbations.
Read the full abstract

Computation and Language

Prove What You Can: Hybrid Neurosymbolic Reasoning for Logical Natural Language Inference

Logical reasoning over natural language requires both semantic interpretation and deductive search. Large language models (LLMs) can propose plausible intermediate derivations, but they often fail to compose them into a complete proof and overpredict uncertainty. Neurosymbolic methods improve search by translating an entire problem into first-order logic and handing it to a prover, but a single bad clause can make the whole translation nonexecutable. We present Hybrid Reasoning, a three-stage pipeline that decouples proposing a proof, validating its steps, and deciding a label. An LLM proposes a structured proof tree with a raw verdict; a theorem prover validates each derivation locally, keeping only the subset it can prove; and an uncertainty-gated repair attempts to re-derive the label from the premises plus the validated derivations, when the raw verdict is UNCERTAIN. Across four benchmarks (FOLIO, ProofWriter, ProverQA, PrOntoQA) and five contemporary LLMs, Hybrid Reasoning is the top-performing method on the two hardest datasets, improving on the strongest baseline by 5.3 points on FOLIO and 4.9 on ProverQA, and reaches ceiling-level accuracy on the easiest. The repair stage operates at 78-98% precision, producing 326 correct relabellings against 19 regressions.
Read the full abstract

Computer Vision

AI-Driven Repair of Missing Automotive Components

Due to supply constraints, replacing damaged automotive components can be expensive and time-consuming, while replacing the entire component can be wasteful when much of it remains usable. This paper presents a pipeline to investigate whether missing geometry can be reconstructed as a repair patch using paired observations of the intact models and synthetically damaged components. Noise-free color, depth, and silhouette images were produced using the camera viewpoints. Depth differences were then used to identify the potential missing regions, while multi-view voxel reconstruction was used to estimate the missing portion. A pretrained DINOv2 model was also added to validate missing candidates and determine whether pretrained visual features could improve reconstruction. We tested the pipeline on 20 automotive CAD models, each with a synthetically removed portion. Each model was tested under four different conditions using two reconstruction methods for a total of 160 trials. The predictions were compared with the missing geometry of the unchanged ground-truth using predefined thresholds of the voxel intersection-over-union over 0.85 and the absolute relative volume error less than 15%. Of the 160 experimental trials, 109 of them satisfied both thresholds. Geometry-only reconstruction passed 16 of 20 components with eight views but only 10 of 20 with four adjacent views. Further analysis showed that 24 out of the 37 failed reconstructions could be categorized as unsuccessful because extra material was incorrectly added to areas that should have been empty. The implementation of DINOv2 into the pipeline did not improve the overall reconstruction accuracy and made prediction generation slower. For 13 verified tests, printing only the repaired sections instead of the entire parts reduced filament use by 93.08% and printing time by 89.39% based on the median results.
Read the full abstract

Artificial Intelligence

TripleGrader: A Web-Based Platform Supporting Collaborative Triple Annotation with Quality Grading and Inter-Rater Agreement Assessment

Knowledge graphs need validated relational triples, and gold benchmarks of such triples are often built by several annotators. Most text annotation platforms focus on span tagging and compute agreement on categorical labels. They are not designed for ordinal quality grading of triples that have already been extracted. We present TripleGrader, an open-source web platform for grading subject-predicate-object triples on a 0 to 10 ordinal scale. In TripleGrader, each triple is scored independently, and peer scores stay hidden until the grader has rated it. Graders then discuss entries in threaded comments and propose revisions, which the session owner approves and publishes as new versions. For agreement, the platform computes quadratic-weighted Cohen's kappa for two raters and a quadratic-weighted Fleiss-type kappa for three or more. Graded sessions and the final benchmark are exported in common formats. We evaluated the platform in three ways. We ran 72 automated tests, including 18 workflow scenarios through the REST API. We also injected 10 error types to check whether prescribed corrections could be applied while preserving unaffected triples. On two small corpora with existing gold benchmarks, we then used the platform to curate the triples extracted by four large language models into each benchmark. Only 10% to 32% of the candidate triples could be kept unchanged, and 53% to 71% needed corrections, partly to match the benchmarks' naming conventions. Per model, 11 to 29 missing triples had to be added. On these corpora, the extracted triples therefore needed substantial human review. The source code is available at https://github.com/razz0n/triple-grader-tool.
Read the full abstract

Hardware Architecture

The World Model Planning Accelerator (WMPA): Hardware-Algorithm Co-Design for Embedded AI in Robotics

Embedded AI is moving from perception to prediction. Robots now use world models$\unicode{x2014}$learned simulators that predict the outcome of candidate actions in a compressed latent space before any is executed. Every predicted step must be computed on board within a tight power and thermal budget, yet the energy these steps fundamentally require, and how far real hardware sits from it, have not been characterized. We derive lower bounds on the arithmetic operations and the CMOS dynamic energy of a latent state transition from information theory and Boolean circuit complexity. Profiling a DreamerV3-style Recurrent State Space Model (RSSM) shows that every rollout operation is memory-bandwidth bound and that a GPU consumes at least $5.9\times10^{10}$ times the CMOS energy floor. Transient 3D thermal simulations show that this inefficiency is also a thermal one, since sustained back-to-back planning heats a millimeter-scale silicon tile to 556 K even behind an ideal copper heat spreader. Guided by this chain from bound to bottleneck, we propose the World Model Planning Accelerator (WMPA), a chip architecture for embedded world model planning that combines near-memory computation, real-time entropy estimation, variable-precision rollout units, and copy-on-write latent scratchpads. With a design target of 40x lower power, the same tile stays near ambient at 306 K in simulation, pointing toward practical planning on thermally constrained robots.
Read the full abstract

Performance

Repeating Work Can Lower Energy per Successful Goal in LLM Agents

When a temporary call failure interrupts a tool-using LLM agent, the framework must choose how much completed work to repeat. Repeating less completed work intuitively saves computation, but recovery mechanisms also change subsequent execution and the chance of success. We compare recovery mechanisms by energy per successful goal (EpG), which divides the energy of all task runs (including failed runs) by the number that finish correctly. In controlled vLLM experiments on one node, we measure board energy above idle for eight recovery options on reliably solved multi-turn tasks, with selected comparisons across three model families. For the model tested across all eight options, with the prefix cache retained, starting over has 22 percent lower EpG than continuing after early failures. Under the same cache condition, continuing has 24 percent lower EpG than starting over after late failures. Among all eight recovery options, repeating only the interrupted turn has the lowest observed EpG after early and late tool failures on this model. Holding recovery fixed, retaining the prefix cache reduces prompt tokens needing fresh computation by 86 percent but reduces recovery energy by only 8 percent at contexts up to 7,333 tokens. These measurements and a condition relating recovery energy and success show when repeating work lowers EpG, helping framework developers choose how their agents recover on their intended models and tasks.
Read the full abstract

All authors Deepak Panigrahy, Stephen Abkin, Aakash Tyagi, David Kebo Houngninou, Rabi Mahapatra

Complex Variables

The Dimension Conjecture for Hypersurfaces Is True

The dimension conjecture in CR geometry is proved for arbitrary real-analytic hypersurfaces. More precisely, for $N\ge2$ the dimension of the Lie algebra of infinitesimal holomorphic automorphisms of a germ of a holomorphically nondegenerate hypersurface in $\mathbb{C}^N$ of the indicated class does not exceed $N^2+2N$. This is exactly the dimension of the corresponding algebra for any Levi-nondegenerate hyperquadric in the same ambient space. Equality holds if and only if the germ is equivalent to a hyperquadric. For a finitely nondegenerate hypersurface with constant positive Levi rank $r$ satisfying $0<r<N-1$, the sharper bound $N^2+r+1=N^2+N-m$ is obtained, where $m=N-1-r$. We also establish the sharper bound $N^2+N-1$ for the automorphism algebra of a nonspherical holomorphically nondegenerate hypersurface; consequently, dimensions $N^2+N,...,N^2+2N-1$ cannot occur (a dimension gap).
Read the full abstract

Machine Learning

Multi-objective drift-aware MLOps agent for model retraining

Deployed ML models face non-stationary data and delayed labels, yet most retraining strategies rely on fixed schedules or brittle drift thresholds that ignore operational cost. In this setting, retraining is a sequential decision problem rather than a one-time threshold choice. We present DARE (Drift-Aware Retraining Engine), an offline, event-driven retraining framework that separates ingestion, monitoring, policy decisions, training, and logging through explicit schemas and deterministic replay. We compare DARE against practical heuristic retraining baselines, including fixed-interval, threshold-persistence, and action-rate-matched fixed scheduling policies. On Airline, the locked LinUCB policy achieves a higher mean F1 while substantially reducing intervention frequency relative to heuristic and action-rate-matched baselines. Additional runtime-calibrated evaluation shows that these gains persist when measured using the cumulative update wall-clock time. These results suggest that conservative sequential retraining policies can improve performance while reducing unnecessary operational overhead under delayed feedback.
Read the full abstract

Software Engineering

Device-Life Boundary Indicators and the Classification of Restored Versus Native Records on Android

A phone that shows calls, messages and photographs from years ago may have been set up recently and carried them over from an earlier handset, so its screen dates do not date the device's own life. Which traces fix where that life begins, how well they agree and how much of what is visible predates it is unmeasured. We read 14 Android full file system extractions (12 Samsung; Android 10 to 16), seven bug reports and 33 iOS extractions (a probe). Every Android device recorded a first boot, but the first evidence of user setup came a median of 98 days later (0 to 563). Of the 13 devices with messages, 7 held some dated before the first boot, a majority of them on 5; 24% of the apps outside the system image were installed as restored. Against the creator field that a migration tool leaves on messages, a boundary at user setup recalled all but 1 of 46,968 such messages, the first-boot boundary 94 to 100%. On synthetic fixtures, the labelling error followed the boundary error. A date on a record dates the device only once the record is shown to have been created on it.
Read the full abstract

All authors Daniel Kwaku Ntiamoah Addai, Selorm Adjabeng, Kingsley Albert Egyir, Sanders Larweh Tetteh

Dynamical Systems

Up or down by half: a family of Collatz-type maps

Each Collatz step changes a positive integer by roughly half: odd inputs go up and even inputs go down. We vary the directions while keeping the step size. Write x=2^m k (k odd), and let S be a set of positive integers. Define T_S(x)=(3x+1)/2 if m=0, 3x/2 if m is in S, and k otherwise. The map T_S^- differs only at odd x, which it sends to (x+1)/2. The usual Collatz map is T_emptyset. Assuming the independence of 2-adic valuations along odd-to-odd steps yields random-walk models with drifts (1+rho(S))log_2(3)-2 for T_S and rho(S)log_2(3)-2 for T_S^-, where rho(S) is an explicit series. Every T_S^- is contracting; T_S is contracting exactly when rho(S)<log_3(4/3), approximately 0.2619. We prove density results, classify interval sets S, and give a criterion ensuring every T_S^- orbit reaches 1 or (2,3). For seven contracting maps T_S, three with rho(S)=1/4 just below the threshold, computations show that every n<2^34 (n<1.668*10^10 for S={1}) reaches one of at most five cycles; stopping-time statistics match the model to 3-5 digits. We study branching maps, where any member of a family may be applied at each step. Boundedness of a family implies boundedness of each member; in particular, boundedness of {T_emptyset,T_{3}} would imply that no Collatz orbit diverges. For {T_emptyset,T_{1}}, a rule depending only on x mod 32 has positive drift (10/17)log_2(9/8) for almost every 2-adic starting point. Thus almost all n have reachable sets containing values larger than n^1.05; residues modulo 2^L, L<=10, cannot prove unboundedness. Finally, we define unified maps in which valuations also determine the direction at odd inputs. The cases x=1 mod 4 govern their behavior: if they go down, every block is dominated by the corresponding Collatz map (and a census finds only the cycles (1), (2, 3) and (3, 5) with smallest element below 10^9); if they go up, the maps behave like T_S.
Read the full abstract

Statistical Methodology

Proximal Difference-in-Differences for Causal Learning under Confounding and Outcome Drift

Estimating long-term treatment effects is essential in regulatory and healthcare decision-making, yet the limited follow-up duration of randomized controlled trials (RCTs) poses a significant challenge. Open-label extension studies attempt to capture additional information for treated subjects, but there is often significant systematic censoring and selection bias due to placebo crossover. While supplementing RCTs with real-world evidence is a common strategy, existing methods often rely on restrictive exchangeability assumptions that fail under latent confounding. We propose a novel proximal framework to integrate real-world evidence into an RCT with an open-label extension to estimate the long-term continued-control trajectory. Our approach successfully identifies long-term effects even in the presence of time-varying latent confounding effects and inter-study outcome drift. We provide a doubly robust locally efficient estimator that provides further resilience to model misspecification. Empirical results demonstrate that our framework maintains validity in complex data-generating processes where standard approaches may be biased.
Read the full abstract

All authors Sihyung Park, Shu Yang, Ilya Lipkovich, Wenyu Ye, Mingyang Shan

Software Engineering

Retention Horizons of Volatile Stores in iOS and Android Extractions

Whether an extraction can answer a question about a given day depends on how far back each store reaches, yet public statements rest on single test devices. We measured the retention horizon, the time from a store's oldest retained record to acquisition, for eight iOS and seven Android stores on 48 full file system extractions and seven bug reports, keeping devices whose clock agreed with the acquisition time (26 Apple devices, 13 Android phones). Horizons ranged from minutes to years and followed the mechanism that bounds each store. The unified log Persist stream reached a median of 2.5 days; interactionC reached 166 to 170 days on 18 of 25 devices. Android stores stayed within their source-code limits (usage statistics 10.2 days, DropBox 3.1, notifications 1.5); bug-report logs reached 3.1 hours at most. The oldest record misled three times: one knowledgeC stream reached back two years, the others a median of 29 days; interactionC read without its write-ahead log showed older records on 18 of 25 devices; and four of seven bug reports began with stray lines up to 16.6 days old above a body of 3.1 hours or less. A horizon belongs to a store, a release and a device.
Read the full abstract

All authors Daniel Kwaku Ntiamoah Addai, Sanders Larweh Tetteh, Selorm Adjabeng, Kingsley Albert Egyir

Statistics Applications

Measuring Public Institutions and their Agents: Disparities and Agreement, with Application to Asylum Decisions in the U.S. Immigration Courts

Public institutions derive their legitimacy from the professionalism and impartiality of their agents. In order to maintain standards of conduct, the performance of an agent is often compared to the performance of her peers. But public institutions typically assign agents to different tasks, and as a result, naive comparisons may reflect the difficulty of the assignment and not an agent's performance. To address this concern, we develop measures of disparity and agreement that compare an agent's performance with the potential performance of randomly selected peers who could have been assigned the same set of tasks. We then give realistic conditions under which disparity is identified from the marginal distribution of each agent's actions. We also show how agreement, which is generally unidentified when agents are assigned to different tasks, can be bounded or estimated under sensitivity assumptions about the dependence between potential outcomes. Finally, we use a nonparametric Bayesian model to examine applications for asylum filed between 2007 and 2014 in the widely studied and often maligned U.S. immigration courts. Our results do not support the picture of widespread partiality asserted in previous empirical work. At both the judge and court level, we find limited disparity and substantial potential agreement.
Read the full abstract

Machine Learning

How Much Is Left Matters: False Positives and False Negatives in Partial-Prompt Contamination Tests

Partial-prompt completion, where a model receives the first part of a benchmark problem and is checked for whether it reproduces the rest, is a widely used test for benchmark contamination. We audit this test on seven open-weight models using MATH-500 and recent AIME 2025/2026 problems, which serve as true negatives. Memorization flags depend strongly on how much text remains: for every model, problems with at most 8 leftover words are flagged several times more often than problems with at least 25 (31.5% vs. 5.5% for Qwen2.5-7B-Instruct), and half to three-quarters of all ROUGE-based flags come from short leftovers. On true negatives, a ROUGE-L >= 0.75 threshold flags up to 6.7% of problems at a 60% prefix, and at an 80% prefix even exact match flags 8.6% (pooled over models; 6.2% on AIME 2026, which postdates every model), against 0% at 40-60%. The test also misses contamination upstream: models distilled from Qwen2.5-Math, whose base is heavily flagged in prior work, show almost no model-specific recall. Calibrating each model against models from other families removes most flags while isolating a consistent Qwen2.5-specific signal at every prefix length. We recommend length-stratified reporting, prefixes of at most 60%, style-matched post-cutoff controls, and cross-family calibration.
Read the full abstract