Leaderboards, eval harnesses and ablations — the contested business of deciding which model is actually better.
arXiv:2610.33581v1 Announce Type: new Abstract: Megapixel visual synthesis with latent diffusion models requires operating far beyond the resolutions at which these models are typically trained. Casc...
arXiv:2610.34221v1 Announce Type: new Abstract: Aesthetic image cropping (AIC) aims to improve image composition and aesthetic quality by removing extraneous or distracting content. Over the past dec...
arXiv:2610.35219v1 Announce Type: new Abstract: Existing multi-agent collaboration (MAC) methods have primarily explored a group thinking collaboration topology, where agents independently generate s...
arXiv:2610.35554v1 Announce Type: new Abstract: Structure-from-motion (SfM) from the dual-fisheye video of a consumer 360-degree camera recovers geometry only up to a similarity transform; in the pip...
arXiv:2610.33684v1 Announce Type: new Abstract: Self-evolving training improves the performance of agents on difficult tasks through verifiable feedback, but it often applies costly deliberation unif...
The paper introduces a method called OptABA-PC that treats the repair of conflicting conditional‑independence tests in constraint‑based causal discovery as a minimum‑cost correction problem. Implemented in Answer Set Programming, it provides sound and complete guarantees, ensures subset‑minimality and cost optimality of the repair, and offers conflict‑based justifications for released facts. Empirical results show that OptABA-PC retains more correct facts, reduces ambiguity, and improves graph reconstruction compared to previous Causal ABA approaches.
Isharah‑Selfie is a large‑scale Arabic Sign Language dataset collected entirely with front‑facing smartphone cameras in unconstrained selfie conditions. It contains 8,997 video clips from 16 signers across 1,493 unique sentences, annotated with gloss sequences for continuous sign language recognition and Arabic sentence translations for sign language translation. The dataset focuses on realistic mobile communication, featuring one‑handed signing, close‑range viewpoints, partial body visibility, camera motion, unconstrained environments, and heterogeneous device resolutions, and introduces signer‑independent and unseen‑sentence evaluation protocols.
The paper proposes a method to capture expert human judgment and reasoning for automated evaluation systems that can scale across high‑stakes domains such as healthcare, defense, finance, governance, and politics. By assembling expert networks and encoding their reasoning into structured editorial standards, the authors demonstrate that automated judges trained on these standards outperform uncalibrated frontier models on source‑quality and neutrality metrics in AI news reporting. The study validates the approach with strong inter‑annotator agreement and high F1 scores, suggesting a viable framework for expert‑calibrated evaluation in any domain where expertise is scarce.
AdaPRA is a process reward agent that uses adaptive rubrics to supervise data science agents, correcting errors such as insufficient exploration, unsound validation, coding bugs, and inefficient budget use. It accumulates knowledge across multiple attempts on the same task and adapts its rubrics to each new task, improving performance on DSPredict-Easy and DSPredict-Hard benchmarks. The approach outperforms existing methods like DataPRM and demonstrates that rubrics mined from one model can benefit another, with code released for reproducibility.
BenchBench‑Cloning is a new benchmark for evaluating large language models on end‑to‑end molecular cloning tasks. It contains 65 tasks drawn from real sequence records, laboratory notebooks, and published studies, split into 50 protocol‑development tasks and 15 construct‑design tasks. Performance is measured by simulation execution for protocols and deterministic or rubric‑based checks for design, with GPT‑6 Astra achieving the highest score of 30.6% among eight tested systems.
TempShap introduces a Shapley-based framework that attributes importance to individual time‑step and feature cells in a time‑series forecast, providing both temporal localization and horizon‑impact profiles. Unlike traditional Shapley explainers that treat each feature as a single player, TempShap’s cell‑level approach captures when and how much a feature matters across the forecast horizon. The method is validated through planted temporal structures, exact cell games, and a closed‑form oracle for linear forecasters, and demonstrates improved deletion‑ranking performance over existing estimators.
The paper introduces ConMem, an algorithm that uses counterfactual causal inference to improve long‑term memory management in large language model agents. It builds a benchmark called MemAlign with 1,079 curated memory cases that capture various misalignment issues such as over‑generalization, misattribution, and privacy concerns. ConMem generates counterfactual interaction scenarios to audit retention decisions, leading to higher accuracy (75.16%) and better balance than existing methods while preserving personalization.
SkillShift-Bench is a new benchmark that evaluates web agents’ skill learning under controlled open-environment shifts, including Contextual, Perceptual, and Execution shifts across three self-hosted web systems. It measures task success, interaction cost, skill use, and skill failure, revealing that skill robustness degrades systematically—especially under Execution Shift. The benchmark also frames shifted skill execution as a nonstationary local executability problem, offering a dynamic-regret analysis and a test-time adaptation method to improve skill execution without relearning.
Graph-Dependent Pooling (GDPool) introduces a dynamic approach to hierarchical graph pooling, allowing each graph to determine its own pooling granularity rather than relying on fixed cluster numbers or ratios. The method employs self-attention to select graph-specific anchors and uses soft node assignment to build coarser graphs, addressing optimization conflicts identified through gradient alignment analysis. Experiments on metal‑organic framework and molecular graph datasets demonstrate that GDPool outperforms existing methods in predictive accuracy, modularity, and exhibits stable behavior across various hyperparameters.
Meta-CLIP is a metadata‑conditioned contrastive learning framework that incorporates DICOM acquisition information to reweight cross‑modal negatives during medical vision‑language pretraining. By defining a bounded pairwise margin over numerical and categorical DICOM fields, the method emphasizes distinctions among studies with similar imaging conditions while keeping the kernel fixed during optimization. Evaluations on chest radiography and volumetric CT show that Meta‑CLIP improves zero‑shot classification performance, achieving higher macro‑AUROC scores than baseline CLIP models on multiple benchmarks.
The paper questions the effectiveness of injecting the Unified Medical Language System (UMLS) into language models for medical question answering (MQA). It reviews current MQA studies, identifies methodological shortcomings such as limited UMLS coverage and weak subgraph retrieval, and finds that UMLS-based graph injection often yields only modest performance gains. The authors propose actionable recommendations to better harness UMLS as an external knowledge source.
ReGE is a framework that transforms recorded interactive task experience into guidance for new attempts. Its implementation, ReGE‑Syn, synthesizes execution records and artifacts into concise, self‑contained instructions using a single coach call, without updating model parameters. Experiments on four benchmarks show that ReGE improves accuracy by roughly 9–10 percentage points for GPT and Claude, and significantly reduces guidance‑construction tokens while maintaining success rates.
The paper introduces CTXAgent, a self‑controlled context management system for large language model agents that compacts conversation history when a token threshold is reached. It evaluates seven frontier models, comparing their ability to manage context with and without dedicated tools, and benchmarks performance across coding, scientific, and software engineering tasks. Results show that well‑managed models keep live context below 100k tokens, maintain high benchmark scores, and achieve significant cost reductions, with Claude Opus 5.5 outperforming its explicit configuration by 34% lower cost.
The paper investigates how to structure verification in agent harnesses that let large language models use tools and iteratively refine their outputs. It evaluates 17 verification designs across four lifecycle layers—initial instructions, mid‑turn reminders, completion gates, and independent review—on coding benchmarks (Terminal‑Bench and SWE‑bench) and a new VisualBench for visual artifacts. The results show that independent review by a critic consistently improves accuracy with modest cost increases, outperforming other designs and offering a practical strategy for test‑time scaling.
CSI-TextBench is a large‑scale dataset and benchmark that aligns WiFi Channel State Information (CSI) with natural language, combining 285,790 samples from three sensing benchmarks into 2.0 million CSI‑text training tuples. It pairs these samples with 2,291 hierarchically structured, label‑derived descriptions across seven tasks, enabling many‑to‑many alignment. Benchmarking five alignment methods shows that contrastive CSI‑text models can match supervised classifiers, generalize to unseen classes with few samples, and answer compositional queries, achieving 86.9% top‑1 accuracy on 140 composite scenes.