Skip to content
AI NEWS JOURNAL
AI, IN PERSPECTIVE

Search AI news

Search headlines, summaries and articles to connect the dots.

Newest first

From Preference Data to Personalization: Tracing Sycophancy in Large Language Models

arXiv:2611.21650v1 Announce Type: new Abstract: Sycophancy in large language models is often measured as agreement with the user, yet agreement can be appropriate when the user's claim is valid, their preference is genuine, or their experience warrants acknowledgment. We study sycophancy as inappropriate agreement: validating a claim, preference, or self-assessment when the context calls for correction, qualification, or resistance. We introduce a validity-aware sycophancy grader and SycoDrift, a persona-targeted evaluation set of 3,298 validity-annotated probes with controlled annotations for user openness and stakes. We use this framework to analyze three sources of agreement pressure: human preference data, automated judge specification, and in-context personalization. First, preference for validating responses in human datasets (Chatbot Arena, HH-RLHF, Community Alignment) is weak and concentrated where agreement is often appropriate. Second, in a selected-value, single-model experiment, emphasized judge rubrics produce consistent but modest increases relative to paired suppressed rubrics, demonstrating a plausible mechanism for training-time amplification. Third, in off-policy evaluations across 14 models, inference-time personalization increases inappropriate agreement, with the largest effect from matched user profiles placed directly before the query.

Availability-Aware Historical Feedback Learning for Electric Arc Furnace Temperature and Chemical Composition Prediction

arXiv:2611.17463v1 Announce Type: new Abstract: Electric arc furnace (EAF) steelmaking plays an increasingly important role in recycled and lower-carbon steel production. However, variability in charged materials, energy input, and operating conditions makes accurate prediction of molten steel temperature and final chemical composition challenging. Intermittent or delayed measurements limit the information available for timely prediction. This study develops an availability-aware historical-feedback learning framework for predicting EAF endpoint temperature and the final concentrations of twelve chemical elements. Eight industrial data sources are integrated into three progressively enriched predictor representations comprising 19 static, 39 temporal, and 47 history-informed features. Historical outcomes are incorporated only when eligible under an assumed 60-minute measurement-release delay. Using 19,738 temperature heats and 3,341 chemistry heats, twelve tuned regression algorithms are evaluated through chronological forward validation. The proposed historical-feedback representation reduces root mean squared error (RMSE) in 147 of 156 matched algorithm-target comparisons. For temperature prediction, ExtraTrees achieves an RMSE of 12.290 degrees C, representing a 5.54 percent reduction compared with its temporal-only counterpart. Historical-feature ablation and feedback-delay sensitivity analyses demonstrate the importance of recent outcome information and its availability. In a separate retrospective later-period evaluation, lightweight adaptive residual correction reduces temperature RMSE from 12.502 to 12.275 degrees C without retraining the underlying model. Exploratory deep learning comparisons and target-specific feature-attribution analyses provide additional insights into predictive behavior. The results establish historical outcome availability and sequential feedback as important considerations for EAF endpoint prediction.

When Should LLM Agents Forget? Memory Filtering Under Query-Critical Resolver Errors

arXiv:2611.17338v1 Announce Type: new Abstract: LLM memory systems often delete facts classified as stale. We show that this hard-filtering step can reduce downstream accuracy when resolver errors delete query-critical facts. Across three models, a controlled sensitivity intervention moves hard-drop from gains of +0.186 to +0.230 at perfect resolution to losses of -0.176 to -0.181 at the lowest tested reliability. Paired targeted interventions show that false-drop causes significantly greater degradation than false-keep, while a non-deleting baseline can still fail when incorrect labels steer generation. On a frozen 102/102 calibration/test split, a simple deployment-matched gate reaches 0.655 accuracy versus 0.641 for always hard-drop and 0.662 for test-informed oracle selection. Global-accuracy crossovers require gold labels and are model- and task-specific diagnostics, not deployment thresholds. Deletion should instead be treated as cost-sensitive: validate query-critical false-drop risk on matched data, or default to label-free retention when such validation is unavailable.

SynthCorp: A Platform for Creating Benchmarks for Data Agents

arXiv:2611.14765v1 Announce Type: new Abstract: Current benchmarks for LLM-driven data agents saturate quickly and cannot target specific capabilities. We present SynthCorp, a platform that creates customizable data estates and benchmarks by simulating how business events produce and modify corporate data over time. Each data estate combines relational databases, a document store, and a data catalog, paired with evaluation tasks and executable gold programs. First, given a scenario and target complexity dimensions, SynthCorp synthesizes reusable event types and compiles each into a program over predefined simulation primitives. Second, SynthCorp schedules event instances on a virtual calendar, interleaving routine and disruption events so that tables, documents, and catalog entries remain consistent. Third, SynthCorp executes the calendar to materialize the data estate and records every operation in a simulation log. Fourth, SynthCorp generates analytical tasks and derives verified gold programs from the log. We use SynthCorp to generate initial evaluation benchmarks comprising seven data estates (up to 1K operational tables) and 256 tasks across five domains. The data estates pass deterministic pipeline checks (such as foreign-key violations, undeclared writes, recurrence violations, and database execution failures) with a near-zero error rate, introduce cross-asset complexities, and cover all requested dimensions. Evaluating three models across four data-context settings shows that the benchmarks separate data contexts and model tiers. Auditing candidate queries against causal provenance in the simulation log verifies gold programs without manual annotation, including on multi-event composite tasks that expose reasoning failures in state-of-the-art agents. We hope the community uses SynthCorp to generate more challenging and useful data estates and benchmarks that foster the development of enterprise data agents.

Linear Probes to Bayesian Manifolds: Geometry of Arithmetic in Large Language Models

arXiv:2611.12998v1 Announce Type: new Abstract: Linear probes are a standard method to study geometric structures, such as helices and circles, from the activations of large language models (LLMs), and such structure is often taken as evidence that the model computes with it. Probe accuracy alone cannot establish this, because it conflates four separate claims. These are (1) that the concept is decodable from the activations, (2) that they form a specific shape, (3) that the shape belongs to the probed concept rather than to a related concept the model also represents, and (4) that the output of the model depends on the shape. We call these claims decodability, shape, ownership, and causal use, and we test them separately with a four-stage Bayesian pipeline. Stage 2 compares candidate shapes and selects the best one only when its evidence passes a predefined threshold. Stage 3 removes a set of related concepts, specified in advance, and tests whether the shape is owned or inherited. Stage 4 intervenes on the shape to test whether the output of the model depends on it. We apply the pipeline to two-digit addition and multiplication in GPT-J 6B, Llama 3.1 8B, and Pythia 6.9B. Nearly every concept is decodable, but the decoded directions overlap substantially across concepts. Every strongly supported comparison selects a ribbon, which is a helix whose radius changes with the value, or a torus. Of 1,330 cells with an ownership verdict, 19% of shapes are owned and 81% are inherited. Owned and inherited shapes move the output about equally often. Intervening on the shape retains approximately 61% of the effect of intervening on the full probe while using at most five dimensions, against a median of nine for the probe. Our results indicate that a probe verdict alone overstates what a model computes with, and we propose that probe results be reported as verdicts on all four claims rather than as a single accuracy.

MICRO: A Surgery-Phase Retinal Image Rigid Registration Benchmark and One-Step Frequency-Structure Similarity Registration Learning

arXiv:2611.13045v1 Announce Type: new Abstract: Precise multimodal retinal image rigid registration (MRIRG) is essential for aligning cross-modal vessel structures, offering critical guidance for ophthalmic surgery. Despite its significance, the advancement of MRIRG is hindered by the lack of publicly available multimodal retinal image rigid registration datasets spanning various ophthalmic surgery phases (e.g., preoperative, intraoperative, and postoperative). To address this issue, we introduce the MICRO dataset, a large-scale MRIRG benchmark, comprising 1,365 cross-modal retinal image pairs with standardized annotations across all ophthalmic surgery phases. Moreover, existing deep registration methods achieve competitive MRIRG results via a two-step segmentation-then-detection framework, but segmentation quality heavily constrains their performance. We argue that this dependency stems from an implicit need for keypoint topology constraints, which have not been exploited for MRIRG through a one-step registration framework. Inspired by the above analysis, we propose a novel One-step Frequency-Structure Similarity Registration (OFS2Reg) learning framework for MRIRG that leverages the potential of cross-modal vessel structure similarities to construct keypoint topology constraints under a shared detection network. To be specific, OFS2Reg applies a well-designed frequency-guided structure preservation descriptor to construct coarse keypoint topology constraints in descriptors by highlighting significant vessel structures. Additionally, we develop a new cross-modal vessel structure similarity learning method that refines cross-modal keypoint topology constraints to precisely align pixel-level cross-modal vessel structures by leveraging cross-modal vessel structure similarity priors.

CulturalCrosswords: A Multilingual Benchmark comprised of Answer Constraints and Grid Completion

arXiv:2611.10044v1 Announce Type: new Abstract: When several answers fit a riddle, exact-match scoring can penalise a valid response. CulturalCrosswords tests how answer constraints affect large language models (LLMs) solving cultural riddles, whose solutions draw on knowledge specific to a community. We constructed 53 crosswords from crowdsourced riddles in 42 languages and evaluated 11 LLM configurations on five versions of each clue set: plain batches, batches with lengths, batches with lengths and partial-answer hints, and crosswords requesting an answer list or a completed grid. On 962 clues with alternative answers, lengths improve agreement with the reference by 22.8 percentage points, compared with 8.4 points when accepted alternatives also receive credit. Part of the gain therefore appears to come from selecting a required answer form. Crossword input increases the share of all crossings with compatible answers by 9.7 points; requiring a completed grid reveals large differences between models. The benchmark distinguishes agreement with a reference from the ability to produce answers that fit together and occupy the correct cells.

DramaGrow: A Regime-Aware Benchmark for Auditing LLM Judges of Multi-Episode Drama

arXiv:2611.09938v1 Announce Type: new Abstract: LLMs can now draft complete multi-episode dramas, and LLM judges are increasingly used to evaluate them. Assessing the reliability of these judges is difficult because drama quality encompasses specification-backed errors, reproducible preferences, and legitimate disagreement among human raters. Most existing evaluations reduce these cases to a single score from one judge call and leave tie calibration, call-to-call variation, and the vote margin underlying the human reference unreported. We introduce DramaGrow, a regime-aware benchmark and audit protocol for AI-generated Chinese five-episode drama. The benchmark assigns verifiable constraint violations, four-of-five preferences, and within-panel disagreements to separate denominators and evidence rules. It contains 800 episodes, 120 byte-preserving interventions, and 1,155 blinded five-rater records, including a frozen extension with 114 supermajority pairs and 21 disagreement pairs. On the frozen sets of 114 preference pairs and 35 constraint interventions, the repeated absolute-score rubrics yield overlapping performance, with mean accuracies of 53.8-55.8% and three-call agreement of 29.8-32.5%, and their accuracy falls from 71.6-71.9% to 34.8-37.1% as the tie band widens. DirectPair-GLM reaches 85.1% preference accuracy, 89.5% constraint F1, and 81.9% Detect+Locate on the same items, and exact subpanel enumeration identifies four-of-five as the smallest stable human vote margin. Evaluation protocol, tie calibration, judge-call variation, and panel margin therefore affect performance as much as rubric design. We will release the specification packages, scripts, interventions, anonymized human votes and rationales, frozen judge runs, and analysis code.

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

arXiv:2611.10327v1 Announce Type: new Abstract: 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.

Not All Misalignment Is Alike: A Decomposed Benchmark for Emergent Misalignment

arXiv:2611.08569v1 Announce Type: new Abstract: Emergent misalignment (EM), in which fine-tuning on a narrow task induces misaligned behaviour on unrelated prompts, has been demonstrated across a wide range of datasets, model families, and training setups. Yet, most studies continue to rely on the evaluation approach introduced in the original EM paper, producing rates that are sensitive to the small set of evaluation questions used, the choice of judge, and the thresholds applied. These evaluations can also conflate misalignment with capability loss and offer little insight into the specific ways misalignment manifests. Differences in datasets, models, and fine-tuning configurations further make findings difficult to compare across studies. We introduce EMERGE, a decomposed benchmark that tracks changes in eight distinct misaligned behaviours from base to fine-tuned models. It evaluates these behaviours on out-of-domain prompts against matched benign controls and tracks capability, coherence, and refusal to account for potential confounders. Applying EMERGE to 8 open-weight models and 13 fine-tuning datasets, we find that models with similar broad misalignment rates can exhibit very different behavioural profiles that a single score cannot capture. We hope EMERGE can help establish a shared and robust standard for measuring EM, making findings more comparable and helping move research from detecting the phenomenon towards understanding it.

A Night at the LLM Zoo: Data Conditioned Flow Matching for Generating Pretrained LLM Weights

arXiv:2611.07103v1 Announce Type: new Abstract: A challenge to Weight Space Learning as it expands from task specific and vision models to Large Language Models (LLMs) is the absence of a well curated and analyzed LLM population. To address this, we train such a model zoo for application to Weight Space Generation by systematically varying the data mixture composition between web text, books, mathematics, coding, and French across populations of 51M, 124M, and 355M parameter GPT-2 Models, each tier with count N = 100. We analyze these populations by cosine similarity, functional diversity, Leave-One-Mixture-Out modeling, and variance decomposition. We show that the zoo contains diverse models whose variation is structured by training mixture composition. We embed the training mixture composition to condition a Flow Matching model of complete pretrained LLM weights in Gram-PCA compressed weight space. By demonstrating zero-shot generalization to held out mixtures, measuring Continuous Ranked Probability Score skill, as well as finding minimal changes in downstream test-set perplexity and general capability benchmark scores, we show that our generator generalizes on the simplex of data mixtures to produce models that perform similarly to those in the zoo ex nihilo. Finally, we produce all source code and artifacts to support the continued expansion of Weight Space Learning and Generation into the LLM domain.

KDASO: Knowledge-Data Dual-Driven Automated Skill Optimization for Liability Adjudication Tasks

arXiv:2611.06972v1 Announce Type: new Abstract: Liability adjudication tasks (e.g., legal charge determination) require analyzing event descriptions against domain-specific skills to determine whether a party should bear liability. While Large Language Models (LLMs) have shown promise in this area by leveraging executable skills for interpretable reasoning, existing approaches face three critical limitations: skill libraries become stale as real-world scenarios dynamically evolve, skill descriptions often suffer from a semantic execution gap where LLMs deviate from intended expert logic, and growing skill collections exceed context budgets causing critical skill omission. To address these challenges, we propose Knowledge-Data Dual-Driven Automated Skill Optimization (KDASO), which continuously evolves a skill library through neighborhood-conditioned induction from recent adjudication data, execution-aligned refinement based on LLM trajectory analysis, and context-budgeted routing for scalable skill selection. Experiments across ride-hailing adjudication, legal liability determination, and logical reasoning judgment demonstrate that KDASO significantly outperforms current general-purpose and domain-specific baselines. KDASO further achieves 8.96% improvement when transferring skills to concept-similar scenarios and maintains consistent performance gains through continuous optimization, underscoring the high transferability and stability of generated skills.

NUMA-Torch: Compiler-Driven NUMA-Aware LLM Prefill on Multi-Socket CPUs

arXiv:2611.07105v1 Announce Type: new Abstract: For sufficiently large transformer-based Large Language Models (LLMs), whose parameters cannot be retained in on-chip caches, prefill throughput on multi-socket CPUs is often limited by sustainable DRAM bandwidth rather than peak floating-point throughput. In the standard PyTorch compilation (i.e., TorchInductor) path, model weights are allocated without considering the Non-Uniform Memory Access (NUMA) topology: specifically, the affinity between memory banks and compute cores is disregarded. This causes a substantial fraction of weight accesses to traverse the inter-socket interconnect, severely degrading the effective memory bandwidth available to each socket. We address this bottleneck through a compiler-level solution that makes TorchInductor NUMA-aware. Treating each NUMA socket as an independent tensor-parallel partition, we introduce NUMA-aware Tensor Parallelism (NUMA-TP). This approach transparently shards weights across sockets and schedules compute kernels such that thread pools and memory allocations remain strictly local to their respective sockets using CPU affinity. We evaluate the approach on GPT-2 (117M, 345M, 762M) and BERT (Small, Large), which exercise standard multi-head attention and MLP, as well as Gemma 3 270M, which combines Multi-Query Attention with a GeGLU-gated MLP. By strictly enforcing data locality, our approach successfully mitigates cross-socket traffic, achieving prefill speedups of up to 2.63x on a dual-socket Haswell platform, up to 1.87x on a dual-socket Cascade Lake platform, and up to 2.27x on a four-socket Haswell-EX platform.

ProcessPrefBench: Evaluating Cross-Task Transfer of Individual Process Preferences from Naturalistic Human-AI Interaction

arXiv:2611.06065v1 Announce Type: new Abstract: In natural human-AI dialogue, users' corrections, requests, and self-descriptions carry reusable signals about how they want work to be done. These signals reveal individual process preferences, user-specific standards for evidence, depth, analytical focus, procedure, and delivery. We ask whether a preference revealed in one dialogue can shape an assistant's later answer when the task does not restate it. ProcessPrefBench uses three months of natural dialogues from two knowledge workers to build 66 source-target cases separately judged for source evidence and target applicability. Sources are preselected, so the benchmark tests application rather than retrieval. We test three models with target-only, full-history, and distilled-preference prompts. Target-specific 0-2 rubrics score preference realization, and a task gate flags severe failures. Compared with target-only prompting, full-history and imperative distilled-preference prompting raise preference realization by 0.204 and 0.244 points on the 0-2 scale, respectively, with no reliable advantage from distillation. Fabricated-evidence failures rise from 5.8% under target-only prompting to 10.1% under full-history and 10.4% under imperative prompting. ProcessPrefBench highlights a tension in learning from dialogue, where drawing on prior interactions can improve preference realization yet also invite fabricated evidence. We hope it helps build assistants that honor a user's process preferences without sacrificing factual reliability.

Assurance-Conditioned Continuous Authorization for Autonomous AI Agents

arXiv:2611.05632v1 Announce Type: new Abstract: Autonomous software agents can act through tools, delegate work, and change their models or execution environments while earlier authorizations remain in flight. We present Assurance-Conditioned Continuous Authorization (ACCA), an approach designed to be harness-independent, for maintaining operational authority according to currently justified assurance. It distinguishes evidence validity from contextual applicability, derives assurance claims from eligible evidence, and resolves a multidimensional effective-authority envelope within a principal-declared ceiling and applicable policy. A typed dependency graph supports selective re-evaluation of authorization-relevant changes. An authority epoch changes when effective authority changes; a distinct optional authorization-basis version supports revalidation of action-specific authorization conditions. Protected effect-time enforcement can reject stale actions before a consequential resource change, subject to the available commit mechanism. A compositional Domain Governance Profile expresses domain-specific semantics without granting authority. Controlled scenario tests implemented as a regression suite using a trading profile exercise default-deny rules, evidence-conditioned contraction and recovery, policy divergence, selective non-response, and stale-action rejection at a local gateway before the simulated-broker boundary. The study establishes implementation-level evidence in one domain, not a cross-domain guarantee of harmful-behavior prevention.

Verified but Vacuous: Three Failure Surfaces of LLM-Driven Verified Code Generation

arXiv:2611.05590v1 Announce Type: new Abstract: What should ``verified'' mean when evaluating LLM-generated code? Many existing pipelines report a single verification number. We argue this collapses three distinct properties measured in a conditional audit sequence: verifier acceptance (Surface A), \emph{syntactic specification binding (Surface B: does the proof obligation bind to the spec?), and \emph{semantic specification fidelity} (Surface C: does the spec capture the informal requirement?). Two empirical demonstrations show the surfaces separate. \textbf{First ($A \neq B$), via Li@Opus (a same-model reproduction of \citet{li2025dafny}'s Dafny-as-intermediate-representation (IR) pipeline under Claude Opus 4.7) on three Python benchmarks}: it reaches \textbf{99.1\% verifier acceptance} on HumanEval while only \textbf{1.9\%} carry a certificate binding to a spec-defined predicate or function. \textbf{Second ($B \neq C$), via VCG on two Rust benchmarks}: we use VCG (which is our own in-house Dafny-as-intermediate pipeline) on Rust because the Rust target lets us enforce fully-verifiable output (every component ships as a Dafny certificate, ensures-binding gate passing by construction). Even inside that construction, \textbf{11.48\% (MBPP-Rust) and 14.82\% (VeriContest LC slice) of structurally-verified samples still fail hidden tests}. On the LC slice of VeriContest, \textbf{VCG's reported End2End rate is 7.79\% (535/6870).} The cited external rates are 5.51\% (GPT-5.5) and 1.88\% (single-shot Opus 4.7). These give descriptive differences of 2.3 percentage points and 4.14$\times$, respectively; task-budget, repair and harness differences prevent a matched-performance claim.

Co-Occurrence Priors Override Visual Spatial Evidence: Understanding False Alarms in Vision-Language Models for Construction Struck-By Hazard Detection

arXiv:2611.04112v1 Announce Type: new Abstract: Vision-language models (VLMs) increasingly support construction hazard monitoring, yet their frequent false alarms remain poorly understood. This study investigates whether false struck-by hazard warnings arise from failures in object recognition, spatial relation encoding, or hazard decisions dominated by learned co-occurrence priors. We evaluated Qwen2.5-VL-7B and InternVL3.5-8B using 355 counterfactual worker-excavator scene pairs with controlled spatial proximity and worker-removal variants, combining behavioral analysis, internal representation probing, causal interventions, and decision correction. Both models recognized the entities and encoded their spatial relations (out-of-fold AUROC: 0.82 and 0.91), assigning higher hazard scores to near than far scenes in 79% and 73% of pairs. Nevertheless, 74% and 100% of far scenes were falsely classified as hazardous. Removing the worker reduced Qwen's hazard rate to 11%, while worker presence exerted 2.4- and 3.2-times greater influence on hazard scores than spatial proximity in the respective models. Attention-head interventions, neuron-level interventions, and activation steering failed to eliminate this co-occurrence bias. However, conditioning hazard decisions on the model's own spatial-relation representation improved Qwen's specificity from 0.36 to 0.58 at matched sensitivity, outperforming threshold recalibration. These findings reveal that VLM struck-by false alarms stem primarily from decision-making failures rather than deficiencies in object recognition or spatial encoding, suggesting that future efforts should prioritize conditioning hazard decisions on encoded spatial evidence rather than further improving visual recognition alone.

Do We Still Need Rotation-Invariant Architectures? A Statistical Robustness Benchmark for Point Cloud Classification

arXiv:2611.03952v1 Announce Type: new Abstract: A substantial line of work builds rotation invariance directly into point cloud architectures. We benchmark 10 architectures (PointNet, PointNet++, DGCNN, RIConv, RIConv++, RISurConv, PointMLP, PointTransformer, and the equivariant VN-PointNet and VN-DGCNN) across 5 domains under pose, noise, and occlusion, and analyze the results with confirmatory hypothesis tests rather than point estimates. The three rotation-invariant (RI) architectures achieve the best pose robustness but rank among the lowest on noise robustness in 4 of 5 domains. We ask whether this cost is intrinsic to invariance or specific to how invariance is obtained. We retrain all 10 architectures unchanged, with a simple invariance loss on the embedding and three training seeds each, and evaluate all models on the independently designed ModelNet40-C. Invariance training transfers narrowly but reliably. It improves exactly the corruption families its perturbations target wherever the baseline has headroom (mean occlusion +10.5, lidar +13.4 percentage points of accuracy). At matched training budget the best standard backbone and the best RI architecture end on par on unseen corruptions (within 2.5 percentage points in every seed), and invariance-trained RIConv++ is the most seed-stable top model. Once rotation augmentation is widened, standard backbones concede only a flat, angle-independent accuracy offset under full-sphere rotation. We release the benchmark along with a package for analyzing point cloud robustness and reproducing our results.

LOLPlay: Benchmarking Real-Time Visual Agents in League of Legends

arXiv:2611.03685v1 Announce Type: new Abstract: Visual agents need to coordinate reasoning and action while their environment evolves during inference. We introduce LOLPlay, a benchmark in native \emph{League of Legends} with 36 champion-specific tasks spanning resource management, combat, cooperation, and objectives. Agents interact through screenshots and mouse--keyboard actions; automated initialization and 82 executable subgoal checks distinguish partial progress from completion. Across 324 task evaluations of three models at three reasoning-effort settings, GPT-6 Astra achieves the highest task success rate of 25.0\% at low and medium effort, and the highest subgoal success rate of 40.3\% at medium effort. However, high effort improves neither metric over low effort for any model. For Astra, longer requests leave fewer interaction rounds, suggesting a cost to additional reasoning in real-time control. Trajectories reveal incomplete subgoal composition and actions executing after the observed scene changes, motivating agents that combine visual reasoning with timely action.

ScienceVision: Can AI Observe and Reason Like a Scientist?

arXiv:2611.04087v1 Announce Type: new Abstract: Scientific discovery builds on accurate observations interpreted through domain knowledge and reasoning. For scientific 3D data, this requires understanding structures and spatial relationships, from protein contacts to anatomical organization. Existing multimodal benchmarks largely evaluate text and 2D images, while scientific 3D evaluations vary in representations and protocols across disciplines, limiting cross-disciplinary comparison. To assess these foundational capabilities, we introduce ScienceVision, a cross-domain benchmark for scientific 3D understanding and reasoning through expert-level visual question answering (VQA). We construct 250 questions, with 50 each from Biology, Chemistry, Medical Imaging, Mechanical CAD, and Architecture, through a pipeline combining synthetic question generation with domain-expert review and annotation. Each task pairs a native 3D artifact with a scientific question, a specified response format, and a task-specific reference answer, requiring domain knowledge to be applied to geometric or topological evidence. We evaluate 12 models on the same questions in three settings: visual question answering with native 3D artifacts accessed through an agent harness's file and code-execution tools; the same workflow augmented with expert-inspired viewer inspection and measurement tools exposed through the Model Context Protocol (MCP); and visual question answering from prescribed rendered 2D images using an agent harness. All settings use the same task-specific deterministic final-answer scoring. The inspection environment supports view changes, target isolation, and slice inspection. In this main comparison, the best mean answer score is 0.53 out of 1. Mean scores in the native-artifact workflow without additional MCP tools exceed those in the rendered-image workflow for every model, by 17.2 percentage points on average.