Transformer
-
Learning to Absorb Corrections: Critic-Weighted Residual Flow for Vision-Based Insertion
2026-10-21 03:00:32Yiqing Yuan, Zhi Li, Jinrong Yang, Mingshan Sun, Mei Yu Zhi, Yong Zhao, Liangliang Ren, Hui ChengAbstract
Vision-based insertion requires correcting alignment errors while retaining useful behavior learned from demonstrations. We present ACT--Flow, an online learning framework that couples an adaptive Action Chunking Transformer (ACT) with a conditional residual flow policy. Following anchor-guided warmup, ACT absorbs successful executed actions through behavioral cloning, while residual targets are recomputed relative to the updated prior. A chunk-level critic learns from successful and failed interactions and evaluates final composed actions. Its estimates weight generated residual candidates for flow matching, complemented by supervision from successful real executions. Candidate scoring is confined to training; deployment uses RGB observations and samples one residual per action chunk. We evaluate the framework on power-plug, USB, and key insertion, together with cable routing and needle threading. Results show a five-task mean success rate of 66% at 20,000 post-initialization interaction steps, exceeding the strongest baseline by 18 percentage points at this checkpoint. Evaluations at matched checkpoints with the residual disabled show ACT's mean success increasing from 16% to 76%. Component ablations further support the complementary roles of behavior absorption, value weighting, and successful-action supervision.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29117
PDF
-
Agentic Music Post-Production
2026-10-21 02:36:51Jaywon Koo, Junghyun Koo, Marco A. Martinez-Ramirez, Wei-Hsiang Liao, Seungheon Doh, Qiyu Wu, Vicente Ordonez, Yuki MitsufujiAbstract
We introduce MixAgent, an agentic pipeline for music post-production that follows the workflow of audio engineers to build interpretable and editable mixing plans guided by a reference song. A grouping agent first routes related input tracks, such as drum recordings, into subgroups. A planning agent then uses retrieved examples to choose and order audio effects, such as volume control and reverb, for tracks, subgroups, and the full mix. To refine the mix, a detector identifies what needs correction, and a critic proposes numerical parameter adjustments from tracks to the full mix. In our experiments, prompting recent multimodal models to listen and revise often worsens the mix, making reliable correction a central challenge. We therefore build MixCritic, a dataset of 45,657 repair critiques derived from parameters fitted to professional mixes, and train the critic with supervised fine-tuning followed by preference optimization on rendered audio outcomes. On MedleyDB, learned refinement reduces error by 19% relative to the initial mix, with preference optimization reducing it by 8.4% over supervised training alone. Retrieved examples further improve effects selection, and a human listening test shows that MixAgent gets higher quality ratings than mixes produced by Qwen3-Omni, Gemini 3.8 Flash and GPT-4o-audio.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29093
PDF
-
Copying Under Repetition: From Prefix Matching to Position Hashing
2026-10-21 02:21:26Anxin Guo, Tianlong Huang, Chenxiao Yang, Zhiyuan LiAbstract
Frontier LLMs can copy random text almost flawlessly, yet they often overrun or truncate long blocks of repeated characters. In repeated strings, matching local content no longer reliably identifies where copying should continue or when it should stop. Nevertheless, we show that a two-layer transformer with $O(\log\log N)$-bit precision and $O(\log N/\log\log N)$ width can copy every string up to a chosen design horizon $N$ and terminate correctly under arbitrary repetition. We use RoPE-style rotations of attention values and outputs (VO-rotations) with small prime periods to encode distances from delimiter anchors modulo those primes. By the Chinese Remainder Theorem, these periodic encodings jointly distinguish all required distances, allowing one retrieval rule to select either the next source token or the end marker. In contrast, we prove a width-independent impossibility result for copying under sequential finite-precision softmax accumulation. It applies to transformers whose only positional input is an additive attention bias that eventually becomes constant, including NoPE and a recursive implementation of ALiBi. Once an identical-token run exceeds the resulting saturation threshold, further repetitions leave the logits unchanged. If this threshold grows sublinearly with $N$, these models cannot both complete a long all-identical copy and stop correctly, a pattern resembling the termination errors of frontier LLMs. In synthetic copying experiments, baseline transformer models fail on long repeated strings even within the training range. Trained VO models, by contrast, generalize beyond the training lengths with near-perfect accuracy up to length $\approx5005$, the selected primes' period product.
Abstract (translated)
URL
https://arxiv.org/abs/2610.26722
PDF
-
Single-Image Food Mass Estimators Do Not Transfer Their Gram Scale Across Datasets
2026-10-21 00:59:03Yongsheng LianAbstract
Estimating food mass from a photograph is central to image-based dietary assessment, but whether trained models transfer to new cameras and foods is unclear. We fine-tuned DINOv2-base on 2,253 weighed Nutrition5k dishes to predict mass from one RGB photograph and compared it with a zero-shot vision-language model, Qwen3-VL-32B, on Nutrition5k and three external datasets: ASA24 studio portions, ECUSTFD whole foods and Food Portion Benchmark meal trays. In domain, the fine-tuned model's mean absolute error was 24.8 g, less than half of Qwen's. On external datasets it did not beat Qwen. For a food it had never seen, it ranked portions by size, but no better than a count of food pixels. Its gram scale was wrong in two ways: it changed with how much of the scene the photograph showed, and it was off by a different factor for each food. Training on other foods or trays from the tested dataset cut the external error by 38 to 65%. DINOv2-large, a larger backbone trained on the same dishes, did not reduce the external error. One labeled photograph of the same food cut the model's ASA24 error to 28.3 g, better than that photograph achieved alone. Prediction intervals that held the true mass for 90% of Nutrition5k dishes held it for at most half of external photographs. Nutrition5k often photographed the same plate several times, so a random test split shared near-identical photographs with training and overstated accuracy. Single-image mass estimators therefore need labeled photographs from the setting where they are used.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29025
PDF
-
Target Regeneration Shrinks the Gains of Draft-Training Objectives in Speculative Decoding
2026-10-21 00:26:37Jungseob Lee, Hogun Seong, Euijun Choi, Sugyeong EoAbstract
Speculative decoding accelerates large language models by verifying drafted tokens in parallel. New draft-training objectives are usually credited with the acceptance they add over cross-entropy on externally written responses, yet many pipelines let the target rewrite them. We ask how much of that gain survives. With a vision-language target and a block drafter, we train three objectives on four versions of one corpus: cross-entropy on corpus tokens, token-level distillation from the target, and distillation gated to positions that would survive verification. Distillation and the gate add large gains on external responses, but after greedy regeneration, distillation adds a smaller gain and the gate little or none. The shrinkage persists across data orders, position weights, verifiers, decoding temperatures and, for distillation, initializations and learning rates. Most of it remains at matched supervised length and at each objective's best measured training length. A second vision corpus shows the shrinkage, as does a text-only replication, for distillation also with an autoregressive drafter. Objective gains should therefore be measured against a target-regenerated baseline and across training lengths. Our code is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29013
PDF
-
Enhancing Cross-Grade Generalization in Banana Variety Classification via Hybrid CNN-Transformers and Knowledge Self-Distillation
2026-10-21 00:19:09Khush Kataruka, Mohammad Muzammila, Zahid Akhtar, Raghavendra Ramachandra, Praveen Kumar ChandaliyaAbstract
In recent years, automated fruit classification systems have become a more critical component in precision agriculture for efficient yield estimation, quality control, and supply chain optimization. However, they still face significant challenges under varying quality, fruit maturity, illumination, texture degradation, morphological attributes, domain shift, and post-harvest handling conditions. In this study, we address the challenging issue of cross-grade banana variety classification. This is a real-world setting where models that have been trained on specific quality grades must generalize to unseen grades to handle significant domain shift. Under a rigorous Leave-One-Grade-Out evaluation protocol, we present a thorough comparative analysis of CNNs, ViTs, and dual-stream hybrid architectures using a large-scale dataset of 31,678 banana images with three varieties and three quality grades. Specifically, to address the drawbacks of single-paradigm frameworks, we present a dual-stream Swin ResNet hybrid framework enhanced with KL-divergence-based self-distillation to align probabilistic feature representations across the CNN and Transformer streams. Comprehensive experimental analyzes show that the presented hybrid framework with KL-divergence based self-distillation achieves better overall performance, i.e., it attains a mean accuracy of 81.25% and significantly surpasses CNN and Transformer baseline models. Qualitative analysis further confirms that the hybrid framework in this study improves discriminative feature transferability, improves attention localization, and stabilizes optimization.
Abstract (translated)
URL
https://arxiv.org/abs/2610.29008
PDF
-
Can Quantum Layers offer Performance and Security Advantages in Vision-Language Models?
2026-10-20 22:28:17Md Khalid Syfullah, Alvi Ataur KhalilAbstract
As quantum computing moves from the noisy intermediate-scale quantum (NISQ) era toward early fault-tolerant quantum computing (EFTQC), quantum-enhanced AI is shifting from a conceptual idea to experimental research. Quantum components are expected to extend the capabilities of existing AI systems, yet it remains unclear how such components will help within large multimodal models. In this study, we try to answer that question for vision-language models (VLMs) using classically simulated quantum layers. With LLaVA-OneVision as the backbone, we evaluate 4- and 8-qubit basic-ring and strongly entangling circuits at four insertion and four replacement locations. Insertion adds quantum layers to the classical model, while replacement substitutes them for classical components. Using the same setup, we compare performance and computational cost between classical and quantum-classical hybrid models. On the Visual Spatial Reasoning dataset, inserting a quantum layer after the visual projector reaches 80.7% development accuracy, 2.28 percentage-points above classical full fine-tuning. Replacing the final decoder feed-forward sublayer retains 79.22% accuracy with 13.07 million fewer trainable parameters. Quantum insertion can thus improve accuracy, and selective replacement can reduce parameters. These benefits depend strongly on where the quantum layer is inserted or which classical component it replaces: earlier replacements lower accuracy, and fewer parameters do not reliably shorten training or reduce computation. To assess potential security benefits, we evaluated the robustness of quantum-classical hybrid models under six attack settings. Overall, our study maps where quantum layers help and hurt within VLMs, offering a starting point for hybrid VLM design; whether these gains generalize remains open. The code is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28964
PDF
-
Fooling LLM Safety Guards into Judging the Wrong Text
Abstract
Safety guards check content before an LLM application acts on it, and most paste that content into a plain-text prompt. Content that imitates the prompt can append a second task, and the guard answers that one instead. We call this target rebinding. Our explanation is that the guard grades the last complete task in its input, here a harmless decoy, so the harmful request goes through unchanged. In a pre-registered test, an appended forged query and decoy clear all 75 held-out AdvBench requests that Shieldstral-1.0-3B flags. The guard is not persuaded: a directive to ignore the text above clears none of them, and in a follow-up run a harmful decoy stops the attack. On our GuardRebind benchmark, each of five guards has an attack family, chosen per guard, that clears at least 93% of its flagged items. Rebinding explains this best on Shieldstral and gpt-oss-safeguard; on ShieldGemma and Llama Guard 3, generic evasion and trailing text respectively do part of the work, and Qwen3Guard falls mainly to a forgery of its whole task. Training the guard to score content no lower than its most harmful part cuts attack success from 75% to 0.8-1.6% on attack families, decoys and forged queries held out of training. All of these attacks append the forged structure after the request, and we added the held-out-decoy evaluation after a first round of test runs. Clean AUROC falls by at most 0.01, but our test of non-inferiority is inconclusive. A score-guided search still beats this trained guard on 47 of 50 items, usually by placing forged structure before the request. On about half of the items where the search wins, the same search also finds benign text, with no forged structure, that clears the guard.
Abstract (translated)
URL
https://arxiv.org/abs/2610.31695
PDF
-
Policy-Mask: Budget-Aware Adaptation of Diffusion Features to Discriminative Tasks
2026-10-20 20:13:31Ivan Listopadov, Maksim Nakhodnov, Aibek AlanovAbstract
Selecting intermediate features from a frozen diffusion model for discriminative tasks is non-trivial because the useful signal is distributed across the denoising trajectory, network depth, and channels. In practice, this requires choosing a timestep, a block, and a channel subset. Existing methods either fix these choices by hand or optimize only a few extraction hyperparameters, most often the timestep, then reuse the same rule across inputs and budgets. We introduce Policy-Mask, a budget-conditioned policy trained on the downstream objective that jointly selects all three under a prescribed channel budget. Without normalization, the channel-score contribution grows with the number of selected channels and dominates the timestep and block updates. We normalize it by subset cardinality, keeping the three policy heads on comparable scales and stabilizing joint training. At matched budgets, Policy-Mask improves over fixed extraction and feature-filter baselines across segmentation, classification, and semantic correspondence, reaching 38.9 mIoU on ADE20K, 75.4% top-1 on CIFAR-100 at 128 dimensions, and 76.3 PCK@0.1 on SPair-71k. Component-wise ablations show that block selection materially affects downstream accuracy, and the highest recorded channel-budget score occurs at 256 retained channels, identifying a practical operating point for compact diffusion representations. We further adapt Policy-Mask to diffusion transformers, including PixArt-$\Sigma$ and SD3.5 MMDiT, by replacing U-Net channels with image-token hidden dimensions.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28889
PDF
-
Real2Sem2Real: 3D Semantic Augmentations for Robot Learning from a Single Human RGB Demonstration
2026-10-20 19:59:24Matthew Strong, Ankush K. Dhawan, Julien Poffet, Giuse Pham, Jinho So, Monroe Kennedy IIIAbstract
Human video scales easily and does not need robot hardware, but existing methods require depth sensing, overlay the robot onto 2D video, or generate data only for the demonstrated objects. We present Real2Sem2Real (R2Sem2R), a real-to-semantic-to-real pipeline that turns one monocular RGB video of a human demonstration into thousands of robot demonstrations across the semantic object instances of a category. R2Sem2R tracks the SE(3) pose of each object with DINOv3-distilled Gaussian splats, outperforming existing methods and FoundationPose with estimated metric depth. R2Sem2R then aligns digital cousins, each generated from one or a few photographs, to the demonstrated objects with a semantic alignment. This alignment matches whole objects and the functional regions of target objects, such as a coffee machine's tray. R2Sem2R maps the human hand to a feasible robot grasp that extends to the cousins. Robot demonstrations are generated either in IsaacSim, for photorealistic rendering, or MuJoCo, for contact physics, and R2Sem2R can generate over 2,000 episodes in about two hours. On two real robots and four tasks, we train a vision-language-action policy on data from one human video per task, and policies trained on our data match or exceed policies trained on teleoperation across diverse object instances. The project page is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28880
PDF
-
DiaLong: Benchmarking Reminiscence in Long-Context Dialogues
2026-10-20 19:55:41Owen Bianchi, Samer Aslan, Ryunosuke SaitoAbstract
How well can transformer-based Large Language Models (LLMs) retain and recall memories in long contexts of dialogue? LLMs are known to forget long-range factual information in their context, yet the field lacks a comprehensive benchmark to quantitatively assess these shortcomings. We present DiaLong, a long-context dialogue dataset and accompanying memory task that standardizes evaluation in this space. It consists of 100 unique conversations created by humans and then augmented with LLMs, extending beyond lengths found in existing datasets. Our selected memory task is fact recall: classify 40 candidate facts about a dialogue as true or false. This task directly probes the model's ability to actively retain and retrieve information, serving as a critical measure of memory within extended conversational contexts. In testing both the latest LLMs and proposed memory enhancements on DiaLong, we demonstrate that more work needs to be done to enable LLMs with better memory performance. By enhancing memory capabilities, LLMs can improve in maintaining coherence, relevance, and human-like interaction over extended conversation. Improving these capabilities will help LLMs fulfill diverse user needs and expectations in various conversational scenarios.
Abstract (translated)
URL
https://arxiv.org/abs/2610.25463
PDF
-
To Manipulate or Not: Opinion Dynamics in LLM Interactions
2026-10-20 19:48:20Mohammad Siahkamari, Phuoc Dat Nguyen, Stephen MacNeilAbstract
Large language models increasingly take part in extended conversations where they can both shape and be shaped by another party's views, yet most persuasion research measures only one side of that exchange. We introduce a controlled LLM-LLM debate framework in which two language-model roles repeatedly discuss a contested topic (climate-change attribution): one is given an assigned stance and, in some conditions, an explicit instruction to persuade, while the other receives no assigned position. Across four ordered model combinations involving GPT-6 Luna and Gemini 2.5 Flash Lite, we track both each role's explicitly reported position and a separate measure of its turn-by-turn behavior inside the conversation. Opinion dynamics differ substantially across model combinations: same-model interactions narrow the opinion gap on average, clearly for only one of the two models, while cross-model interactions widen it. Explicit reports and in-dialogue behavior capture distinct aspects of opinion dynamics, and a blanket instruction to persuade does not reliably increase influence. These findings suggest that conversational influence depends not only on assigned roles and initial positions but also on the participating models, with implications for how opinion change is measured in extended LLM interactions.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28875
PDF
-
Small Average Error, Wide Individual Error: General-Purpose LLMs Estimating Energy Intake from Meal Photos in a Feeding Trial
Abstract
Photo-based estimation of food energy could ease dietary assessment, and general-purpose multimodal large language models (LLMs) make such estimates easy to obtain. Most evaluations use one photo taken before eating and compare with energy served, while dietary assessment needs energy eaten. We evaluated eight models from three vendors (Claude Haiku, Sonnet and Opus 5.5; GPT-6 luna, 6.1 sol and 6 astra; Gemini 3.8 Flash and 4 Argon) on 802 meals from ACETADA, a controlled feeding trial with weighed intake, with the before photo only, before and after photos in separate calls (subtraction), or both photos in one request. Three results ran against intuition. Subtraction under-estimated intake for six of eight models, because the models still report hundreds of kilocalories for a nearly empty tray. Scaling the before-only estimate by the average share eaten (0.87) matched much of the gain from the after photo; beyond that constant, the second photo lowered meal-level error by only 2.2 to 4.4 points, mainly in meals with food left. With one photo, two models had a near-zero intake bias only because they under-estimated the served energy and ignored uneaten food. In ablations, adding the after photo improved accuracy under both prompt formats for seven models, while prompt effects varied across models. Over 148 feeding days, five models kept the 90% confidence interval of their mean error within a pre-set 10% margin; one model also met the pre-set 30% limits of agreement, narrowly (-28% to +27%; limits outside 30% cannot be excluded), and the other seven did not. Protein was under-estimated by every model, most in protein-dense meals; repeated calls changed error by at most 1.4 points; a high reasoning effort helped only the smallest model. A low average bias is not evidence of accurate individual intake estimation; such estimates should support, not replace, assessment by a dietitian.
Abstract (translated)
URL
https://arxiv.org/abs/2610.26538
PDF
-
A General Topological Symmetry Index for Critical Orbits In Neural Network Loss Landscapes
Abstract
A systematic understanding of the global structure of neural-network loss landscapes remains challenging due to their high dimensionality. Symmetries arising from network architectures and learning objectives provide a way to investigate aspects of this structure. Although a growing number of studies have drawn attention to these symmetries, a systematic characterization of them, as well as of the critical orbits they give rise to, remains incomplete. To address this gap, in this work we establish a topological symmetry index, based on the methodologies from the family of equivariant degree theories, to decode these symmetries. The frameworks apply to arbitrary network architectures and include methods applicable to degenerate critical orbits. The index detects the isotropy of the critical points and quantifies the corresponding net topological contributions. Several examples, including softplus MLPs, Transformers, and equivariant neural networks, are presented to illustrate the computations.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28864
PDF
-
Depth-guided Geometric Adaptation for Few-shot Image Classification
2026-10-20 19:29:42Zhiqi Miao, Estefan\'ia Talavera, Weichuan Zhang, Jiapan GuoAbstract
Few-shot image classification (FSIC) frequently suffers from visual ambiguity, where background clutter and RGB appearance similarities hinder discriminative feature matching in data-scarce settings. To mitigate this, we propose DAFIC, a geometry-guided framework that incorporates monocular depth into representation learning, episodic adaptation, and dense support-query matching. DAFIC first learns geometry-aware token representations through RGB-D self-supervised pre-training. Subsequently, the Depth-guided Geometric Adaptation (DGA) module uses depth-derived geometry to refine patch representations with geometry-enriched class context while suppressing background interference. Spatial Bidirectional Patch Matching (SBiM) then performs bidirectional dense matching on the refined features, reducing sensitivity to spurious correspondences and background clutter. Extensive experiments on four benchmarks show that DAFIC achieves strong performance against existing methods, with ablation and robustness analyses supporting the effectiveness of coupling geometry-conditioned episodic adaptation with bidirectional matching for few-shot visual recognition.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28856
PDF
-
Speedrunning Relational Foundation Model Pretraining
2026-10-20 19:29:31Gabriel Wedson Mendon\c{c}a de Souza Veloso, Luan Borges Teodoro Reis Sena, Denis Oliveira Correa, Francisco Galuppo AzevedoAbstract
Progress on foundation model pretraining is gated by who can afford a trial, and relational pretraining is no exception. Recipes are developed inside individual labs, reported under private protocols, and do not accumulate. Speedrun leaderboards remove that barrier by decentralizing the search rather than the training run. The problem is fixed, the stack is open, entries are ranked by wall-clock time to a quality target, and one GPU is enough to hold the record. We introduce modded-nanoRT, the first such leaderboard for Relational Transformers. Which database to hold out decides whether the benchmark can resolve improvements at all, so we fix it by calibration, ranking six candidates by how sharply a 45-minute learning curve separates improvement from seed noise. We open the leaderboard with typed trilinear attention, a relational operator that conditions attention scores and values on the foreign key traversed, together with Muon and SIGReg. The recipe reaches the target in 305.63 seconds on one NVIDIA H200, 8.11x faster than the reference implementation, and wins 13 of 15 pairs outside the development database. We will release the benchmark, its frozen artifacts, and the record runs. Our code is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28855
PDF
-
Beyond Correct Answers: How Compositional Cues Shape Visual Judgments
2026-10-20 19:16:00Xiaoyu Ma, Zheyuan Zhang, Kaiwen Shi, Yanfang YeAbstract
Visual question answering requires models to assign attributes and relations to the correct objects, a problem known as visual binding. Structured prompts make these assignments explicit through labeled fields. Yet an improvement from the full prompt does not reveal which labels help or how they interact. First, we isolate field-label effects across ten models and six tasks by holding the image and claim fixed. These comparisons reveal differences that overall accuracy can hide. With a relation label present, adding object-role labels helps Qwen3 but hurts Qwen2 on Spatial. Second, we compare auxiliary hints that preserve or swap the question's role assignments against a no-hint baseline. Across two image sets, Qwen3's advantage with preserving hints mainly reflects errors caused by swapped hints. Preserving hints remain near baseline. Field order further affects the harm. Third, we test clearer hint wording and direct selection of the relation shown in the image. Clearer hints improve text matching but do not consistently improve visual judgments. In contrast, relation selection partly reduces errors for some hints in Qwen3 and GPT-4.1-mini.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28843
PDF
-
Decision language models as affordable and rapid alternatives to LLM judges of clinical generative AI
2026-10-20 18:46:47Rotem Gershon, Seffi Cohen, Dean Ariel, Havatzelet Bilavsky-Yarden, Amiel Sberro, Shai Yitzhaki, Osnat Tausky, Naama Golan, Maayan Rabinovitch, Ariel Avraham Hasidim, Itay Arianu, Maya Dagan, Ran Balicer, Noa DaganAbstract
Automated judges of clinical artificial intelligence (AI) outputs are necessary, but large language model (LLM) judges are costly and slow at scale. We compared Jev, a decision language model that returns structured judgments, with frontier and open-weight LLM judges in four use cases. On 118 pediatric answers, Jev's agreement with physicians' ordinal quality ratings (Gwet's AC2 0.50; expected score 0.53) matched or exceeded that of seven of 12 LLM judges, including frontier models, Claude Opus 5.5 (0.49) and GPT-6 (0.38). On physician-graded binary HealthBench criteria, Jev matched or exceeded four of six LLM judges. On MedHELM's open-ended tasks, Jev's agreement with a three-LLM jury approached or exceeded the jurors' internal agreement. On HealthBench Professional, Jev matched the benchmark's reference judge, GPT-5.4, on 91% of criteria (other judges, 91-93%). Frontier judges cost 73-294 times more than Jev and took 7-57 times longer. Jev may enable efficient clinical AI evaluation at scale.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28826
PDF
-
Register Anything Model for Generalizable and Robust Point Cloud Registration
2026-10-20 18:40:50Zheng Qin, Junhua Xi, Yan Huang, Jiahong Lai, Anwen Huang, Qiong Li, Guangda Zhang, Kai XuAbstract
We present Register Anything Model (RAM), a unified model for generalizable 3D point cloud registration that robustly estimates 6-DoF transformations across diverse domains spanning sensors and environments. Existing registration methods are typically tailored to a single domain, with carefully tuned architectures and hyperparameters, and thus degrade markedly when applied to novel domains. We bridge this gap by training a single model on diverse data from indoor, outdoor, and object domains. A central challenge is the severe geometric discrepancy across domains, i.e., differences in scene scale, sensor noise, point density, and structural patterns, which hinders processing with a unified point-based model. To address this, we introduce a geometry-aware normalization strategy that jointly voxelizes and normalizes point clouds into a standardized geometric space, adaptively based on their geometric properties to preserve salient local structures while standardizing resolution. Built on this normalization, RAM follows a hierarchical coarse-to-fine paradigm, and employs an efficient geometric transformer that models local geometric consistency instead of globally over all points. This design mitigates discretization noise introduced by voxelization, and improves computational efficiency. Extensive experiments on 12 benchmarks demonstrate state-of-the-art performance in terms of inlier ratio and registration recall. Notably, our method outperforms the counterparts trained on each separate domain, showing superior generalization capability.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28821
PDF
-
Combining Long-Term Memory and LLMs Comes at a High Cost: From Prompt Injection to Persistent Compromise
2026-10-20 18:33:05Pedro H. Barcha Correia, Matheus Azevedo de S\'a, Paulo H. Gomes de Senna, Giovanni Mandel Martignago, Victor Takashi Hayashi, Charles Christian Miers, Marcos A. Simplicio JrAbstract
The widespread adoption of generative artificial intelligence in recent years has brought a parallel rise in prompt injection and jailbreak attacks. By embedding malicious instructions, these exploits manipulate language models to bypass intended behavior and operational boundaries. In traditional chatbots, such attacks are confined to a single session because state memory is not preserved across conversations. However, persistent memory features in modern conversational interfaces enable cross-session attack propagation, establishing a threat vector known as memory injection. This work explores this emerging vulnerability in leading LLM-as-a-service platforms, highlighting its real-world severity and the need for countermeasures. While existing literature focuses on violations tailored to specific platform features, we present a general threat model. We demonstrate that all NIST attack goals are collectively achievable across ChatGPT Instant 5.3, Claude Sonnet 4.6, and Gemini 3 Flash under this model. We also show that memory injections can exploit multiple attack vectors, including personal memories, projects, skills, organization instructions, and Model Context Protocol (MCP)-connected resources. Our findings reveal that all evaluated services can be manipulated to persistently spread disinformation, generate harmful content, or simulate offline unresponsiveness, even in fresh sessions. In the particular case of Claude, shared-memory features allow an attacker to compromise multiple users within an organization and also exfiltrate user prompts. To address these vulnerabilities, we derive defense mechanisms from our proposed threat model and foundational cybersecurity principles. Our analysis indicates that nearly half of these mitigations remain unimplemented across the target services. Datasets and attack artifacts are available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28810
PDF
