Foundation models such as GPT and Claude have shown strong capabilities in spatial understanding and embodied intelligence. These advances make agentic robotics a promising path toward artificial general intelligence in robotics. However, we find that using foundation models to control robots poses serious safety risks, and comprehensive safety evaluation is still lacking. We introduce RoboRisk, a comprehensive benchmark for evaluating safety in agentic robotics. Its 92 simulated tasks and 13 physical-robot tasks fall into five categories: Task-level Refusal (T), Plan Repair (P), Execution-time Hazard (E), Injection (I), and Sequential Memory (S). Prior benchmarks cover at most two of these categories, and several evaluate only text outputs. In contrast, RoboRisk covers all five and executes every task in closed loop. We evaluate six state-of-the-art foundation models on RoboRisk: GPT-6 Astra, GPT-6.1 Sol, GPT-6 Luna, Claude Opus 5.5, Gemini 4 Argon, and Grok 4.7. These models generally show high harm rates. We also evaluate two robot-specific APIs: the Dream API and PokeBot's RoboAPI. We further propose the Lease-Enforced Agentic Safety Harness (LEASH), which offers essential safety infrastructure for agentic robotics and can be integrated into nearly any agentic architecture. LEASH leverages a decision model (Jev or OpenAI's Decisions API) to implement task, action, and commit gates, which review the task, each action, and each irreversible operation in their visual context. The planner and all three gates share a common safety policy. Experiments on RoboRisk in simulation and on physical robots show that LEASH effectively reduces harm across all six foundation models. It largely preserves their task completion performance and adds little time and token overhead.
https://arxiv.org/abs/2611.14115
We present our system for the SHROOM-visions 2026 shared task on character-level hallucination span detection in image-conditioned vision-language model (VLM) responses, evaluated across English, French, Italian, and Chinese. The detector is built on a frozen Gemma-4-12B-it backbone with trainable LoRA adapters (approximately 10.7M parameters). Because the backbone is a causal decoder, hidden states read from a single copy of an answer cannot attend to subsequent tokens; we therefore repeat the candidate answer within one prompt and extract features from its second occurrence, giving every position approximate bidirectional context in a single non-autoregressive pass. A character-level locator combines intermediate hidden states to predict hallucination probabilities, BIO boundaries, and a response-level abstention score, while separate language-specific classifiers assign one of five hallucination types to each predicted span. On the closed test set, our system reaches a macro-averaged character IoU of 0.4404, more than doubling the official baseline (0.1854), and ranks 5th-6th on IoU across the four languages. On the English tuning split, controlled ablations attribute a gain to the repeated answer that is comparable to that of LoRA adaptation (its held-out gain is smaller and concentrated on hallucinated answers), and they show that the detector relies on the image-answer pairing: shuffling images across examples drives IoU down to the level of the predict-nothing floor or below it in every language. In deployments where the same Gemma-4-12B-it model also generates answers, its base weights can be reused, making the approach practical for local question-answering systems that cannot run a separate verifier.
https://arxiv.org/abs/2611.14100
We study how to generate diverse, physically executable robot demonstrations for dexterous vision-language-action (VLA) learning. Existing demonstration-generation methods typically rely on costly teleoperated examples. Meanwhile, human hand--object datasets provide a reusable source, but existing transfer methods mainly reproduce given interactions. Turning these interactions into VLA training data further requires coverage of different objects and manipulation locations, together with executable arm and finger actions, and visual observations. To achieve this, we present DexSeed, a framework for low-cost generation of dexterous robot data from human hand--object interactions. Our key idea is to expand human sequences while preserving their grasps and realizing the resulting motions under arm-tracking dynamics in physics simulation. Specifically, we organize source sequences by grasp mode with shared language descriptions. InteractionMesh-based retargeting and arm trajectory augmentation adapt these sequences to the robot hand, different object geometries, and pick-and-place locations while preserving local hand--object relationships. An RL finger-control policy produces executable rollouts, while the arm tracks precomputed trajectories. Rendered observations and recorded actions from successful rollouts are paired with grasp descriptions to fine-tune a VLA. Experiments show improved contact fidelity in retargeting and more reliable whole-arm execution than the evaluated baselines. Simulation evaluation and real-world deployment further demonstrate that a VLA fine-tuned solely on the generated demonstrations can execute language-specified grasps. Project page: this https URL.
https://arxiv.org/abs/2611.14092
We present Sable, an argument system and single-GPU prover for the statement that a committed, integer-quantized transformer produced a given token sequence from a given prompt, bit for bit. Sable fixes an integer function F_O with no floating-point step, in which every rounding, normalization, softmax and sampling decision is determined by its inputs, and proves it with a constraint system whose relations provably never wrap modulo the 31-bit KoalaBear prime. The proof batches zerochecks and LogUp-GKR lookups into one sumcheck per domain, and commits to every witness tensor and weight in one multilinear commitment built from Reed-Solomon codes, BLAKE3 Merkle trees and a mixed-length BaseFold opening. The prover streams the model layer by layer on one GPU. On one NVIDIA RTX PRO 6000 Blackwell GPU, Sable proves all 64 layers of Qwen2.5-32B in 92.5 s for 4,096 tokens and 398 s for 16,384 tokens, and covers Qwen3-32B and the mixture-of-experts Qwen3-30B-A3B (4,096 tokens in 121.8 and 84.7 s). A proof costs 41 to 50 times the serving time, or 6.4 cents per 4,096-token request at public GPU prices, and 40 to 64 times less time per parameter and token than zkLLM. On WikiText-2, F_O's greedy token matches the BF16 model's at 86.3% of positions; on Qwen3, F_O is within 1.6 points of BF16 on MMLU and 1.7 points on GSM8K. A verifier with a compiled core checks a proof on one CPU core in 19 to 56 s. In the random-oracle model, the interactive proof has round-by-round soundness error 2^-103. The proof is not zero-knowledge, so the system targets open-weight models. Code: this https URL
https://arxiv.org/abs/2611.14083
Fine-grained visual perception requires Vision-Language Models (VLMs) to locate and interpret subtle evidence within complex scenes. Constructing supervision for this ability is challenging: questions must have correct, unambiguous answers that depend on actual fine-grained visual details rather than coarse cues or language priors. Verifying these properties makes manual annotation labor-intensive and automatic generation difficult for models that themselves struggle with fine-grained perception. Yet the same models can often recognize such details once the region is zoomed in. We exploit this global-to-zoom perception gap and introduce ZoomGain, a data construction pipeline that constructs questions about the full image from visual information gained only by zooming in. It verifies each answer on the zoomed view with multiple models, ensures that each question remains unambiguous in the full image and cannot be answered from language priors, and specifically generates high-quality distractors for multiple-choice questions. With this pipeline and human-only screening (no manual QA editing), we build ZoomGainBench, a 1,000-question bilingual benchmark on which the best of 28 VLMs achieves only 61.08. Failure analysis reveals that models favor a particular type of distractor when answering multiple-choice questions incorrectly. Beyond evaluation, training on just 9K ZoomGain samples substantially improves the visual perception capabilities of our 4B/9B models, enabling them to outperform several hundred-billion-parameter models across various perception benchmarks. The trained 9B model achieves SoTA performance on several benchmarks, such as 90.25 on HR-Bench 4K. These results establish global-to-zoom perception gaps as a practical basis for scaling fine-grained visual supervision for both evaluation and training. Our code is available at this https URL.
https://arxiv.org/abs/2611.14076
In this paper, we propose a new paradigm in wildfire management for detecting both visible and obscured fire using Deep Learning (DL). The essence of our method is the joint spatial and temporal analysis of RGB videos, embracing the fact that temporal patterns of smoke motions in consecutive video frames can indicate obscured fire flames. This perspective contrasts the traditional approach of independently analyzing images and individual video frames. To this end, we propose a frame-wise transformer architecture that applies an attention mechanism between the consecutive frames while preserving spatial information within frames as opposed to regular attention mechanisms like the non-local block, which combines spatial and temporal information into a single vector. We also optimize the temporal pattern lengths under different spatial resolutions to achieve the highest performance while reducing the computational cost. To investigate the applicability of our method in aerial image processing, we applied it to a curated version of the FLAME2 dataset that includes dual-mode RGB/IR videos. The IR video feed is used to extract ground truth, while the model is trained only on RGB videos to maintain applicability to lowcost commercial drones (which are equipped only with regular cameras) and facilitate processing existing RGB-only datasets. Our proposed method achieves a Segmentation Foreground Dice score of 92.61%, an Object Detection Precision and Recall rate of 93.21% and 91.73%, and clip-level classification accuracy of 98.02%, respectively. We also investigated the efficacy and generalizability of our method through ablation study
https://arxiv.org/abs/2611.14074
A vision-language model may take left-right direction from the orientation of image content within patches, its vision transformer's (ViT) position code, or token positions and order in its language model (LLM). Ablations show the ViT code is necessary; we test sufficiency by mirroring only a position code. In Qwen2.5-VL-7B, Qwen3-VL-4B and InternVL3-2B, mirroring ViT width positions reproduces the flipped-image answer on 84.3-93.2% of CV-Bench left-right questions (Qwen2.5-VL-3B: 87.5%, partly a left fallback). With two options, this also counts disruption; a half-grid shift flips most answers too. With the ViT code intact, LLM-side id and token-order mirrors change few left-right answers in all three series (a weak id test in Qwen2.5-VL). In Qwen2.5-VL-7B the effect sits in the global-attention ViT blocks, mostly the last. On What'sUp, flipping content and token order, not the position code, keeps more on/under than left-right answers at the flipped-image answer. Post hoc analyses qualify these results. On CV-Bench items with located objects, random permutations flip 18.9-31.0 points more answers when reversing the objects' order than when keeping it. On What'sUp, in Qwen2.5-VL and Qwen3-VL, the share of the left-right changes made by random permutations that land on the mirror relation is mostly below chance. Counted as returns to the original answer, the What'sUp contrast of support with left-right relations is clear only in InternVL3-2B. Most remaining errors give the same relation whichever object is asked about, suggesting that repair may target relation binding, object localization or use of the reference object.
https://arxiv.org/abs/2611.14060
Emergencies can occur where there is no phone signal and no ambulance nearby. A small language model on a phone can be helpful in such situations, but its advice needs to be safe and accurate. We built HandsOnAid, a 4-bit Gemma 2B model fine-tuned with QLoRA on FirstAidQA dataset, a corpus of 5,550 first-aid question-answer pairs. We compared seven configurations, including the base, 4-bit and 8-bit fine-tuned models, and versions with a length floor, safety gate, or retrieval. We used a 41-pair question-answer bank with medically reviewed references that followed ANZCOR guidelines, and a strict 0-5 score rubric, with twelve safety overrides. Four frontier LLM judges (GPT-5.6 Sol, Claude Opus 5, DeepSeek V4 Pro, GLM-5.3) blindly scored every answer, with 45 hidden controls used to check judge reliability. Fine-tuning improved the mean score by +1.08 over the base model to 2.34. The best configuration scored 2.51 out of 5 which was the fine-tuned 4-bit model with retrieval. A medical professional also conducted a blind evaluation of 40 answers across the base, the 4-bit fine-tuned, and the retrieval configurations, and the scores were similar to the LLM judges' average scores. A keyword audit found limited coverage of several details of ANZCOR guidelines in the training corpus, that are required by the rubric, which suggests that missing information in the training data could be a cause of the low scores. On a smartphone device (Pixel 7 Pro) with six CPU threads and no network connection, the 4-bit fine-tuned model produced its first displayable text in 6.86 seconds and a complete answer in 14.88 seconds on average over ten questions.
https://arxiv.org/abs/2611.14059
World Action Models (WAMs) have recently emerged as a promising paradigm for robotic manipulation, endowing policies with strong capabilities for modeling temporal and physical dynamics. Existing WAMs, encompassing both video-first action prediction and joint video-action modeling, directly transfer global physical representations from world models to action generation. However, global scene representations may underemphasize local robot-object interactions, creating a mismatch between representations optimized for world evolution and those required for manipulation. To bridge this gap, we propose LaST0.5, a novel WAM that introduces an interaction-centric physical representation between world modeling and action generation. Specifically, LaST0.5 learns a compact interaction-centric latent space under the supervision of visual interaction cues between the robot and the environment. This reasoning space is instantiated as a latent expert that bridges the world expert and the action expert within a Mixture-of-Transformers architecture. Furthermore, LaST0.5 performs adaptive world-to-action transfer by jointly predicting future representations and progress values, using progress consistency to adjust action execution and trigger future resampling. Finally, we pretrain LaST0.5 on a large-scale robotic data corpus comprising over 14k hours of data, establishing interaction-centric physical reasoning as a scalable and effective paradigm. LaST0.5 achieves 99.5% on LIBERO, surpasses the previous state of the art on ten evaluated RLBench tasks, and demonstrates robust long-horizon execution and precise dexterous manipulation in real-world experiments.
https://arxiv.org/abs/2611.14048
Misogyny is among the most prevalent forms of online hate speech directed at Brazilian users, yet open and auditable resources dedicated to it in Brazilian Portuguese remain scarce. We introduce MISO-BR, an open, documented and versioned corpus of 896 balanced social media texts annotated for binary misogyny, released with its construction protocol and reproducible baselines. The corpus combines lexicon- and target-guided scraping from X with items imported from OLID-BR, a large language model triage stage whose prompt and decoding parameters are released, and manual annotation by three annotators following a written guide. Inter-annotator reliability reaches a Fleiss' kappa of 0.788 over 816 items. Under a single fixed protocol we compare four classical classifiers (Multinomial and Bernoulli Naive Bayes, a linear SVM and a random forest) with fine-tuned BERTimbau Base and Large. The classical models cluster between 0.651 and 0.676 macro-F1 and are precision-oriented but recall-poor. The transformers reach 0.833 macro-F1 and accuracy (Wilson 95% CI for accuracy [0.77, 0.88]), a gap of roughly 16 points, while Base and Large are indistinguishable on the test set despite a 2.6x difference in training cost. We argue that the value of the resource lies less in its headline scores than in the auditability of its pipeline, and we state three limits that users must reckon with: the artificial class balance, whose effect on precision under realistic prevalence we quantify; the association between provenance and label, which the released source field makes measurable; and the selection bias induced by lexicon-guided sampling and automatic triage. The corpus, annotation guide, models and experimental scripts are publicly available.
https://arxiv.org/abs/2611.14036
Users of model APIs and third-party platforms often rely on the declared model name to identify the model responding to their requests. However, the actual model may differ from the declared one, motivating identity verification based on observable outputs. Previous studies have shown that preferences in simple numerical-choice tasks can serve as model fingerprints. We investigate whether a single fixed-length integer sequence can combine multiple numerical observations to identify its source model. For this task, we introduce NumPrint, a benchmark of 5,036 answers from 51 models, collected using Chinese, Japanese, and English prompts that request 294 integers between 1 and 355 in one response. We propose the Shallow MLP and Causal Transformer (SMCT) to capture aggregate numerical preferences and behavioral patterns in generated sequences. In 51-class closed-set evaluation, SMCT achieves 92.72% accuracy from a single sequence, outperforming all methods compared in this study. Using two and four answers increases accuracy to 96.06% and 97.12%, respectively. These results show that a single integer sequence contains useful model fingerprints and that additional answers further improve identification.
https://arxiv.org/abs/2611.14010
Language-model agents increasingly tackle long-horizon tasks whose interaction histories exceed the model's active context. They must therefore decide what to write down and what to read back, and when to do either. Recent work has begun to use on-policy distillation (OPD) to make memory control part of the policy, relying on memory tools and compaction protocols that lie outside either model's pre-training. However, this weakens the distillation on both sides: the teacher cannot give useful guidance on an interface it has never used, and the student samples too few meaningful memory actions for that guidance to take effect. To study this problem, we introduce Coding Agent Memory Gym II (CAMG-II), a suite of long-horizon agentic-RL environments spanning Shop, Coding, AutoResearch, General, and Visual. Alongside each environment's native task interface, CAMG-II provides ordinary Bash access and an episode-persistent workspace, enabling the policy to create, revise, search, and reuse files as memory across context replacements. We also introduce Coding Agent Memory Post-Training II (CAMP-II), which trains a single policy jointly across all five environments with fully asynchronous OPD to learn this file-based memory behavior directly from a frozen teacher with a prior for file operations. On CAMG-II, CAMP-II obtains 2.97 times the gain of the strongest OPD baseline. The memory behavior CAMP-II learns generalizes to two external benchmarks, Terminal-Bench 2.1 and SWE-Bench Pro V2, with improved success rates.
https://arxiv.org/abs/2611.14009
Large Vision-Language Models (LVLMs) have demonstrated remarkable capabilities but raise significant privacy concerns due to their abilities to infer sensitive personal information from images with high precision. While current LVLMs are relatively well aligned to protect universal privacy, e.g., credit card data, we argue that privacy is inherently personalized and context-dependent. This work pivots towards a novel task: can LVLMs achieve Inference-Time Personalized Privacy Protection (ITP$^3$), allowing users to dynamically specify privacy boundaries through language specifications? To this end, we present SPY-Bench, the first systematic assessment of ITP$^3$ ability, which comprises (1) 32,700 unique samples with image-question pairs and personalized privacy instructions across 67 categories and 24 real-world scenarios, and (2) novel metrics grounded in user specifications and context awareness. Benchmarking the ITP$^3$ ability of 21 SOTA LVLMs, we reveal that: (i) most models, even the top-performing o4-mini, perform poorly, with only $\sim$24% compliance accuracy; (ii) they show quite limited contextual privacy understanding capability. Therefore, we implemented initial ITP$^3$ alignment methods, including a novel Noise Contrastive Alignment variant which achieves 96.88% accuracy while maintaining reasonable general performance. These results mark an initial step towards the ethical deployment of more controllable LVLMs. Code and data are available at this https URL
https://arxiv.org/abs/2611.13997
Flow-matching models can produce high-quality image restorations, but their learned dynamics may amplify small input perturbations and make reconstructions unstable. We introduce LipsFlower to reduce this amplification. We derive stability bounds for Flow-from-$Y$, which initializes the ordinary differential equation from a noise-perturbed observation, as well as for observation-conditioned posterior sampling, which conditions the velocity field on the observation throughout integration. Based on this analysis, we develop LipHDiT, a Lipschitz-aware hourglass diffusion transformer that serves as the vector-field architecture of LipsFlower. LipHDiT incorporates stabilized approximate spectral normalization, bounded modulation, and convex residual mixing. We also use adversarial finite-radius gain regularization at sampled interpolation states to penalize excessive sensitivity to the state and, for conditional sampling, the observation. We test both formulations on blind face restoration, measuring adversarial robustness and the perceptual quality of the reconstructions. LipsFlower reduces perturbation amplification while keeping restoration quality competitive with the evaluated baselines. These results suggest that Lipschitz-aware flow matching can improve the stability of generative image restoration.
https://arxiv.org/abs/2611.13995
Which linguistic units minimise the bitrate required to encode speech content: phones, syllables, words, maybe larger units, or something in between? This question has implications for the development of speech and language systems and theories of human language processing. Using a transcribed corpus of 960 hours of English audiobooks, I compare the information rates of phones, syllables, words, multi-word combinations and intermediate tokenisations. While all units are equivalent under a perfect language model, it makes sense to compare these under limited modelling budgets; I use $n$-gram and transformer language models. The lowest bitrates are achieved using words, with slightly longer units sometimes performing best. The results show that, when modelling capacity is limited, units larger than syllables but smaller than long multi-word phrases provide the most efficient representation in terms of information rate.
https://arxiv.org/abs/2611.13990
In Mixture-of-Experts (MoE) models, the nature of expert specialization remains controversial. Reports of domain-associated expert preferences contrast with analyses attributing routing primarily to token identity or syntactic patterns. To clarify these differing accounts, we look beyond expert-level routing to investigate the neuron-level organization of domain-specialized computation using Routing-Aware Magnitude Profiling (RAMP), a measure developed for this analysis. Using mixed-domain calibration data, we identify domain-specialized neuron sets with 58-74% overlap across multi-source samples of the same domain, compared with 15-24% across domains. To understand how this organization evolves during training, we trace neuron importance throughout pre-training and post-training across three fully open MoE families. Neuron importance continues to reorganize between pre-training checkpoints despite comparatively stable expert-level importance rankings, while post-training largely preserves neuron rankings. Building on these neuron-level perspectives, we further introduce RAMP-based Pruning (RAMPr) to select neurons across experts for balanced multi-domain capability preservation. At 75% neuron retention, RAMPr preserves average 91-98% of original full performance across six models. At 50% retention, our pruning method can reduce resident GPU memory by 36-43% and improves batched prefill throughput by 17-39% effectively. Our code is available at this https URL
https://arxiv.org/abs/2611.13975
Text-guided segmentation models receive an image and a short caption about the lesion, and on public benchmarks the caption improves accuracy considerably. On colonoscopy and breast ultrasound, however, the size, count and location words of each caption were computed from the mask that later scores the prediction, and the chest X-ray captions were written by benchmark annotators. We tested what two published models take from them. On colonoscopy and chest X-ray, it is mostly the lesion's extent. Keeping only the size word or the lung-zone phrase recovers 84 to 98 percent of the improvement over an empty caption, and removing it loses 52 to 77 percent. The rest matters little, and a polyp called a rock is segmented just as well. We call this extent leakage. On breast ultrasound, nearly half the captions also name the diagnosis, which carries part of the benefit. Retraining on three datasets shows that models pick up this habit from their training captions, though only partly on breast ultrasound, and that it weakens their segmentation from the image alone. The leak can be measured without training a model, since caption words alone predict lesion area (coefficient of determination 0.54 to 0.77 on six of seven test sets). Without the test mask, the best method we tested, which estimates extent from the image, reaches 74.59 Dice on colonoscopy against the published 77.23. We propose an evaluation protocol that keeps what a caption says about the image apart from what it copies from the label.
https://arxiv.org/abs/2611.13970
Natural fractures (NFs) and drilling-induced tensile fractures (DITFs) in acoustic image logs provide insights into rock mass structure and in situ stress, but manual characterization is time-consuming. Deep learning can automate this task, although expert-labeled data are scarce. We investigate whether adapting a pretrained vision foundation model (VFM) outperforms task-specific models trained from scratch with limited annotations. Using 80 annotated acoustic image samples, we fine-tuned DINOv3, pretrained on satellite imagery, and evaluated it on four boreholes excluded from fine-tuning. The adapted model estimated NF and DITF parameters comparably to human interpretation. Relative to U-Net, YOLOv8, and Swin Transformer models trained from scratch, it yielded fewer false positives and negatives and more accurate fracture counts and orientations. However, its performance advantage generally narrowed as training sample size decreased. We also assessed decoder architecture, fine-tuning strategy, and image enhancement. A progressive upsampling decoder with full fine-tuning performed best among the DINOv3 configurations tested, while fracture-focused image enhancement did not consistently improve accuracy. These results demonstrate the potential of VFM adaptation for fracture characterization in acoustic image logs with limited labeled data and suggest a transferable workflow for other annotation-scarce geoscientific imaging tasks.
https://arxiv.org/abs/2611.13911
Under a fixed frame budget, video understanding with vision-language models (VLMs) depends on whether the sampled frames provide sufficient evidence for the question. Sufficient frame evidence supports direct answering, whereas less informative observations may benefit from reasoning that connects the available cues. We introduce EVAThink, which enhances video understanding through evidence sufficiency-guided training for adaptive thinking. First, Recoverable Evidence Gap Mining constructs evidence-dense and evidence-sparse video views that share the same question and answer under an identical frame budget. It retains pairs where sparse views yield lower direct-answer accuracy than dense views but benefit from explicit reasoning, providing supervision for response-mode selection. Second, Evidence-Guided Two-Stage Reinforcement Learning uses stage-specific rewards to guide training with evidence-based mode preferences and then refine response efficiency, while first-token advantage regularization mitigates mode collapse. Trained on only 6.7K curated samples, EVAThink consistently improves upon the base model across video understanding benchmarks under both 32- and 64-frame budgets, while outperforming strong baselines on average. The code, model, and datasets are available at \href{this https URL}{EVAThink}.
https://arxiv.org/abs/2611.13907
Three-dimensional scene graphs (3DSGs) can provide robots with persistent metric-semantic representations for spatial reasoning, but the existing methods have primarily been developed for indoor and terrestrial environments. In this paper we investigate how a representative outdoor open-vocabulary method transfers to maritime environments and which adaptations are needed. Our experiments show that zero-shot representative outdoor 3DSG descriptions agree with our labeled ground truth for a majority of geometry-supported objects on our data, but the resulting map exhibits considerable physical-entity fragmentation, and its terrestrial higher-level assumptions don't capture maritime spatial organization needs. We therefore keep the underlying mapping method and post-process its accumulated evidence. We show a metric-semantic consolidation method followed by geometry-grounded multi-view vision-language model (VLM) reasoning that reconciles remaining fragments into physical entities and groups entities through observations into local places. In parallel, we construct maritime macro regions using map priors and onboard water and structure observations, and associate them with the local places. The resulting representation reduces physical-entity fragmentation beyond both the baseline and our geometry-semantic consolidation component alone, supports correspondence of fixed local places across repeated maritime traversals, and improves contextual retrieval on our reviewed queries over tested baselines. These results show how an outdoor open-vocabulary map can be adapted into persistent maritime scene memory with our methods for object/place identity and spatial abstractions.
https://arxiv.org/abs/2611.13903