arXiv subjects

Victor Sanchez

Publications and source records attributed to Victor Sanchez.

At least 19 recordsLinked to original sources

arXiv2026

Revisiting Chain-of-Thought Reasoning under Limited Supervision: Semi-supervised Chain-of-Thought Learning

Chain-of-thought (CoT) reasoning has emerged as an effective approach for activating latent reasoning capabilities in large language models. However, most existing CoT methods use reasoning chains mainly as inference-time prompts, while the generated reasoning traces are rarely reused as semi-supervised learning signals. In this report, we define \textbf{Semi-supervised Chain-of-Thought Learning} and propose \textbf{Semi-CoT}, a simple framework that uses unlabeled questions to construct pseudo reasoning supervision. Semi-CoT samples multiple pseudo-CoTs for each unlabeled question, estimates answer-level semantic entropy, and selects low-entropy reasoning chains as reliable pseudo-CoT demonstrations. This extends the self-training view of CoT from inference-time refinement to semi-supervised pseudo-supervision. Pilot experiments on AQuA, SVAMP, GSM8K, and MultiArith show that the entropy gate selects high-precision pseudo-CoTs, with pseudo-answer precision ranging from $91.36\%$ to $100\%$. Semi-CoT also gives small gains on SVAMP and GSM8K, while AQuA shows negative transfer and MultiArith reaches a ceiling. These results suggest that unlabeled questions can provide reliable pseudo reasoning signals, but their effective use still requires stronger demonstration selection or student training.

cs.AI↗

arXiv2026

Semi-Supervised Vision-Language-Action Model

Vision-Language-Action (VLA) models enable robots to predict actions directly from visual observations and language instructions, but adapting them to new environments still depends on costly action-labeled demonstrations. To reduce this dependence, we study semi-supervised VLA adaptation under limited supervision signals, where only a small portion of trajectories contain robot actions and the remaining trajectories provide action-unlabeled vision-language observations. Unlike standard semi-supervised learning, the missing supervision is an embodied action signal that must be visually grounded, language-consistent, physically feasible, and temporally stable. To address this problem, we propose SemiVLA, a self-distilled teacher-student framework that learns from reliable pseudo-actions on unlabeled trajectories. SemiVLA introduces a VLA-specific reliability controller to assess vision-language alignment, action feasibility, and temporal transition consistency, and further updates the teacher through a Bottleneck-Projected Alignment Update to avoid noisy feedback contamination. With OpenVLA as the backbone, SemiVLA consistently improves multiple PEFT strategies across LIBERO and CALVIN. Under 10\% labeled trajectories, SemiVLA with Selective LoRA achieves 89.0\% average success on LIBERO, outperforming supervised LoRA by 8.0 points without extra inference cost.

cs.CV↗

arXiv2026

TAPM-Net: Trajectory-Aware Perturbation Modeling for Infrared Small Target Detection

Infrared small target detection (ISTD) remains a long-standing challenge due to weak signal contrast, limited spatial extent, and cluttered backgrounds. Despite performance improvements from convolutional neural networks (CNNs) and Vision Transformers (ViTs), current models lack a mechanism to trace how small targets trigger directional, layer-wise perturbations in the feature space, which is an essential cue for distinguishing signal from structured noise in infrared scenes. To address this limitation, we propose the Trajectory-Aware Mamba Propagation Network (TAPM-Net), which explicitly models the spatial diffusion behavior of target-induced feature disturbances. TAPM-Net is built upon two novel components: a Perturbation-guided Path Module (PGM) and a Trajectory-Aware State Block (TASB). The PGM constructs perturbation energy fields from multi-level features and extracts gradient-following feature trajectories that reflect the directionality of local responses. The resulting feature trajectories are fed into the TASB, a Mamba-based state-space unit that models dynamic propagation along each trajectory while incorporating velocity-constrained diffusion and semantically aligned feature fusion from word-level and sentence-level embeddings. Unlike existing attention-based methods, TAPM-Net enables anisotropic, context-sensitive state transitions along spatial trajectories while maintaining global coherence at low computational cost. Experiments on NUAA-SIRST and IRSTD-1K demonstrate that TAPM-Net achieves state-of-the-art performance in ISTD.

cs.CV↗

arXiv2026

OpenSocInt: A Multi-modal Training Environment for Human-Aware Social Navigation

In this paper, we introduce OpenSocInt, an open-source software package providing a simulator for multi-modal social interactions and a modular architecture to train social agents. We described the software package and showcased its interest via an experimental protocol based on the task of social navigation. Our framework allows for exploring the use of different perceptual features, their encoding and fusion, as well as the use of different agents. The software is already publicly available under GPL at https://gitlab.inria.fr/robotlearn/OpenSocInt/.

cs.AI↗

arXiv2025

Hierarchical Bayesian constitutive model selection for high-strain-rate soft material characterization

The high-fidelity characterization of soft, tissue-like materials under ultra-high-strain-rate conditions is critical in engineering and medicine. Still, it remains challenging due to limited optical access, sensitivity to initial conditions, and experimental variability. Microcavitation techniques (e.g., laser-induced microcavitation) have emerged as a viable method for determining the mechanical properties of soft materials in the ultra-high-strain-rate regime (higher than 10^3 s^{-1}); however, they are limited by measurement noise and uncertainty in parameter estimation. A hierarchical Bayesian model selection method is employed using the Inertial Microcavitation Rheometry (IMR) technique to address these limitations. With this method, the parameter space of different constitutive models is explored to determine the most credible constitutive model that describes laser-induced microcavitation bubble oscillations in soft, viscoelastic, transparent hydrogels. The target data/evidence is computed using a weighted Gaussian likelihood with a hierarchical noise scale, which enables the quantification of uncertainty in model plausibility. Physically informed priors, including range-invariant, stress-based parameter priors, a model-redundancy prior, and a Bayesian Information Criterion motivated model prior, penalize complex models to enforce Occam's razor. Using a precomputed grid of simulations, the probabilistic model selection process enables an initial guess for the Maximum A Posteriori (MAP) material parameter values. Synthetic tests recover the ground-truth models and expected parameters. Using experimental data for gelatin, fibrin, polyacrylamide, and agarose, MAP simulations of credible models reproduce the data. Moreover, a cross-institutional comparison of 10% gelatin indicates consistent constitutive model selection.

physics.flu-dyn↗

arXiv2025

ZIP: Scalable Crowd Counting via Zero-Inflated Poisson Modeling

Most crowd counting methods directly regress blockwise density maps using Mean Squared Error (MSE) losses. This practice has two key limitations: (1) it fails to account for the extreme spatial sparsity of annotations - over 95% of 8x8 blocks are empty across standard benchmarks, so supervision signals in informative regions are diluted by the predominant zeros; (2) MSE corresponds to a Gaussian error model that poorly matches discrete, non-negative count data. To address these issues, we introduce ZIP, a scalable crowd counting framework that models blockwise counts with a Zero-Inflated Poisson likelihood: a zero-inflation term learns the probability a block is structurally empty (handling excess zeros), while the Poisson component captures expected counts when people are present (respecting discreteness). We provide a generalization analysis showing a tighter risk bound for ZIP than MSE-based losses and DMCount provided that the training resolution is moderately large. To assess the scalability of ZIP, we instantiate it on backbones spanning over 100x in parameters/compute. Experiments on ShanghaiTech A & B, UCF-QNRF, and NWPU-Crowd demonstrate that ZIP consistently surpasses state-of-the-art methods across all model scales.

cs.CV↗

arXiv2025

OptiScene: LLM-driven Indoor Scene Layout Generation via Scaled Human-aligned Data Synthesis and Multi-Stage Preference Optimization

Automatic indoor layout generation has attracted increasing attention due to its potential in interior design, virtual environment construction, and embodied AI. Existing methods fall into two categories: prompt-driven approaches that leverage proprietary LLM services (e.g., GPT APIs) and learning-based methods trained on layout data upon diffusion-based models. Prompt-driven methods often suffer from spatial inconsistency and high computational costs, while learning-based methods are typically constrained by coarse relational graphs and limited datasets, restricting their generalization to diverse room categories. In this paper, we revisit LLM-based indoor layout generation and present 3D-SynthPlace, a large-scale dataset that combines synthetic layouts generated via a 'GPT synthesize, Human inspect' pipeline, upgraded from the 3D-Front dataset. 3D-SynthPlace contains nearly 17,000 scenes, covering four common room types -- bedroom, living room, kitchen, and bathroom -- enriched with diverse objects and high-level spatial annotations. We further introduce OptiScene, a strong open-source LLM optimized for indoor layout generation, fine-tuned based on our 3D-SynthPlace dataset through our two-stage training. For the warum-up stage I, we adopt supervised fine-tuning (SFT), which is taught to first generate high-level spatial descriptions then conditionally predict concrete object placements. For the reinforcing stage II, to better align the generated layouts with human design preferences, we apply multi-turn direct preference optimization (DPO), which significantly improving layout quality and generation success rates. Extensive experiments demonstrate that OptiScene outperforms traditional prompt-driven and learning-based baselines. Moreover, OptiScene shows promising potential in interactive tasks such as scene editing and robot navigation.

cs.CV↗

arXiv2025

Advancing Video Anomaly Detection: A Bi-Directional Hybrid Framework for Enhanced Single- and Multi-Task Approaches

Despite the prevailing transition from single-task to multi-task approaches in video anomaly detection, we observe that many adopt sub-optimal frameworks for individual proxy tasks. Motivated by this, we contend that optimizing single-task frameworks can advance both single- and multi-task approaches. Accordingly, we leverage middle-frame prediction as the primary proxy task, and introduce an effective hybrid framework designed to generate accurate predictions for normal frames and flawed predictions for abnormal frames. This hybrid framework is built upon a bi-directional structure that seamlessly integrates both vision transformers and ConvLSTMs. Specifically, we utilize this bi-directional structure to fully analyze the temporal dimension by predicting frames in both forward and backward directions, significantly boosting the detection stability. Given the transformer's capacity to model long-range contextual dependencies, we develop a convolutional temporal transformer that efficiently associates feature maps from all context frames to generate attention-based predictions for target frames. Furthermore, we devise a layer-interactive ConvLSTM bridge that facilitates the smooth flow of low-level features across layers and time-steps, thereby strengthening predictions with fine details. Anomalies are eventually identified by scrutinizing the discrepancies between target frames and their corresponding predictions. Several experiments conducted on public benchmarks affirm the efficacy of our hybrid framework, whether used as a standalone single-task approach or integrated as a branch in a multi-task approach. These experiments also underscore the advantages of merging vision transformers and ConvLSTMs for video anomaly detection.

cs.CV↗

arXiv2025

RefComp: A Reference-guided Unified Framework for Unpaired Point Cloud Completion

The unpaired point cloud completion task aims to complete a partial point cloud by using models trained with no ground truth. Existing unpaired point cloud completion methods are class-aware, i.e., a separate model is needed for each object class. Since they have limited generalization capabilities, these methods perform poorly in real-world scenarios when confronted with a wide range of point clouds of generic 3D objects. In this paper, we propose a novel unpaired point cloud completion framework, namely the Reference-guided Completion (RefComp) framework, which attains strong performance in both the class-aware and class-agnostic training settings. The RefComp framework transforms the unpaired completion problem into a shape translation problem, which is solved in the latent feature space of the partial point clouds. To this end, we introduce the use of partial-complete point cloud pairs, which are retrieved by using the partial point cloud to be completed as a template. These point cloud pairs are used as reference data to guide the completion process. Our RefComp framework uses a reference branch and a target branch with shared parameters for shape fusion and shape translation via a Latent Shape Fusion Module (LSFM) to enhance the structural features along the completion pipeline. Extensive experiments demonstrate that the RefComp framework achieves not only state-of-the-art performance in the class-aware training setting but also competitive results in the class-agnostic training setting on both virtual scans and real-world datasets.

cs.CV↗

arXiv2025

DiffFake: Exposing Deepfakes using Differential Anomaly Detection

Traditional deepfake detectors have dealt with the detection problem as a binary classification task. This approach can achieve satisfactory results in cases where samples of a given deepfake generation technique have been seen during training, but can easily fail with deepfakes generated by other techniques. In this paper, we propose DiffFake, a novel deepfake detector that approaches the detection problem as an anomaly detection task. Specifically, DiffFake learns natural changes that occur between two facial images of the same person by leveraging a differential anomaly detection framework. This is done by combining pairs of deep face embeddings and using them to train an anomaly detection model. We further propose to train a feature extractor on pseudo-deepfakes with global and local artifacts, to extract meaningful and generalizable features that can then be used to train the anomaly detection model. We perform extensive experiments on five different deepfake datasets and show that our method can match and sometimes even exceed the performance of state-of-the-art competitors.

cs.CV↗

arXiv2025

From Age Estimation to Age-Invariant Face Recognition: Generalized Age Feature Extraction Using Order-Enhanced Contrastive Learning

Generalized age feature extraction is crucial for age-related facial analysis tasks, such as age estimation and age-invariant face recognition (AIFR). Despite the recent successes of models in homogeneous-dataset experiments, their performance drops significantly in cross-dataset evaluations. Most of these models fail to extract generalized age features as they only attempt to map extracted features with training age labels directly without explicitly modeling the natural ordinal progression of aging. In this paper, we propose Order-Enhanced Contrastive Learning (OrdCon), a novel contrastive learning framework designed explicitly for ordinal attributes like age. Specifically, to extract generalized features, OrdCon aligns the direction vector of two features with either the natural aging direction or its reverse to model the ordinal process of aging. To further enhance generalizability, OrdCon leverages a novel soft proxy matching loss as a second contrastive objective, ensuring that features are positioned around the center of each age cluster with minimal intra-class variance and proportionally away from other clusters. By modeling the ageing process, the framework can enhance generalizability by improving the alignment of samples from the same class and reducing the divergence of direction vectors. We demonstrate that our proposed method achieves comparable results to state-of-the-art methods on various benchmark datasets in homogeneous-dataset evaluations for both age estimation and AIFR. In cross-dataset experiments, OrdCon outperforms other methods by reducing the mean absolute error by approximately 1.38 on average for the age estimation task and boosts the average accuracy for AIFR by 1.87%.

cs.CV↗

arXiv2024

Interact with me: Joint Egocentric Forecasting of Intent to Interact, Attitude and Social Actions

For efficient human-agent interaction, an agent should proactively recognize their target user and prepare for upcoming interactions. We formulate this challenging problem as the novel task of jointly forecasting a person's intent to interact with the agent, their attitude towards the agent and the action they will perform, from the agent's (egocentric) perspective. So we propose \emph{SocialEgoNet} - a graph-based spatiotemporal framework that exploits task dependencies through a hierarchical multitask learning approach. SocialEgoNet uses whole-body skeletons (keypoints from face, hands and body) extracted from only 1 second of video input for high inference speed. For evaluation, we augment an existing egocentric human-agent interaction dataset with new class labels and bounding box annotations. Extensive experiments on this augmented dataset, named JPL-Social, demonstrate \emph{real-time} inference and superior performance (average accuracy across all tasks: 83.15\%) of our model outperforming several competitive baselines. The additional annotations and code will be available upon acceptance.

cs.CV↗

arXiv2024

Enhanced Prototypical Part Network (EPPNet) For Explainable Image Classification Via Prototypes

Explainable Artificial Intelligence (xAI) has the potential to enhance the transparency and trust of AI-based systems. Although accurate predictions can be made using Deep Neural Networks (DNNs), the process used to arrive at such predictions is usually hard to explain. In terms of perceptibly human-friendly representations, such as word phrases in text or super-pixels in images, prototype-based explanations can justify a model's decision. In this work, we introduce a DNN architecture for image classification, the Enhanced Prototypical Part Network (EPPNet), which achieves strong performance while discovering relevant prototypes that can be used to explain the classification results. This is achieved by introducing a novel cluster loss that helps to discover more relevant human-understandable prototypes. We also introduce a faithfulness score to evaluate the explainability of the results based on the discovered prototypes. Our score not only accounts for the relevance of the learned prototypes but also the performance of a model. Our evaluations on the CUB-200-2011 dataset show that the EPPNet outperforms state-of-the-art xAI-based methods, in terms of both classification accuracy and explainability

cs.CV↗

arXiv2024

LLplace: The 3D Indoor Scene Layout Generation and Editing via Large Language Model

Designing 3D indoor layouts is a crucial task with significant applications in virtual reality, interior design, and automated space planning. Existing methods for 3D layout design either rely on diffusion models, which utilize spatial relationship priors, or heavily leverage the inferential capabilities of proprietary Large Language Models (LLMs), which require extensive prompt engineering and in-context exemplars via black-box trials. These methods often face limitations in generalization and dynamic scene editing. In this paper, we introduce LLplace, a novel 3D indoor scene layout designer based on lightweight fine-tuned open-source LLM Llama3. LLplace circumvents the need for spatial relationship priors and in-context exemplars, enabling efficient and credible room layout generation based solely on user inputs specifying the room type and desired objects. We curated a new dialogue dataset based on the 3D-Front dataset, expanding the original data volume and incorporating dialogue data for adding and removing objects. This dataset can enhance the LLM's spatial understanding. Furthermore, through dialogue, LLplace activates the LLM's capability to understand 3D layouts and perform dynamic scene editing, enabling the addition and removal of objects. Our approach demonstrates that LLplace can effectively generate and edit 3D indoor layouts interactively and outperform existing methods in delivering high-quality 3D design solutions. Code and dataset will be released.

cs.CV↗

arXiv2024

Beyond Isolated Frames: Enhancing Sensor-Based Human Activity Recognition through Intra- and Inter-Frame Attention

Human Activity Recognition (HAR) has become increasingly popular with ubiquitous computing, driven by the popularity of wearable sensors in fields like healthcare and sports. While Convolutional Neural Networks (ConvNets) have significantly contributed to HAR, they often adopt a frame-by-frame analysis, concentrating on individual frames and potentially overlooking the broader temporal dynamics inherent in human activities. To address this, we propose the intra- and inter-frame attention model. This model captures both the nuances within individual frames and the broader contextual relationships across multiple frames, offering a comprehensive perspective on sequential data. We further enrich the temporal understanding by proposing a novel time-sequential batch learning strategy. This learning strategy preserves the chronological sequence of time-series data within each batch, ensuring the continuity and integrity of temporal patterns in sensor-based HAR.

eess.SP↗

arXiv2024

WorkBench: a Benchmark Dataset for Agents in a Realistic Workplace Setting

We introduce WorkBench: a benchmark dataset for evaluating agents' ability to execute tasks in a workplace setting. WorkBench contains a sandbox environment with five databases, 26 tools, and 690 tasks. These tasks represent common business activities, such as sending emails and scheduling meetings. The tasks in WorkBench are challenging as they require planning, tool selection, and often multiple actions. If a task has been successfully executed, one (or more) of the database values may change. The correct outcome for each task is unique and unambiguous, which allows for robust, automated evaluation. We call this key contribution outcome-centric evaluation. We evaluate five existing ReAct agents on WorkBench, finding they successfully complete as few as 3% of tasks (Llama2-70B), and just 43% for the best-performing (GPT-4). We further find that agents' errors can result in the wrong action being taken, such as an email being sent to the wrong person. WorkBench reveals weaknesses in agents' ability to undertake common business activities, raising questions about their use in high-stakes workplace settings. WorkBench is publicly available as a free resource at https://github.com/olly-styles/WorkBench.

cs.CL↗

arXiv2024

CLIP-EBC: CLIP Can Count Accurately through Enhanced Blockwise Classification

We propose CLIP-EBC, the first fully CLIP-based model for accurate crowd density estimation. While the CLIP model has demonstrated remarkable success in addressing recognition tasks such as zero-shot image classification, its potential for counting has been largely unexplored due to the inherent challenges in transforming a regression problem, such as counting, into a recognition task. In this work, we investigate and enhance CLIP's ability to count, focusing specifically on the task of estimating crowd sizes from images. Existing classification-based crowd-counting frameworks have significant limitations, including the quantization of count values into bordering real-valued bins and the sole focus on classification errors. These practices result in label ambiguity near the shared borders and inaccurate prediction of count values. Hence, directly applying CLIP within these frameworks may yield suboptimal performance. To address these challenges, we first propose the Enhanced Blockwise Classification (EBC) framework. Unlike previous methods, EBC utilizes integer-valued bins, effectively reducing ambiguity near bin boundaries. Additionally, it incorporates a regression loss based on density maps to improve the prediction of count values. Within our backbone-agnostic EBC framework, we then introduce CLIP-EBC to fully leverage CLIP's recognition capabilities for this task. Extensive experiments demonstrate the effectiveness of EBC and the competitive performance of CLIP-EBC. Specifically, our EBC framework can improve existing classification-based methods by up to 44.5% on the UCF-QNRF dataset, and CLIP-EBC achieves state-of-the-art performance on the NWPU-Crowd test set, with an MAE of 58.2 and an RMSE of 268.5, representing improvements of 8.6% and 13.3% over the previous best method, STEERER. The code and weights are available at https://github.com/Yiming-M/CLIP-EBC.

cs.CV↗

arXiv2024

Data-Agnostic Face Image Synthesis Detection Using Bayesian CNNs

Face image synthesis detection is considerably gaining attention because of the potential negative impact on society that this type of synthetic data brings. In this paper, we propose a data-agnostic solution to detect the face image synthesis process. Specifically, our solution is based on an anomaly detection framework that requires only real data to learn the inference process. It is therefore data-agnostic in the sense that it requires no synthetic face images. The solution uses the posterior probability with respect to the reference data to determine if new samples are synthetic or not. Our evaluation results using different synthesizers show that our solution is very competitive against the state-of-the-art, which requires synthetic data for training.

cs.CV↗