Telecomunicaciones, información y comunicación

Ω-QVLA: Robust Quantization for Vision-Language-Action Models via Composite Rotation and Per-step Scaling

- Vision-Language-Action (VLA) models unify perception, reasoning, and control within a single policy, yet their multi-billion-parameter backbones and diffusion-based action heads make on-device...

OmniVerifier-M1: Multimodal Meta-Verifier with Explicit Structured Recalibration

- Visual outcomes are increasingly central to multimodal large language models, making reliable and fine-grained verification essential for scaling generalist foundation models. In this work, we...

Personal Visual Memory from Explicit and Implicit Evidence

- Long-term memory is increasingly important for personalized AI agents, yet existing benchmarks and methods remain largely text-centric. Even when images are included, the user-specific information...

AREA: Attribute Extraction and Aggregation for CLIP-Based Class-Incremental Learning

- Class-Incremental Learning (CIL) is important in building real-world learning systems. In CLIP-based CIL, the model performs classification by comparing similarity between visual and textual...

Affective Music Recommendation: A Rollout-Based World Model for Offline Preference Optimization

- Functional music applications, from consumer focus and sleep aids to clinical interventions, share a distinctive recommendation problem: success is defined by the listener's affective state, but...

Beyond Binary: Sim-to-Real Dexterous Manipulation with Physics-Grounded Contact Representation

- A primary bottleneck in contact-rich manipulation is the difficulty of collecting real-world data. Sim-to-real reinforcement learning offers a scalable alternative, but the simulation-reality gap...

Self-Improving Language Models with Bidirectional Evolutionary Search

- Search has been proposed as an effective method for self-improving language models and agentic systems, both for post-training sample generation and for inference. However, widely used methods such...

Gamma-World: Generative Multi-Agent World Modeling Beyond Two Players

- World models for interactive video generation have largely focused on single-agent settings, where future observations are generated from a single control signal. However, many generated environments...

VLMs May Not Globally Enhance Human Alignment over LLMs During Natural Reading

- Large language models (LLMs) have become increasingly useful computational models of human language processing, but it remains unclear whether vision-language learning makes text representations more...

From Pixels to Words -- Towards Native One-Vision Models at Scale

- Current vision-language models (VLMs) typically stitch together separate image encoders and language decoders via multi-stage alignment, a modular framework that inevitably fragments pixel-level...

Entropic and operational characterizations of dynamic quantum resources

- We offer new methods for characterizing general closed and convex quantum resource theories, including dynamic ones, based on entropic concepts and operational tasks. We propose a resource-theoretic...

SwarmHarness: Skill-Based Task Routing via Decentralized Incentive-Aligned AI Agent Networks

- Vast quantities of compute (GPU cycles on personal workstations, idle inference servers, and edge devices between jobs) go unused because no incentive-aligned protocol exists for their owners to...

Principled Algorithms for Optimizing Generalized Metrics in Multi-Label Learning

- Many real-world classification tasks require predicting multiple labels per instance, necessitating the optimization of complex evaluation metrics such as the $F$-measure and Jaccard index. While the...

Multi-Mixer Models: Flexible Sequence Modeling with Shared Representations

- Softmax attention is the cornerstone of modern large language models, but its memory scales linearly and compute quadratically with sequence length. Linear recurrent models, such as linear attention...

Sampling Random Graphs from the Colored Configuration Model

- A fundamental step in knowledge discovery is statistically assessing data mining results. In network analysis, such evaluation compares the outcome of a given procedure with the outcomes obtained...

Extrapolative Weight Averaging Reveals Correctness-Efficiency Frontiers in Code RL

- Linear interpolation between fine-tuned checkpoints has been shown to trace the Pareto front between competing objectives, but whether extrapolative weight averaging can extend such frontiers to new...

LLM Zeroth-Order Fine-Tuning is an Inference Workload

- Zeroth-order (ZO) fine-tuning is attractive for large language models because it replaces backpropagation with forward objective evaluations. Existing implementations nevertheless execute ZO...

Human Label Variation as Stable Signal: Learning Annotator-Specific Explanation Behavior via Cross-Annotator Preference Optimization

- Free-text explanations extend human label variation (HLV) beyond label disagreement by revealing the reasoning and preferences behind annotators' decisions. We study whether large language models...

CaMBRAIN: Real-time, Continuous EEG Inference with Causal State Space Models

- Electroencephalography (EEG) is a critical, non-invasive method to monitor electrical brain activity. EEGs can span anywhere from a couple seconds to multiple hours, posing a major hurdle for...

Skill-Conditioned Gated Self-Distillation for LLM Reasoning

- On-policy self-distillation (SD) improves LLM reasoning by using teacher-side privileged information (PI) to turn sparse verifier outcomes into dense token-level supervision. Existing methods usually...

Can Large Language Models Handle Discourse Particles? A Case Study of Colloquial Malay

- Discourse particles, such as \textit{well} and \textit{kind of}, are crucial components that enable LLMs to ``speak'' more like humans. They are used to convey emotions, intentions, and...