Sección

Investigación

Papers y preprints de los feeds de investigación que seguimos.

6415documentos

Complementary Roles of Activation and Parametric Memory in Few-Shot Learning

At test time, large language models (LLMs) can encode historical information in activation memory (i.e., KV caches) and parametric memory (i.e., updated parameters). While activation memory is generally considered effective for factual recall and parametric memory for learning new tasks, their interplay remains unclear. In this work, we systematically investigate the role of memory in few-shot learning through controlled experiments. We find that activation memory is superior for recalling facts, whereas parametric memory does not consistently outperform activation memory in task learning. Moreover, our experiments show that the composite task, Conditional Arithmetic, requires the synergy of both memory types. Through neuron-level analysis, we find that the model activates distinct sets of neurons when accessing the same historical information through activation versus parametric memory. When both memory types are combined, the model recruits neurons from both sets, which is crucial for solving Conditional Arithmetic. These findings suggest that neither memory mechanism alone is sufficient for this composite task, highlighting the importance of their collaboration.

arXiv AI Papershace 2 meses

hyperbolix: Hyperbolic Deep Learning in JAX

We present hyperbolix, an open-source library for hyperbolic deep learning in JAX, built on Flax NNX. To our knowledge, it is the first comprehensive, general-purpose hyperbolic deep learning library in JAX. It includes six manifolds with a common interface: Euclidean space, the Poincaré ball, the hyperboloid, the κ-stereographic model, mixed-curvature product spaces, and the proper velocity space. We implement layer families that cover linear layers, convolutions, attention, normalization, positional encoding, regression, and vector quantization. These building blocks span methods ranging from Ganea's original hyperbolic neural networks to recent fully hyperbolic architectures such as Hypformer and Lorentzian ResNet. Additionally, hyperbolix contains Riemannian optimizers implemented as optax transformations, wrapped distributions, and hyperbolic dimensionality-reduction techniques. Its API uses idiomatic JAX: Manifolds are stateless, with curvature being passed at call time, while manifold operations act on single points, with jax.vmap enabling batch operations. The precision of every checked operation is tested against a closed-form NumPy/SciPy transcription from the source paper or a finite difference, for both float32 and float64. On the hyperboloid, standard formulas for two-point operations, such as the distance, lose precision far from the origin, because they subtract two large, nearly equal terms. hyperbolix replaces these subtractions with cancellation-free formulas that stay accurate in float32 at distances where prior implementations return NaN. hyperbolix is available under the MIT license at https://github.com/timoklein/hyperbolix .

arXiv AI Papershace 2 meses

Controlling Collectives of AI Agents in Reasoning Space with Spatial Transformers

Large Language Models (LLMs) introduce an exciting new paradigm for planning and navigation in robotics, but fail on even simple multi-robot tasks as team sizes grow. We propose COMPASS, a scalable, decentralized multi-robot architecture for controlling large collectives of agentic robots with reasoning space feedback control. Feedback is generated locally on each robot by a spatial transformer which aggregates multi-hop messages across the fleet into a learned feedback token. Our experiments find that collectives of language models demonstrate performance gains from structured diversity of the input command, which can cancel biases; an advantage that is held across scale. Compared against a centralized frontier LLM policy and a language-only communication ablation, we find that the coupled design of COMPASS decisively produces cohesive flocking formations that accurately fly the commanded intent. We show that reasoning feedback works best when composed with a compact learned token. Our ablations show that hand engineered feedback with raw state appearing in the language channel obliterates cohesion. COMPASS generalizes zero-shot to unseen instructions of ambiguous meaning while commanding flocks up to 16 times its training scale, flying up to 1024 robots under natural language commands.

arXiv AI Papershace 2 meses

Beyond Poetry: Can Large Language Models Generate Classical Arabic Maqamat?

Large language models (LLMs) have shown strong performance in creative text generation, yet their ability to produce culturally grounded and stylistically constrained literary forms remains underexplored. Prior work has focused largely on modern language varieties and poetry, while classical prose traditions such as maqama remain largely unstudied. The maqama is a classical literary genre characterized by rhymed prose (saj), dense rhetorical ornamentation, and episodic narrative structure, making it a challenging testbed for evaluating whether LLMs can move beyond surface fluency toward deeper literary competence. In this paper, we present the first controlled evaluation study of maqama generation with LLMs, comparing five models under zero-shot, few-shot, and rule-based prompting, and evaluating outputs through both human annotation and an LLM-as-a-judge framework across dimensions such as rhetorical richness, saj density, structural coherence, and stylistic authenticity. Our results show that prompting strategy plays a strong role in stylistic quality: few-shot prompting most consistently improves saj density, while its effects on rhetoric and coherence vary by model, with the strongest models (GPT-4o and GPT-5.4-mini) benefiting most from rule-based prompting on these dimensions, though zero-shot prompting yields the highest aggregate scores across all five models. We further observe systematic differences between models in stylistic alignment with Arabic maqama conventions, and corroborate our findings with a second independent LLM judge, paired statistical significance testing, and non-LLM proxy measures of saj.

arXiv AI Papershace 2 meses

Do Center Biases Propagate? Robustness of Pathology Foundation Models in Whole-Slide Image Classification

Pathology foundation models (PFMs) have transformed computational pathology through powerful representation learning from histopathological images. PFMs provide rich, discriminative representations for whole slide image (WSI) analysis, enabling tasks such as slide-level classification under multiple instance learning (MIL). However, these representations may also encode non-biological signals associated with acquisition centers, potentially introducing spurious shortcuts into downstream predictions. In this work, we evaluate center-associated robustness in WSI classification using a controlled training setting with increasing class-center correlations quantified by Cramér's V. We benchmark six PFMs across four datasets and two MIL aggregators, while evaluating ComBat as a robustification strategy. We further introduce the Area Under the Cramér's V Curve (AUCC) to jointly capture absolute classification performance and its degradation as spurious correlation increases. Results show that center-related information encoded by PFMs propagates to WSI-level predictions, with robustness depending on both the PFM representation and MIL aggregation strategy. Additionally, ComBat harmonization does not provide consistent robustness gains across datasets.

arXiv AI Papershace 2 meses

From Agent Output to Authorized Transition

Agentic engineering systems can edit repositories, run tools and tests, build firmware, synthesize schematics, and prepare deployable or manufacturable artifacts. The assurance problem is therefore shifting from whether an agent can produce an output to whether an engineering lifecycle is justified in acting on claims about that output. Current products and standards provide sandboxes, approvals, hooks, traces, policy enforcement, attestations, bills of materials, and assurance representations, but these capabilities remain fragmented. This paper presents the Agile-V Assurance Spine, a cross-domain transition contract for software, firmware, and PCB engineering. Evidence is admitted only when it establishes required properties through an authoritative source profile, is bound to the exact artifact and frozen policy baseline, remains current with respect to declared dependencies, and satisfies risk-appropriate independence and authority. Gate decisions are recorded as receipts; approvals and exceptions are exact-scope and time-bounded; and authorization is rechecked at the effect boundary before merge, deployment, flashing, release, or fabrication. A bounded review of contemporary research, commercial platforms, open-source infrastructure, and standards positions the model relative to evidence-gated lifecycle control, continuous assurance, runtime admission, provenance, and AI/ML inventories. The paper contributes a precise vocabulary, compositional architecture, domain profiles, mapping to open-source implementations, and an adversarial evaluation agenda. It does not claim regulatory conformity or demonstrated production superiority.

arXiv AI Papershace 2 meses

Log-Depth Recurrent Language Modeling

Language modeling using Transformers has become commonplace despite their fixed computational depth and quadratic runtime with respect to input tokens. Recurrent models on the other hand offer linear depth but no parallel execution. In this work, we extend balanced-tree recursive operators from sequence encoding to autoregressive prediction, enabling all prefix representations to be computed with logarithmic depth and linear runtime. Our experiments provide an initial characterization of this model class, demonstrating robust length extrapolation and performance approaching that of ALiBi-based Transformers, highlighting its potential as an alternative architecture for language modeling.

arXiv AI Papershace 2 meses

Support-Compiled Feature Folding: More Evidence at Lower Memory Across Tabular Foundation Models

Tabular foundation models face a feature-side scaling dilemma: full-width pairwise mixing grows quadratically with the number of columns, whereas feature selection saves memory by discarding evidence. We introduce Support-Compiled Feature Folding (SCFF), a training-free inference framework that resolves this dilemma without changing the frozen backbone. SCFF routes support-ranked features through bounded leaves of the native feature encoder, support-checks the residual evidence, and merges the encoded messages before a single contextual prediction. It thereby converts quadratic feature-interaction work into linear-in-width work with a bounded local working set, without ensembling predictions or training new parameters. On the exhaustive 18-dataset wide-table slice of fixed AMLB-29, TabZilla, and TabArena snapshots, SCFF improves dataset-macro accuracy and NLL on all six evaluated backbones. All four matched-width comparisons retain favorable 95 percent dataset-bootstrap intervals on locked folds, with relative error reductions up to 26.1 percent. Median paired GPU-memory savings are 2.09x to 2.36x, and the ratio of separately observed maximum peaks reaches 34.3x. Under a measured peak-memory ceiling, SCFF uses the saved budget to preserve more support-selected evidence, improving accuracy by 4.06 and 3.72 points over the widest feasible single leaf on predeclared wide-Core strata of TabICLv2 and TabPFN-3.

arXiv AI Papershace 2 meses

Transferable Evidence Reconstruction for Longitudinal Glucose Representations

Long physiological recordings contain many routine measurements, while predictive information is often concentrated in rare events, sustained burden, and recurring temporal patterns. Masked autoencoding recovers measurements; contrastive learning aligns views. We study self-supervision that explicitly prioritizes structured signal evidence. We introduce transferable evidence reconstruction (TER), which constructs evidence from unlabeled recordings, fits a fresh low-capacity reader on one recording group, and requires that reader to recover the same evidence in another group without refitting. Differentiating through this cross-group test learns representations with transferable evidence-decoding rules; the evidence guides self-supervision but is not used as a downstream feature. For continuous glucose monitoring (CGM), an observation-aware daily encoder and clock-aware multi-day memory bind glucose level and change to recorded time while organizing up to seven days of history. On the 14-task leaderboard, TER improves the strongest prior overall PR-AUC/ROC-AUC/Macro-F1 scores by 5.51/4.43/2.80 percentage points and sets a new best metric on 12/14 tasks. These leaderboard gains are 2.0-2.9 times the respective gaps between the two strongest baselines. With public pretraining data, folds, and the linear probe matched, TER outperforms our GlucoFM reproduction by 6.09/5.52/2.72 points. Target-reader ablations, same-history controls, and cross-person readouts support the combination of structured evidence, cross-group reader fitting, and learned multi-day organization.

arXiv AI Papershace 2 meses

PASTABench: Proactive Assessment of Sequential Trajectories for Agent Safety

As Large Language Models (LLMs) evolve into autonomous agents that alter real-world states, ensuring operational safety across multi-step workflows has become a critical challenge. While recent work has moved beyond single-turn evaluation toward multi-turn paradigms, key limitations persist: step-level methods treat actions in isolation, missing how risks accumulate, while trajectory-level evaluations operate post-hoc, offering no opportunity for timely intervention. To address these limitations, we formalize Decoupled Proactive Safety Monitoring along three dimensions: whether to intervene, when to intervene, and what the risk is. We introduce PASTABench, a benchmark of 1,139 multi-turn trajectories spanning 5 risk categories and 13 subcategories. We further propose the Optimal Intervention Window (OIW), anchored by annotated Earliest-Signal and Trigger turns, to quantify intervention timeliness. Evaluation of 16 LLMs reveals that proactive intervention remains largely unsolved, with the best model achieving only 40.74% optimal-timing interventions. Fine-grained diagnosis further uncovers pervasive lexical overfitting: competitive safety scores of smaller models mask keyword hypersensitivity rather than genuine risk comprehension, as their proactive capability largely collapses once hazard vocabulary is neutralized.

arXiv AI Papershace 2 meses

Improving Ensemble Filters with Flow Matching

Data assimilation estimates a dynamical state from partial and noisy observations. Classical ensemble filters are efficient but restrict analysis updates through finite sample covariance and affine Gaussian distribution. We introduce the Flow Ensemble Filter (FlowEF), which uses conditional flow matching to transport the forecast ensemble from a classical baseline filter to an analysis ensemble. FlowEF uses a localized Gaussian source during training, transports forecast ensemble members from a baseline filter at deployment, and conditions its velocity field on ensembles from that baseline filter and the observation. The proposed model therefore learns a nonlinear update while mapping each baseline ensemble independently. For sparsely observed dynamical systems, FlowEF improves both deterministic and probabilistic metrics over all four classical ensemble filters. It also achieves the best performance among the state-of-the-art generative data assimilation models.

arXiv AI Papershace 2 meses

SoLiD26: A First Principles Solid-Liquid Interface Dataset for Machine-learned Interatomic Potentials

Machine-learned interatomic potentials (MLIPs) for solid-liquid interfaces in advanced materials applications, e.g., electrochemistry, catalysis and corrosion, require training data that samples both liquid environments, the solid and the interface itself. We present SoLiD26, a curated solid-liquid interface dataset, containing 15.4 million first-principles atomic structures with up to 576 atoms and 15 chemical elements for training and evaluating MLIPs. The structures were compiled from density functional theory (DFT) calculations performed in studies of solid-liquid interfaces, with most configurations originating from ab initio molecular dynamics (AIMD) simulations. Each record contains atomic species, positions, simulation cell, periodic boundary conditions, potential energy and atomic forces. SoLiD26 includes aqueous coinage metal interfaces, electrode-electrolyte systems, and selected bulk reference structures, calculated with VASP using the PBE functional and D3 dispersion corrections. We describe the data ingestion and preparation pipeline used to construct the dataset. The application of SoLiD26 for training and evaluating MLIPs is demonstrated with a suite of MACE models on a simple training, validation and test split. The dataset enables development and benchmarking of MLIPs for structurally and chemically heterogeneous solid-liquid interfaces.

arXiv AI Papershace 2 meses

Evaluating Open-Weight LLMs for Turkish Domain Documents Under Retrieval and Hardware Constraints

Most Turkish-capable large language models (LLMs) are evaluated using general-purpose benchmarks rather than long, structurally complex domain documents. This paper evaluates five open-weight 7B-8B models for Turkish document question answering under a resource-constrained local deployment setting. The primary benchmark contains 100 systematically validated questions derived from a 109-page industrial R&D report, and the evaluation protocol is replicated using a second 112-page public-sector report and an independently constructed 100-question set. All models are evaluated locally on an NVIDIA RTX 3050 laptop GPU with 6 GB VRAM using controlled prompting, decoding, and 4-bit quantisation. The principal methodological contribution is an evidence-annotated evaluation protocol that separates retrieval failure from downstream model reasoning failure without requiring additional model calls. On the primary benchmark, end-to-end accuracy ranges from 49% to 75%. Seven lexical, dense, and hybrid retrieval configurations are additionally compared using 95% Wilson intervals and exact paired McNemar tests; none significantly outperforms the character TF-IDF baseline on either document. Evidence recall saturates differently across the two reports, showing that retrieval and effective context capacity can be binding constraints for some documents but not others. These results demonstrate that model selection, retrieval behaviour, and hardware limits must be evaluated separately when deploying open-weight LLMs for Turkish domain documents.

arXiv AI Papershace 2 meses

Shared Global KV with Layer-Specific Local History

Decoder-only Transformer language models cache keys and values (KV) to reuse past computation during generation. Sharing KV across layers saves storage but reduces the diversity of representations available across depth. We study what local memory should retain alongside shared global KV, separating historical content from the input source used to form it. At 126M parameters and 2K context, an eight-seed study finds about 1.4% lower held-out test perplexity with local history than with a current-token local branch. Capacity, entry-count and training-compute controls support the value of historical content. In a two-seed comparison, this value persists when adjacent layers share local inputs while retaining independent projections; source sharing also shortens exact cache-construction dependencies. Against GQA and adjacent-layer KV sharing, equal bounded learning-rate searches and new-seed confirmation yield better same-source likelihood with larger caches and higher long-request latency. The ordering against adjacent-layer sharing persists after equal-token adaptation to 8K, with a short-context cost. The eight-seed external-book history effect remains uncertain, and downstream outcomes vary by task. We derive a sufficient suffix schedule that reduces upper-layer construction work while preserving the complete cache in exact arithmetic.

arXiv AI Papershace 2 meses

Controlled Attribute-Specific Summarization of Interrogative Dialogues

Effective summarization of interrogative dialogues is a critical task in forensic and investigative settings, requiring high factual accuracy, coherence, and attribute-specific relevance. In this work, we introduce CASPER, a novel Chain-of-Thought Attribute-Specific Prompting for Evaluative Summarization framework that leverages structured prompting and iterative refinement to generate high-quality summaries of interrogator-witness interactions. We construct MINDSum, a dataset extending the MIND corpus, comprising 6,000 utterance pairs annotated with event details, factual statements, character descriptions, and fillers. CASPER employs RoleEval, a hierarchical evaluation mechanism where multiple roles (officer, inspector, senior inspector) iteratively assess summaries based on predefined criteria. By integrating entity extraction and structured feedback loops, CASPER significantly improves factual consistency and contextual completeness compared to existing baselines. Experimental results demonstrate that our framework outperforms standard summarization models on both lexical (ROUGE) and semantic (BERTScore) metrics, while human evaluation confirms its alignment with expert reasoning. Our findings underscore the potential of controlled summarization in high-stakes domains, paving the way for AI-driven forensic intelligence.

arXiv AI Papershace 2 meses

Learning from Failures: Heterogeneous Graph Memory for Small Language Model Tool-Using Agents

Small and medium-sized language models offer cost-effective executors for tool-using agents, making them attractive for local and large-scale deployment. However, in long-horizon and stateful environments, they often make structural errors such as missing required observations, performing premature writes, repeating failed calls, and violating action preconditions. These errors can lead to incorrect state updates, policy violations, and costly or irreversible consequences, making reliable tool execution a critical deployment challenge. Existing fine-tuning approaches require substantial data and computation, while flat memory may retrieve failed actions without preserving their causal context or safety conditions. In this paper, we propose FRESH, a Failure-aware Retrieval framework over Experience-Structured Heterogeneous graphs, which transforms historical successes and failures into structured external experience for tool-using agents. By explicitly modeling the dependencies among tasks, actions, errors, repairs, and execution conditions, FRESH helps frozen language models reuse reliable strategies, avoid recurring failures, and make safer decisions in stateful tool interactions. Experiments on τ-Bench and AppWorld with multiple open-source models show that FRESH consistently improves task success and tool-use reliability over no-memory agents and representative memory-based baselines.

arXiv AI Papershace 2 meses

Compliant with Local Controls, Collectively Discriminatory. A Governance Architecture for Multi-Agent AI in Regulated Finance

Financial institutions are beginning to deploy agentic workflows in credit, fraud, collections, compliance, and operational control. Governance remains largely component-centric: each model or agent is specified, tested, authorized, and monitored locally. That is insufficient when institutional risk arises from the joint behavior of many locally acceptable components. We call this gap constitutional non-compositionality: local compliance checks need not compose into acceptable collective outcomes such as bounded disparate impact, market integrity, or traceable accountability. We propose ARIA as a finance-specific reference architecture and falsifiable research agenda for agent-population governance. It organizes six capabilities across normative-accountability, execution-control, and assurance-learning planes: policy specification, population-level observed-versus-expected behavior monitoring (M2), bounded authority, runtime containment, adaptive policy change, and preserved human oversight competence. Two simulations illustrate shared-signal thin-file exclusion under local controls and earlier warning from observed-versus-expected distributional monitoring in a constructed drift regime. The contribution maps these controls to fair-lending, EU AI Act, model-risk, and conduct-supervision evidence needs, and closes with a validation agenda rather than a production-effectiveness claim.

arXiv AI Papershace 2 meses

Task-Induced Riemannian Metrics for Vision Transformer Feature Spaces

Methods operating on Vision Transformer (ViT) feature spaces typically rely on Euclidean distance or cosine similarity. This assumes that every direction is equally meaningful, but there is no reason to believe the true task geometry has this property. The task-sensitive geometry of the feature space is given by the pullback metric g(F) = J(F)^J(F), where J is the Jacobian of the decoder's output fed to a task-specific distance, with respect to the features. Storing the full g is infeasible at modern scales, and for dense outputs such as depth maps even forming J is impractical. We show that whether a low-rank approximation of this metric can be learned depends on the model-decoder pair, and we characterize this with a matrix-free diagnostic κ_{cap}(r) computable with a low number of Jacobian-vector products. For tractable pairs, we develop the Spectral Pullback Network (SPN), which learns a low-rank version of the metric from randomized power iteration, and we distill it into a 310K-parameter importance head that predicts token importance directly from the features. When the Jacobian spectrum is too spread out for a low-rank approximation, passing the decoder's input features through a VAE bottleneck can restore tractability. Across DPT, DINOv2, CLIP, and VGGT backbones, κ_{cap}(r) predicts which learned-metric architectures are viable. The importance head reaches Spearman ρ= 0.998 on DINOv2 CLS, and our geometric token pruning reduces the additional depth error of ToMe-based token selection by 25\% on DPT depth at prune ratio 0.5, without fine-tuning the ViT. Project page: https://cyberiada.github.io/TaskInducedViTs/

arXiv AI Papershace 2 meses

PCQC: Privileged Counterfactual Question Credit for Multi-Turn Medical Dialogue

Large language models (LLMs) have made substantial progress on medical question-answering, yet effective medical dialogue also requires learning to ask questions that uncover relevant patient information. To train such dialogue policies, a common pipeline combines supervised fine-tuning with reinforcement learning (RL) based on final diagnostic correctness. However, this outcome-based supervision does not directly distinguish the contributions of individual questions and provides no question-level feedback for unexecuted alternatives. To address this gap, we introduce PCQC (Privileged Counterfactual Question Credit), which uses privileged patient information during training to learn from questions never asked. During training, PCQC makes alternative questions directly comparable at the same dialogue state by using privileged patient facts to construct their answers. A frozen diagnostic scorer evaluates the diagnostic utility of each resulting question-answer pair by how strongly it supports the correct diagnosis. PCQC turns these comparisons into relative question credit that teaches the policy which questions to favor, directly supervising both executed and unexecuted questions alongside outcome-based RL without requiring complete rollouts for the unexecuted alternatives. Extensive experiments across four medical benchmarks demonstrate that PCQC achieves 63.10% mean diagnostic accuracy, outperforming GRPO and ATPO by 4.38 and 4.21 percentage points, respectively. These gains are achieved with 33.1% fewer inquiry turns than GRPO.

arXiv AI Papershace 2 meses

Relative Discharge Stage (RDS) Classification: A Practical Indicator of Battery Discharge Progress

Accurate remaining discharge time (RDT) prediction is challenging in real-world battery applications because future load profiles are unknown and highly dynamic. To address the uncertainty of continuous RDT regression, this paper introduces Relative Discharge Stage (RDS), a battery-management indicator that represents the remaining discharge condition using five interpretable classes: Normal, Good, Moderate, Low, and Recharge Required. Unlike state of charge (SOC), which reflects the current charge level, RDS characterizes the remaining discharge process without requiring future-current information during inference. A physics-informed RDS classification framework is proposed, combining SOC estimation with lightweight temporal learning. The SOC-estimation component includes second-order ECM state and terminal-voltage prediction, hysteresis and OCV temperature correction, core-temperature estimation, and AEKF state correction, supported by OCV evaluation, online STC-ECM parameter adaptation, and pretrained neural residual-voltage correction. The measured current, terminal voltage, surface temperature, and estimated SOC are arranged into a sliding observation window and processed by a lightweight temporal convolutional network. Experiments on two public lithium-ion battery datasets demonstrate robust RDS classification, with accuracy exceeding 80% under varying load and thermal conditions.