Sección

Investigación

Papers y preprints de los feeds de investigación que seguimos.

6415documentos

From Memory Hits to Answerable Evidence: Query-Conditioned Memory Compilation for Long-Term Agent Memory

Long-term agent memory requires recovering answer-relevant evidence dispersed across turns and sessions. Existing systems increasingly use structured memories and adaptive retrieval, yet often pass retrieved or expanded units directly to the reader, coupling memory addressing with evidence presentation and leaving the final context redundant or incomplete. To address this gap, we propose MemCompile, a query-conditioned evidence compilation framework for long-term conversational memory. Given a query, the Address Plane locates relevant support through compact source-linked descriptors and capacity-aware Cells, while a normalization boundary resolves heterogeneous retrieval hits to stable SourcePacket identities. Its Evidence Plane then allocates a bounded evidence budget across complementary packets, admits limited additional support, and renders the retained claims associated with each selected packet into an answer-ready context. MemCompile achieves 80.52% Judge Accuracy on LoCoMo and extends the empirical quality-cost Pareto frontier to a higher-quality regime. Its strong performance on LongMemEval-S further suggests that the benefits of query-conditioned compilation generalize across long-term conversational-memory benchmarks.

arXiv AI Papershace 28 días

Self-Reflective Chain-of-Thought (SRCoT): Evidence-Based Text Generation via a Closed-Loop Retrieve--Reflect--Revise Pipeline on Edge LLMs

Large language models (LLMs) can generate fluent outputs that lack source support or conflict with local operational constraints. We introduce Self-Reflective Chain-of-Thought (SRCoT), a modular architecture that organizes generation as a closed Retrieve--Reflect--Revise loop. SRCoT separates environment state, persistent knowledge, accumulated experience, and short-term memory into provenance-labeled stores. BM25 retrieval supplies local context; subsequent passes expose intermediate drafts, missing-information markers, follow-up queries, and optional claim--evidence mappings. These artifacts inform retrieval, revision, clarification, or abstention decisions. We examine the architecture through qualitative, trace-based case studies in two fully synthetic settings: emergency medical services and Munich-Sim, an emergency-operations simulation. Local prototype runs on an Apple M4-family laptop with 128 GB memory illustrate source-labeled documentation, explicit gap marking, and structured dispatch-style outputs. The traces also reveal provenance-assignment errors, incomplete evidence coverage, and unsafe assumptions that remain possible despite prompting constraints. These cases demonstrate report-level inspectability and local execution, not comparative superiority, hallucination reduction, clinical validity, or operational safety. The architecture separates retrieval and control from the model interface, but behavior across different models remains to be evaluated. SRCoT is a research proof of concept requiring human review, with its principal contribution lying in an inspectable control procedure rather than a new model.

arXiv AI Papershace 28 días

Inferring L1 Backgrounds from L2 English with LLM Token Probabilities

Native language identification (NLI) infers a non-native speaker's first language (L1) from their second language (L2) production. We investigate a token probability approach that scores each text sample as a continuation of a prompt prefix specifying a candidate L1, then predicts the candidate with the highest score. We evaluate our approach using Llama 3.1 8B and 70B on three datasets: TOEFL9, Write and Improve 2024, and BULATS-L1. We compare its performance with previous prompting-based classification approaches. Token-probability scoring substantially improves on direct classification, particularly with the 8B model, while the 70B model achieves state-of-the-art performance across all datasets. Our results indicate that pretrained LLMs' token probabilities contain information useful for distinguishing L1 backgrounds from L2 English production.

arXiv AI Papershace 28 días

MobileInject: Evaluating Reasoning-to-Action Security of MLLM-Based Mobile Agents under Cross-Modal Consistency Injection

Multimodal large language model (MLLM)-based mobile agents are rapidly evolving from conversational assistants into autonomous systems that perceive, reason over, and act on smartphone environments. This capability also exposes a new security risk: adversaries can embed malicious objectives in cross-application mobile content and steer agent behavior through indirect multimodal injection. Although prompt injection has been studied in desktop and web settings, its cross-modal properties in mobile ecosystems remain poorly understood, and existing security analyses are still limited. To close this gap, we introduce MobileInject, the first benchmark for evaluating mobile agents against multimodal injection attacks and reasoning-to-action security risks. MobileInject provides a high-fidelity interactive mobile evaluation framework covering 11 scenarios, 9 security-risk categories, and 108 executable risk instances. It further introduces Cross-Modal Consistency Injection to exploit corroboration across modalities, two enhanced attack methods, Multi-Role Collaborative Deception and Reasoning-Camouflage Injection, to intensify trust manipulation and reasoning hijacking, and a Reasoning Security Diagnosis Framework for fine-grained failure analysis. Systematic experiments show that mobile agents are broadly vulnerable to multimodal injection, and that many failures stem from reasoning hijacking at early stages. These findings establish MobileInject as a systematic cross-modal benchmark for studying and improving the security of mobile agents. Code release details are provided in the Appendix Open Science section.

arXiv AI Papershace 28 días

SimOPD: Relational Knowledge Transfer for On-Policy Distillation

A key challenge in on-policy distillation (OPD) is extracting effective supervision from teacher knowledge. Supervision from token probabilities may underutilize internal knowledge, while direct hidden-state matching forces the student to shift to the teacher's feature space and complicates cross-architecture transfer. Inspired by contrastive learning's use of relationships between instances, we propose SimOPD to transfer teacher internal knowledge through response similarities. Specifically, SimOPD uses paired memory banks of student-generated responses to extract supervision from the teacher's internal representations as similarity distributions. The student then matches these distributions within its own representation space, retaining representational freedom and removing the need for cross-model projection. Extensive experiments show that SimOPD improves the overall average over the strongest baselines by up to 2.6 and 3.1 percentage points in same- and cross-architecture settings.

arXiv AI Papershace 28 días

Designing for Imperfect Prompts: A Simulation-Based Evaluation of an LLM-Based Enterprise Data Assistant Using Knowledge Graphs

Enterprise data assistants using large language models (LLMs) promise democratized data access in organizations. Yet in our real-world deployment within the CFO unit's data lake at an industry partner, we observed that business users frequently pose imperfect prompts that include domain-specific vocabulary and lack schema references or structural information. This Design Science Research study examines automatic resolution using a semantic mediation layer. We design and evaluate two design variants: schema-based mediation via semantic similarity, and knowledge graph-based mediation leveraging an automatically constructed knowledge graph from the data lake. Our simulation-based evaluation using real-world data, supported by LLM-as-a-Judge assessments and human validation, shows that knowledge graph-based mediation outperforms the schema-based approach in Precision, Recall, F1, and preserving user intent. We propose a design principle for knowledge graph-based mediation, demonstrating that structured knowledge representation bridges user intent and database structure and guides the design of enterprise data assistants across skill levels.

arXiv AI Papershace 28 días

Look, Reason, Draw a Region: Recursive Refinement for Spatial Context Prediction

Spatial context prediction requires visual grounding and spatial reasoning to identify vacant regions relative to objects in a scene. Spatial relationships and physical constraints remain challenging for large vision-language models (VLMs). Recursive reasoning has demonstrated strong performance on discrete and abstract reasoning tasks, motivating its use for improving spatial context prediction in natural images. Weight sharing enables higher accuracy with a small trainable parameter count. We introduce Spatial-TRM, a compact recursive model with frozen vision-language features, a convex region decoder, and just 11.2M trainable parameters. This weight-tied reasoner supports six spatial relationships, three reference frames, and open-vocabulary anchor objects. On the RoboSpatial-Home benchmark, Spatial-TRM achieves 29.51% context accuracy, a gain of 9.02 percentage points (44.0% relative) over a separately trained single-step model with the same parameter count. Including the frozen encoders, the model has 877.8M total parameters, approximately 15fewer than the reported 13B size of RoboPoint. These results show that repeated refinement improves spatial context prediction from pretrained vision-language features at a fixed trainable parameter count on this benchmark.

arXiv AI Papershace 28 días

Optimist Deep Research: An Iterative, State-Centric Agent for Autonomous Literature Research

Most retrieval-augmented systems treat literature search as a single retrieval step followed by summarization. Real literature research works differently: the investigator's understanding of the question changes as they read, new subquestions appear, and evidence is weighed, not just collected. We present Optimist Deep Research, an open-source agent that runs literature research as an iterative evidence-acquisition process over an explicit, mutable research state. A research question is broken into a tree of subquestions, each served by search intents that are translated into source-specific queries over arXiv, OpenAlex, Crossref, PubMed, Semantic Scholar and the web. Results are merged into a canonical, provenance-preserving document registry through a conservative four-tier deduplication hierarchy. The registry is expanded by citation snowballing until a saturation criterion is met, and then screened, ranked and allocated to subquestions round-robin. Documents are read in parallel into a structured evidence store of claims with support, location, type and strength. A critic reviews the state along eight dimensions and returns research actions rather than a binary verdict: it can add, split, rephrase, deprioritize or close subquestions, so the research plan is an output of the investigation as well as an input. The final report is written section by section from the evidence store, and every citation resolves to a retrieved document. We describe the design, the deterministic machinery that keeps a language-model pipeline auditable, and a case study that read 28 of 237 retrieved documents in about three and a half minutes. We also report weaknesses the case study exposed.

arXiv AI Papershace 28 días

How the elastic axis folds the transonic flutter boundary: shock energy transfer and conformally calibrated active learning

Isogai's case A, a pitch-plunge aerofoil whose elastic axis lies half a chord ahead of the leading edge, has a folded transonic flutter boundary in inviscid flow: near the bottom of the transonic dip several flutter speeds coexist at one Mach number. We map this boundary, and the family obtained by moving the elastic axis, with coupled unsteady Euler simulations, a Gaussian-process surrogate of the damping coefficient trained by batch active learning, and split conformal prediction on independent random simulations, which turns the surrogate into stability calls with a distribution-free error rate. Three failure modes of time-domain damping labels that leave no trace in the residuals are documented, and the half-degree mean incidence left by the common one-degree start is shown to raise the flutter speed at the dip by 2.8%, more than any numerical setting. With 80 simulations the learned case A boundary labels 64 of 65 independent runs correctly, and calibrated calls are wrong in at most 1.6% of cases. Over Mach number, elastic axis and flutter speed index, 193 simulations and 40 calibration runs show that the fold exists only for elastic axes more than about 0.4 chord ahead of the leading edge; it closes between 0.39 and 0.33 chord ahead, on two meshes. Chordwise energy budgets, closed against the solver loads and the growth rate within a few per cent, explain the boundary. The lower branch is crossed through the reduced frequency at unchanged shock phase. On the upper branch the shock is pinned at the trailing edge, and moving the elastic axis aft brings the pivot of the flutter mode towards the section, so that the damping of the supersonic surface falls faster than the residual shock input and the fold closes.

arXiv AI Papershace 28 días

FlashReFocus: Interactive Image Refocusing in Seconds

Single-image refocusing changes the focal plane and depth of field of a photograph after capture. It is inherently interactive: users try many focal points and blur strengths on one image, so each edit should be fast and change only the blur. Recent generative methods, however, rerun large diffusion models for every edit. We observe that the two sub-problems differ in kind: recovering the all-in-focus image is ill-posed and needs a generative prior, whereas rendering blur is largely determined by depth and the focal plane. We therefore propose FlashReFocus, which runs diffusion once per photograph and renders every edit with a lightweight network. To address the ill-posed restoration, a one-step diffusion deblurrer with a decoder refiner recovers fine detail efficiently while balancing fidelity and perceptual quality. For precise control, the renderer is conditioned on an explicit defocus map and uses defocus-bias attention, which lets pixels at similar defocus share information without interference from pixels at other depths; in-focus content stays unchanged, and editing the map yields tilt-shift and two focal planes. To further test generalization to unseen cameras, we collect 3CReal, a three-camera benchmark of 89 pairs. FlashReFocus achieves the best perceptual fidelity for zero-shot bokeh rendering on EBB! and 3CReal, refocuses 148.3x faster than GenRefocus, and takes about half a second per further edit, 462x faster.

arXiv AI Papershace 28 días

Learning angles for quantum approximate optimization

The Quantum Approximate Optimization Algorithm (QAOA) and its variants are leading algorithms to help solve challenging combinatorial optimization problems. QAOA maps these problems to Ising Hamiltonians such that their ground state is the desired solution. Candidate solutions are then drawn from a variational quantum circuit. The performance of QAOA hinges on a set of variational angles. However, finding good angles is an instance-specific and computationally expensive non-convex classical optimization task. Here, we learn the mapping from the problem instance to high-quality QAOA angles. Our learned models replace the per-instance optimization, that we call the solver, with a single forward pass that is several orders of magnitude faster than the solver with a near identical quality. Our models further generalize to unseen topologies such as Barabási-Albert and 2D grid graphs. Experiments on quantum hardware confirm that QAOA samples with angles produced by our open-source models are indistinguishable from solver angles. Crucially, models sensitive to the topology of the problem instance increase the quality of the QAOA angles. Our work amortizes a major bottleneck in QAOA by an up-front data generation and training cost. Importantly, this cost can be amortized over time in settings where many similar problems are optimized on a regular basis, as is often the case in applications for industry.

arXiv AI Papershace 28 días

The Optimal Ranking Tradeoffs Between Two Ranking Scores in Two-Class Classification

The accuracy, true negative rate (negative class recall), true positive rate (positive class recall), negative predictive value (negative class precision), positive predictive value (positive class precision), class-specific intersections over unions, class-specific F-scores, Cohen's kappa, and balanced accuracy are scores that one intuitively wants to maximize when developing machine learning methods. However, they lead to different rankings of such methods. So, how can we establish a ranking that is an optimal tradeoff between any pair of these scores? It turns out that all these scores are special cases of ranking scores: a continuum of scores parameterized by the relative importance values assigned to the true negative, false positive, false negative, and true positive outcomes. Since all these scores are known to induce meaningful rankings, we seek the optimal tradeoffs among them. We consider the tradeoff between two scores to be optimal when it minimizes the Fréchet variance regarding the Kendall distance (also known as the bubble-sort distance) between rankings. For any given set of classifiers to rank, the optimal tradeoff lies at the center of the shortest (geodesic) path on the manifold of all rankings induced by the ranking scores. However, in the parameter space of the ranking scores, it is rarely achieved by averaging the importance values corresponding to the initial scores. In this paper, we present an exact analytical solution for determining a ranking score that is an optimal tradeoff for any given pair of initial scores, in the common case where the class priors are the same for all compared performances. We also experimentally investigate what happens when performance follows some distributions with fixed priors.

arXiv AI Papershace 28 días

Distributed Performance Boosting through Neural Control for Plug-and-Play Networks

We address the problem of improving the performance of pre-stabilized networked systems subject to plug-and-play (PnP) operations, namely additions or removals of subsystems. While distributed baseline controllers can preserve stability after a PnP transition, augmenting them with expressive neural policies can invalidate the formal stability guarantees. To overcome this issue, we build on recently proposed Performance Boosting controllers based on the Internal Model Control (IMC) principle, which reduce closed-loop stability to a finite _2-gain requirement on an auxiliary operator. In the distributed setting, this operator is formed by interconnecting local neural operators that exchange information only between neighboring subsystems. We then derive free parametrizations of this interconnection, namely explicit maps from unconstrained trainable variables to interconnected neural operators satisfying a prescribed network gain bound for every parameter value. The obtained distributed neural controller can therefore be trained by unconstrained gradient-based optimization without stability constraints or projection steps, while retaining the prescribed gain bound at every training iterate. We develop two complementary parametrizations. A neighbor-based compositional parametrization uses local coupling bounds, so that after a PnP event only subsystems whose bounds change need to update their gain parameters. A network-norm parametrization instead uses a bound on the overall coupling matrix and can reduce conservatism, at the price of requiring network-wide information. A PnP DC-microgrid case study shows that the proposed neural layer improves voltage transients under topology changes and load variations while preserving nominal closed-loop stability.

arXiv AI Papershace 28 días

CoTA++: Understanding the Repeat Curse in dMLLMs from an Information Flow Perspective

Diffusion-based multimodal large language models (dMLLMs) decode by iterative denoising with bidirectional attention, which precludes the standard autoregressive key--value cache. Practical deployments therefore rely on approximate caching that reuses intermediate states across denoising steps. Such caching, however, often makes a response repeat one token at adjacent positions in a run that breaks the sentence, a failure we term the Repeat Curse. Tracing the context-token information flow reveals three disruptions. Within a step, caching disperses the attention that anchors a position on nearby context tokens. Across steps, similarity-based refresh is not staleness-aware, so some positions are decoded against stale key--value states. Across layers, the deep-layer entropy of the context fails to converge where repetition arises. We accordingly propose CoTA++, a training-free, plug-and-play framework that counteracts each disruption where it originates: Context-Token Attention Re-anchoring re-forms the attention routing against the current context, Decode-Aware Refresh reserves part of the cache's refresh budget for the positions about to be committed, and Context-Token Entropy-Guided Voting discounts the confidence of a position by the deep-layer entropy of its context, which reorders the unmasking. The three components act at non-overlapping stages of a decoding step and compose freely with existing dMLLMs and caches. Across three dMLLMs and two caching backends, CoTA++ reduces the sample repetition rate of long responses from up to 99.6\% to at most 17.0\%, recovers most of the accuracy lost to caching on nine benchmarks, and keeps the speedup of caching at under 15\% overhead, making approximate caching reliable for dMLLMs.

arXiv AI Papershace 28 días

Data-Efficient Classifier Pretraining Through Diffusion Models

Machine learning with limited data is susceptible to overfitting, especially with modern deep neural networks. Prior to the introduction of AlexNet, reconstruction-based pretraining with auto-encoder and denoising auto-encoder were found to be beneficial, partially due to the limited size of the training sets at the time. Motivated by these traditional approaches, we propose DiffPT, a simple diffusion modeling approach to pretrain a given classifier backbone for limited domain-specific training data. We experiment on time-series data, standard natural-image benchmarks, and medical imaging datasets, and contrast the diffusion approach to a multitude of modern and traditional un/self-supervised methods. While contrastive pretraining performs better on CIFAR-10 and Tiny ImageNet in low-data regimes, DiffPT is particularly effective on medical imaging and time-series tasks, where it outperforms self-supervised baselines across a range of low-data regimes. Our results suggest that DiffPT is a simple and effective approach for data-efficient classifier pretraining on different modalities.

arXiv AI Papershace 28 días

Annotrace: A Workbench for Step-Level Annotation of Agent Traces

Every study of why LLM agents fail begins by labelling traces by hand, and every such study builds its own throwaway tool. Observability platforms record agent traces but offer no structured labelling; annotation tools label text but have no notion of a step, a delegation or a tool call, leaving it unclear what the smallest unit of execution to label should be. Annotrace reads six trace formats, including OpenTelemetry GenAI, LangSmith and Langfuse, and turns each run into an ordered, nested list of steps. The annotator writes their own labelling codebook; a frontier model proposes a starting set from a sample of their own traces, answers first, and is asked exactly the questions the human is asked, making annotation a joint human-AI process. Annotrace measures agreement between any two annotators, human or AI. On a multi-agent question-answering corpus it reproduces a previous manual effort's 235 annotation units exactly, and two models labelling the same traces agree 91% of the time at κ= 0.82.

arXiv AI Papershace 28 días

Verbalising Model's Computation for Chain-of-Thought Faithfulness Detection

Current Chain-of-Thought (CoT) verification methods detect reasoning faithfulness from model outputs or static activations, but are limited in explaining how unfaithfulness develops during generation. We introduce VERITAS to verbalise the model's computation: how different reasoning anchors emerge, transfer across layers, and surface in the output. VERITAS adopts the layer-wise Jacobian projection to produce a verbalised account of the model's internal reasoning evolution at each generated CoT token. By training a classifier on the extracted features along the traces, we show that these structured features provide insightful perspectives on how and where reasoning diverges toward unfaithfulness. Experiments on four datasets from FaithCoT-Bench and BonaFide show that VERITAS achieves a significant performance gain over LLM-as-Judge, with a 28.4% gain in balanced accuracy, and an average improvement of 16.0% over activation probing. Further analysis shows VERITAS can be generalised to different datasets and model architectures. Our work shows that verbalising model computation provides a richer and more interpretable lens on CoT faithfulness, moving beyond detection toward a mechanistic understanding of how reasoning fails.

arXiv AI Papershace 28 días

DeDiCa: A Scalable Deconfounding Discrete Causal Generative Model

We introduce DeDiCa, a practical framework for efficient causal inference with discrete variables and hidden confounding that revisits classical probabilistic modeling through a tensorized parameterization of causal Bayesian networks, enabling tractable inference. Given a causal graph, DeDiCa is trained once on observational data by maximum likelihood, exactly marginalizing over hidden confounders, and can then estimate any (proxy-)identifiable interventional query. Despite this deliberately simple parameterization, DeDiCa matches state-of-the-art neural causal generative models in ATE estimation, while supporting memory-efficient CPU training and exact interventional inference in milliseconds. Compared with causal foundation models, DeDiCa scales more effectively with additional observational samples and achieves competitive accuracy at substantially lower inference time and memory. DeDiCa provides a tractable, accurate, and scalable approach to causal inference in discrete settings with hidden confounding.

arXiv AI Papershace 28 días

NestedVLA: Learning to Consolidate and Generate Skills for Vision-Language-Action Models

Vision--Language--Action (VLA) models promise generalist robotic agents, yet consolidating skills learned across heterogeneous tasks and environments remains difficult: training data is fragmented across platforms, joint multi-task training suffers from interference, and model merging tends to introduce parameter conflicts or require architectural changes. We instead advocate a paradigm in which skills are first learned independently and then consolidated post hoc. To this end, we propose NestedVLA, which replaces parameter aggregation with parameter generation: a nested hypernetwork, conditioned on the target task, synthesizes task-adaptive action-expert parameters that compose with a frozen base VLA. Task-specific compressed VLM residuals are retained separately. This formulation enables flexible task-conditioned skill reuse while mitigating interference. Across three RL benchmarks (MetaWorld, ManiSkill, CALVIN), NestedVLA outperforms the strongest model-merging baseline by 41 pp on average, matches per-task fine-tuning within 1 pp on ManiSkill, and substantially closes the gap on CALVIN, pointing toward a storage-efficient route to skill consolidation. Code is available at https://github.com/WanyuGroup/NeurIPS2026-NestedVLA.

arXiv AI Papershace 28 días

Beyond Marginal Influence: Interaction-Aware Memory Selection for LLM Agents

LLM-agent memory selection often uses retrieval rankings based on record relevance or individual influence. Such rankings cannot identify sufficient subsets under redundant or jointly necessary evidence. We study per-decision memory sufficiency: selecting a small subset that preserves full-memory behavior. To separate search error from model error, we introduce JointCoreBench, a benchmark with nine interaction families, six workflow domains, natural-language decisions, and exact sufficient-set oracles, and JointCore, a relevance cost-ordered group-deletion baseline that tests blocks before individual records. On 6,912 exact conditions, cost-ordered JointCore recovers a minimum-cost sufficient set in 100.0% of cases, reduces subset evaluations from 30.00 to 11.13 (62.9%), and achieves 87.8% active-memory savings. Across four 12B-70B model families and 432 tasks per model, its unweighted mean is 10.28 versus 16.00 model queries and 50.6% versus 43.1% minimum-cost recovery relative to cost aware singleton deletion. Both methods preserve full-memory behavior under the deterministic primary interface. Tests with physical deletion, UNKNOWN masks, PENDING sentinels, counterfactual values, and alternative labels identify interface dependent behavior. JointCore is an efficient benchmark baseline; the study does not establish deployment-ready memory compression.