Noticias de IA - ítems

arXiv:2610.34044v1 Announce Type: cross Abstract: Entropy in reinforcement learning with verifiable rewards (RLVR) moves up or down depending on choices that should be neutral, such as whether over-length responses are masked or counted as failures. We trace this to sampler mismatch, where rollouts are drawn with top-k/top-p truncation or at a temperature the learner does not apply, and prove that a constant advantage shift then changes the entropy of a softmax policy, at first order and in expectation, with a sign fixed by the sampler: positive shifts sharpen the policy under any sampler whose likelihood ratio to the policy increases with token probability, including top-k, top-p, min-p, and temperatures below one, and flatten it under flatter samplers. In group-based RLVR with token-level losses, the baseline-controlled part of this drift is set by the token-weighted net advantage, which masking and failing over-length responses can push to opposite signs; the remainder, a covariance between advantages and the sampler's sharpness gap, can feed back over long horizons. On a 2B model, the entropy slope is linear in shifts of the net advantage under truncated sampling and indistinguishable from flat under exact sampling, and token-level importance ratios leave 98% of the drift. In 100-step runs, a net-zero baseline only delays an entropy explosion that exact sampling does not show, a fixed positive shift delays it further, and a baseline weighted by the sharpness gap keeps entropy stable without changing the learner. On a 9B reasoning model over ten steps, the uncorrected rule loses held-out accuracy, whereas zeroing the net advantage or renormalizing the learner gains 8.2 and 8.8 points over it, with no held-out effect under exact sampling.

arXiv:2610.34043v1 Announce Type: cross Abstract: Autonomous agents are increasingly deployed across sequences of related tasks, yet they often fail to retain and improve their knowledge over time. Existing approaches convert interaction trajectories into memories or reusable skills, but experience is not automatically reliable knowledge. Automatically collected skills and memories can propagate errors, mislead, and corrupt useful knowledge. To address this gap, we introduce Knowledge Evolution from Expertise and Lessons (KEEL), an agent-agnostic framework that uses verifier evidence to transform successful behavior into reusable skills and failures into corrective lessons. As new evidence accumulates, individual skill claims can be supported, contradicted, replaced, or retired. On SkillFlow benchmark, KEEL improves task success by 4.2 to 21.7 percentage points over the no-knowledge baseline, compared with gains of 2.4 to 9.6 points from SkillFlow's default method. Ablations show that skills and lessons are complementary. These results show that reliable continual agent learning depends not simply on accumulating experience, but on continually validating and evolving the knowledge derived from it. Our code is available at: https://github.com/microsoft/KEEL-TTCL26.

arXiv:2610.34021v1 Announce Type: cross Abstract: Structured scientific spectra encode rich physical information while obeying hard constraints, such as conservation laws and characteristic spectral geometry. Accurate prediction of these spectra is central to materials discovery, yet existing methods often treat them as unconstrained sequences and therefore fail to enforce the underlying physical structure. Taking electronic density of states (eDOS) as the prototypical example, we introduce DeepSciReasoner, a general paradigm for predicting scientific spectra under physical constraints. The framework combines structure-aware spectrum decoding with constraint-preserving physical reasoning, allowing predictions to capture rich spectral structure while respecting the underlying physics. We evaluate DeepSciReasoner on eDOS, phonon density of states (phDOS), X-ray absorption near-edge structure (XANES), and Raman spectra, where it substantially improves prediction accuracy while maintaining physical consistency. These results establish DeepSciReasoner as a reusable blueprint for structured scientific spectrum prediction under hard physical constraints.

arXiv:2610.34018v1 Announce Type: cross Abstract: We present BriLLM-MetaPred, the first engineering implementation of the MetaPred machine-learning framework established by a mathematical theory of AGI, instantiated for language modeling. The theory argues that MetaPred is both necessary and sufficient as a machine-learning primitive for general cognition. Taking this theoretical conclusion as a design premise, we report an implementation of the primitive for language modeling. A single continuously evolving dynamical state serves as both the context representation and the state from which predictions are read out. Continuous measurement of its own state reduces finite-depth inconsistency, maintains internal consistency, and provides an endogenous dynamical self-supervision signal in fixed-point form. The single dynamical state of MetaPred intrinsically couples training and inference. MetaPred therefore exhibits symmetry across almost all key elements of machine learning, an intuitive and observable property of the mathematical optimum over the space of all machine-learning paradigms. BriLLM-MetaPred is neither another recurrent language model nor an extension of BriLLM. It provides a bridge between two prior contributions: the construction and motivation of the MetaPred primitive provided by AGI theory, and the SiFu learning paradigm provided by BriLLM (Zhao et al., 2025). The latter's node-signal-flow design enables the primitive to run on available software and hardware accelerators. Our implementation of BriLLM-MetaPred achieves predictive performance comparable to a GPT baseline of the same size. This provides initial evidence for the engineering feasibility of the MetaPred framework established by the mathematical theory of AGI. MetaPred is the first machine-learning paradigm derived entirely from mathematical theory rather than assembled and incrementally improved through human design---and perhaps the last such paradigm.

arXiv:2610.34005v1 Announce Type: cross Abstract: On-policy self-distillation (OPSD) trains a student on its own reasoning trajectories using token-level supervision from a reference-conditioned teacher initialized from the same model. However, access to the reference can induce teacher behaviors that undermine independent reasoning in thinking models, leading to reference hallucinations, where the student cites an unavailable reference to justify reasoning steps, and reconsideration suppression, where the teacher discourages verification and revision at uncertain steps. We introduce Steered-Teacher OPSD (ST-OPSD) to jointly address these failure modes through activation steering. ST-OPSD extracts two directions from model-generated reasoning trajectories and applies them to the teacher's hidden states during distillation target computation, suppressing reference-invoking behavior and encouraging reconsideration with marker-family calibration. The student learns from the resulting teacher distributions without being steered. Experiments on AIME and HMMT show that ST-OPSD improves accuracy over vanilla OPSD and other interventions while restoring reconsideration and reference-hallucination rates to base-model levels.

arXiv:2610.33968v1 Announce Type: cross Abstract: Generative Flow Networks (GFlowNets) are trained to sample objects from compositional spaces with probability proportional to a reward function, but ineffective exploration can cause learned samplers to concentrate on a few high-reward regions. An intuitive remedy is to train multiple GFlowNets independently in parallel and combine their samples. However, this does not prevent multiple models from collapsing onto the same high-reward areas, leaving large regions of the state space unexplored. With this in mind, we propose Hierarchical Cooperative GFlowNets (HCG), an approach that dynamically partitions the search space to coordinate discovery without relying on hand-crafted boundaries. HCG imposes an ordering in which the highest-ranked model may explore the entire space, while lower-ranked models are progressively pushed toward regions not covered by higher-ranked ones. We provide a theoretical characterization of how specific learning objectives can shift probability mass away from designated areas, instantiating this mechanism within our architecture. A shared classifier coordinates the hierarchy by partitioning the search space, keeping the GFlowNets independently parameterized and enabling asynchronous training. Furthermore, we derive a normalizer-weighted aggregation rule for the ensemble and prove that it recovers the target reward-proportional distribution under the stated assumptions. Finally, experiments across multimodal distributions, sequence generation, combinatorial optimization, and biological sequence design show that HCG improves mode coverage and target approximation, translating increases in the number of coordinated samplers into broader, complementary exploration.

arXiv:2610.33964v1 Announce Type: cross Abstract: AI foundation models play a pivotal role in advancing heart disease diagnosis from electrocardiogram (ECG) data, particularly for rare diseases where the lack of sufficient labeled data hinders supervised learning. However, most self-supervised learning methods treat ECGs as generic sequence or image data without exploiting underlying physiological structure. To address this limitation, we propose QECGBERT, a self-supervised ECG language model that explicitly encodes physiological semantics by tokenizing the P, QRS, T, and isoelectric background waveforms. Vector-quantized variational autoencoder (VQ-VAE) tokenization is applied to each 12-lead ECG, with tokens enriched by continuous waveform features to preserve morphology lost during quantization without increasing sequence length. This design enables compact BERT masked language modeling with short sequences, improving scalability and reducing computational cost relative to long-sequence approaches. QECGBERT applies BERT pre-training to the MIMIC-IV dataset containing over 800,000 unlabeled ECG recordings, and fine-tuning on labeled PTB-XL, CPSC2018, and Chapman--Shaoxing--Ningbo datasets for evaluation. The results demonstrate that QECGBERT achieves comparable or superior performance to published benchmarks, offering a scalable, physiologically aware approach to ECG foundation modeling for cardiac diagnosis.

arXiv:2610.33962v1 Announce Type: cross Abstract: Tool-using large language models (LLMs) must decide whether external information warrants action. We investigate whether their expressed credibility judgments reliably indicate behavioral resistance to misinformation. We evaluate five LLMs on 196 constructed false notices paired with legitimate counterparts and 204 archived attacks adapted to simulated tool use. When asked to act on the constructed notices, models issue harmful tool calls in 27%--99% of cases they have rated untrustworthy. Removing the preliminary rating step increases harmful execution by 15 percentage points on average, showing that credibility elicitation itself changes the behavior being evaluated. Interventions that improve credibility discrimination do not consistently yield safer actions. A simple forewarning significantly reduces harmful execution in four of five models, but neither tested generated-example prebunk outperforms it in the main evaluation. These reductions can also compromise legitimate task completion. Supervised fine-tuning on generated false notices reduces harmful execution on held-out archived attacks under the rating-first protocol, with model-dependent utility costs and weaker protection without rating elicitation. Our findings motivate evaluating misinformation defenses through both harmful actions and legitimate task completion, while accounting for the behavioral effects of the evaluation protocol.

arXiv:2610.33956v1 Announce Type: cross Abstract: Continual learning is the ability of LLM agents to improve over time by acquiring, refining, and applying knowledge from experience while retaining useful prior knowledge. It is essential for agents learning over long time horizons, yet current evaluations largely assess individual task performance rather than improvement through accumulated experience. We introduce Continual Learning Exam (CLE), a large-scale benchmark developed by 50+ domain experts to measure experience-driven improvement in realistic professional workflows. CLE comprises 120 human-authored tasks, each organised as a five-level ladder, yielding 600 levels spanning 89 professional roles across 12 occupational domains. Agents must pass each level to unlock the next, with later levels building on experience accumulated during earlier assignments. An evaluation agent assesses each submission against human-authored rubrics. We compare a single-submission setting with one that provides diagnostic feedback after failure and allows one revision. Across 17 harness-model configurations using frontier LLMs, feedback and revision yield a median completion gain of 5.3 percentage points. Yet the best-performing system clears only 47.7% of levels and completes all five levels in just 23.3% of tasks, highlighting the difficulty of sustained learning from experience for existing frontier LLMs. As such, the dataset can also be used to post-train LLMs to strengthen their continual learning capabilities.

arXiv:2610.33950v1 Announce Type: cross Abstract: Large language models increasingly act as partners in everyday communication. In these interactions, useful assistance may require not only answering, but also asking the right questions that surface knowledge gaps, assumptions, and alternatives. In this study, we refer to such epistemic question-asking as model curiosity, and introduce a theory-grounded taxonomy that distinguishes curiosity-oriented questions from generic model questioning. Across four conversation datasets, we find that curiosity-oriented questions remain rare (17-28% of all questions), but are associated with more exploratory and epistemically rich subsequent user responses, with varying patterns across datasets. We further conduct a human study to evaluate the usefulness of model curiosity, finding that curiosity can encourage exploration, especially in advice, creative, and research contexts. These findings have implications for building LLMs that ask more selectively to support deeper exploration and human-AI collaboration.