arXiv:2606.05541v1 Announce Type: new
Abstract: Cross-linking mass spectrometry (XL-MS) has emerged as a powerful quantitative technique for probing intra-protein structural information as well as protein-protein interactions at an unprecedented scale. XL-MS data yield information on the pairwise spatial proximity of proteins through inter-molecular linkers. However, systematic methods for adapting such data for coarse-grained interacting particle models remain limited. Predominant focus is put on directly fitting radial distribution functions (RDFs), while numerous observables, e.g. coordination numbers, which are functionals of the RDF, cannot be uniquely inverted. In this work, we develop a framework for parameterizing interaction potentials from such observables in potentially phase-separated mixtures, as encountered in XL-MS results. We establish a connection between this problem and the inverse Henderson problem and adapt algorithms such as Iterative Boltzmann Inversion and Iterative Monte Carlo to its numerical solution. We derive exact and low-density limit gradient approximations and propose two new algorithms based on an adaptation of the predictor-corrector~framework. In total, we evaluate several optimization algorithms on biologically realistic ten-component test systems. We demonstrate that for homogeneous fluids, all methods achieve exceptional efficiency and accuracy. Critically, we further demonstrate successful parametrization in a challenging three-phase system. Here, three algorithms, namely Adam and gradient descent employing the low-density derivative as well as Newton's method with the exact gradient, reliably recover the correct parameters. These results establish a clear pathway from XL-MS experiments to coarse-grained protein models for systems where phase separation governs biological function, potentially enabling new investigations of biomolecular condensates and protein aggregation.
Science Journals
arXiv:2606.05724v1 Announce Type: new
Abstract: Long-form narrative QA requires reasoning over evolving story worlds rather than isolated passages: answers may depend on earlier goals, changing character states, social relations, causal triggers, temporal position, and later consequences. Existing retrieval and graph-augmented generation methods improve evidence access, but their units--chunks, entities, relations, summaries, or tool actions--do not directly encode how evidence functions in a story. We introduce Narrative Knowledge Weaver(NKW), a source-grounded framework that aligns textual evidence, atomic facts, canonical graph structure, entity profiles, interactions, episodes, and storylines. At query time, NKW uses text, graph, and narrative tools with post-retrieval reading skills to assemble evidence and audit actor, scope, polarity, state, and temporal constraints. Across STAGE, FairytaleQA, and QuALITY, NKW is strongest on screenplay-level story-world QA while remaining competitive on more passage-centered benchmarks. Ablations, question-type analyses, graph-asset statistics, and case studies show complementary benefits for character, scene, temporal, causal, and narrative-progression reasoning.
arXiv:2606.05586v1 Announce Type: new
Abstract: In remote sensing object detection, Convolutional Neural Networks (CNNs) excel at capturing local details while Vision Transformers (ViTs) are better at global context modeling. However, existing detectors typically rely on a single fixed backbone or a manually designed hybrid architecture, and thus fail to adaptively exploit these complementary strengths across inputs of diverse complexity. To address this limitation, we propose Backbone Module Composition via Reinforcement Learning (BMCR). BMCR dynamically assembles input-adaptive inference paths from reusable modules decomposed from off-the-shelf CNN and ViT backbones. To enable such cross-family composition, we first construct an extensible module toolbox. Specifically, we decompose representative CNN and ViT backbones into reusable functional modules and encapsulate each module with explicit structural, semantic, and computational metadata for compatibility-aware assembly. To bridge the gap between grid-based CNN features and token-based ViT representations, we design a lightweight Optimal Transport (OT) based transition interface that ensures distribution-aware alignment while respecting spatial consistency. The backbone composition process is then formulated as a sequential decision problem, in which a policy network progressively selects task-relevant modules according to intermediate multi-scale observations. To stabilize the joint optimization of reusable modules and the routing policy, we further develop an Adaptive Module Cooperative Optimization (AMCO) strategy that coordinates module updating, routing exploration, and reward assignment during training. On DOTA-v1.0, DOTA-v1.5 and DIOR-R, BMCR achieves 79.31\%, 73.41\% and 71.86\% mAP, respectively, surpassing strong static and dynamic baselines by up to 2.5 points while maintaining competitive efficiency.
arXiv:2606.05553v1 Announce Type: new
Abstract: Role-playing language agents (RPLAs) should play characters whose values and behavior evolve as the story progresses, not maintain a fixed persona. Existing benchmarks measure factual recall at a given chapter, not whether responses align with the character's psychological trajectory, especially in scenarios the source text never explores. We introduce ArcANE (Arc-Aware Narrative Evaluation), an automatically constructed benchmark spanning 17 novels and 80 principal characters. A Character Arc segments the narrative into phases along a psychological axis, and each probe poses the same scenario across phases, spanning both situations within the source text and situations beyond it. Across six models and six context modes, conditioning on the Character Arc tops every other context strategy on every model, and the gap is largest on scenarios outside the source text where retrieval has nothing to find. We further fine-tune open-weight models on the same data to obtain ArcANE-8B/32B, which widen the Arc advantage even more on scenarios outside the source text.
arXiv:2606.05747v1 Announce Type: new
Abstract: Cellular networks are undergoing a revolutionary transform with the advent of O-RAN architectures and AI/ML solutions. O-RAN's Non-Real-Time and Near-Real Time RAN Intelligent Controllers open the door to the implementation of automated control-loops that can provide RAN optimisations in numerous scenarios and use cases, and which can be further empowered by AI-driven approaches. Energetic sustainability has raised as one of the main optimisations targets due to the impact of mobile networks on global energy consumption. To this end, the BeGREEN project aims at enhancing the energy efficiency of beyond 5G networks by defining novel AI/ML-based methods at RAN and edge infrastructure. This paper presents BeGREEN Intelligent Plane, a novel framework which implements and exposes AI/ML workflows to O-RAN-based optimisations targeting energy efficiency. We also describe an exemplary application of the Intelligent Plane and its AI Engine, which aims at providing AI-driven cell on/off control.
arXiv:2606.05755v1 Announce Type: new
Abstract: Energy efficiency is a critical concern in the deployment and operation of 5G networks, particularly due to the low utilization of 4G and 5G carriers during off-peak hours. While considerable research has focused on designing energy-efficient cell on/off switching strategies that avoid disrupting user connectivity, the integration of operator-specific policies to guarantee particular Quality of Service (QoS) levels has received limited attention. This paper presents a machine learning (ML)-based energy saving strategy, trained using a real-world dataset from a European mobile operator, that enforces operator-defined policies that jointly consider strong throughput requirements and maximum outage tolerance constraints. By tuning the model's class ratios during training, the proposed solution enables operators to manage the trade-off between energy savings and QoS policy compliance prior to deployment in live networks. Evaluation results show that the method provides substantial energy savings while maintaining policy-compliant service levels under realistic 5G operating conditions.
arXiv:2606.05756v1 Announce Type: new
Abstract: Graph Neural Networks (GNNs) have demonstrated remarkable performance across a range of applications involving graph-structured data, particularly in high-stakes domains. However, the opaque nature of their decision-making processes limits their trustworthiness and broader adoption. Existing post-hoc explanation methods aim to improve explainability by identifying subgraphs that influence GNN predictions and adopt mixup strategies to alleviate the out-of-distribution (OOD) issue caused by using subgraphs for prediction. Yet, these approaches typically rely on soft masks, which are inherently unable to fully eliminate label-irrelevant information, allowing redundant structures to leak into the mixup process and hindering the resolution of the OOD problem, thereby degrading explanation fidelity. In this work, we propose HPME, a Hard-Perturbation Mixup Explanation framework grounded in a generalized Graph Information Bottleneck, which leverages graph pooling to extract discrete explanatory subgraphs and to yield an information-capacity bound to thoroughly compress label-irrelevant components. Furthermore, we introduce a novel mixup strategy built upon structure-level replacement, generating in-distribution explanations to effectively mitigate the distribution shift. Extensive experiments on diverse tasks demonstrate that HPME achieves state-of-the-art performance in generating robust and interpretable explanations across both synthetic and real-world datasets.
arXiv:2606.05760v1 Announce Type: new
Abstract: Deepfake videos are increasingly challenging the credibility of online content. Many existing detection methodology relies on complex, resource-intensive models, which limit their practical use. The study introduces the ExpSpeech-Net deepfake detection (SqN-R-DFD) model, which utilizes SqueezeNet and RNN (Recurrent Neural Network) as its backbone, providing a lightweight and efficient deepfake detection framework that simultaneously analyzes facial expressions and speech patterns. The approach incorporates advanced feature extraction, such as ISLBT-based features for image and MPNCC for signals, along with a smart feature-selection strategy using SASMA (Sandpiper-Assisted Slime Mould Algorithm), ensuring optimal and balanced input to the detection models. By combining SqueezeNet and an RNN, subtle inconsistencies in deepfake videos are captured effectively. The framework achieves 94.5% accuracy, precision of 99.3%, and F-measure of 96.8%, outperforming conventional methods. This demonstrates that integrating multiple modalities with intelligent preprocessing and feature selection enables practical, real-time deepfake detection suitable for everyday applications.
arXiv:2606.05787v1 Announce Type: new
Abstract: Protecting proprietary RAG databases from unauthorized redistribution is challenging: existing watermarking methods either inject fabricated relations between real entities, polluting the knowledge base with misinformation, or embed fragile lexical patterns that adversarial paraphrasing easily removes. We propose SentinelRAG, a watermarking framework that embeds style-consistent but fictitious knowledge entries into the RAG database. Our key insight is that synthetic knowledge describing fictitious entities is unlikely to be retrieved by legitimate queries, yet can be reliably triggered through targeted probes known only to the data owner. Experiments on four datasets ranging from 2.9k to 8.8M documents demonstrate that SentinelRAG achieves statistically significant detection $p < 10^{-5}$ across all tested configurations at only a 0.1% injection rate. Compared to the state-of-the-art, our method significantly reduces the false detection rate while maintaining negligible interference with legitimate user queries.
arXiv:2606.05585v1 Announce Type: new
Abstract: High tides are a threat to damage the coast and onshore structures. To investigate mitigation strategies, we simulate waves and a flood-like situation from two-dimensional (2D) dam-break flow with a ramp section at the end of the channel using smoothed particle hydrodynamics (SPH). We analyse the effects of ramps with various topographies to reduce the pressure on structures exerted by the wave. Structures of ramp surfaces influence flow behaviour significantly, absorbing kinetic energy of the wave. Increasing the ramp angle reduces the impact on the structure. A wave with a large velocity intensifies the flow impact, rendering the effects on all topography of the ramp almost insignificant. The ramp experiences the highest force exerted by the fluid on the bottom section. These insights enhance the understanding of ramp-induced energy dissipation and provide valuable implications for hydraulic engineering and structural resilience.
arXiv:2606.05814v1 Announce Type: new
Abstract: The support vector machine (SVM) is a widely used classifier, but choosing an appropriate loss function remains difficult. Convex losses such as the hinge loss and least-squares loss are sensitive to outliers, while bounded non-convex losses often lead to high computational cost. To address this, we propose a hybrid truncated loss function ($L_{\mathrm{ht}}$) that is both sparse and bounded, and build the $L_{\mathrm{ht}}$-SVM model for single-view classification. We introduce the P-stationary point and use it to establish the first-order necessary and sufficient optimality conditions. Based on these conditions, we design an alternating direction method of multipliers with a working-set strategy that reduces computational cost and achieves global convergence. We further extend $L_{\mathrm{ht}}$-SVM to multi-view learning by adding structural information and view weights, resulting in Mv$L_{\mathrm{ht}}$-SVM, which follows both the consensus and complementarity principles. Experiments on synthetic, real-world, and image datasets show that $L_{\mathrm{ht}}$-SVM achieves higher accuracy with fewer support vectors and better noise robustness than five single-view methods, while Mv$L_{\mathrm{ht}}$-SVM outperforms six multi-view methods in accuracy, precision, recall, and F1-score.
arXiv:2606.05817v1 Announce Type: new
Abstract: Consistency training encourages models to behave similarly across different contexts, and has shown promise for reducing misalignment. We broaden the scope of consistency training in two ways. First, we introduce two new internal consistency targets: MLP Consistency Training (MLPCT), which matches post-activation MLP states, and Attention Consistency Training (AttCT), which matches per-head attention distributions. Second, we apply consistency training to four additional safety threats: persona in-context learning attacks, adversarial frustration, prefill attacks, and conditional misalignment. Across several models and threat settings, we find that consistency training reduces misalignment well beyond the sycophancy and jailbreak settings studied in prior work. We also find cases of cross-threat generalization, where training against one failure mode improves robustness to another, and identify a shared residual-stream mechanism underlying ACT, MLPCT, and AttCT, while distinguishing BCT as mechanistically distinct. Our results suggest that consistency training is a flexible and extensible framework for alignment, capable of unifying defenses against a broader class of model pathologies.
arXiv:2606.05602v1 Announce Type: new
Abstract: AI assistants in human-AI collaboration often correct suboptimal human actions through behavioral feedback (e.g., alerts or steering-wheel nudges in assistive driving). Such interventions can mitigate immediate errors, but long-term improvement requires addressing the underlying misconceptions that cause repeated mistakes. We introduce SENSEI, a framework that infers user misconceptions from interaction behavior and provides targeted, minimal yet sufficient suggestions to correct them. Our approach departs from action- or trajectory-level interventions by operating over a structured knowledge representation to localize and correct the sources of erroneous behavior. Across three long-horizon tasks with diverse misconceptions and corresponding behaviors, SENSEI demonstrates zero-shot compositional generalization, disentangling multiple overlapping misconceptions despite training only on single-misconception cases. A user study further shows that our method identifies real human misconceptions and provides effective guidance that improves long-horizon task performance, successfully correcting $90\%$ of student misconceptions. Code and project page are available at https://misoshiruseijin.github.io/SENSEI/.
arXiv:2606.05605v1 Announce Type: new
Abstract: How does a system that merely predicts the world come to distinguish its own causal influence from everything else? We trace this transition in a minimal 192-dimensional GRU through 40 controlled experiments arranged as a developmental sequence, adding components one at a time and tracking whether the system can distinguish self-caused from world-caused changes.
The developmental path reveals four conditions that must be satisfied in strict order: (1) persistent state forming stable attractors, (2) a causal action loop linking output to input, (3) proprioceptive feedback that makes implicit causal knowledge explicit, and (4) asynchronous awakening - perceptual learning must consolidate before action learning begins. We propose agency gain (A = Err_world - Err_self), the predictive advantage of knowing one's own action, as a metric to track this process. The self-aware predictor consistently outperforms the self-blind predictor across periodic (sinusoidal) and chaotic (Lorenz) environments, and the metric survives ablation of all auxiliary components. Only forward-sampled action selection produces meaningful agency gain; two gradient-based alternatives degenerate.
Equally significant are 12 falsified hypotheses mapping where development stalls: predictive coding alone does not produce self-represent
arXiv:2606.05821v1 Announce Type: new
Abstract: Service discovery is essential in wireless communications. However, existing protocols provide limited privacy protection, leaking sensitive device information and opening routes to network attacks. This paper proposes a private service discovery protocol, called PriSrv, which enables both service providers and clients to specify fine-grained authentication policies before establishing connections. PriSrv achieves this via a dual-layer matching architecture: an outer layer filters mismatched entities using public attributes, while an inner layer handles mutual authentication using selectively disclosed private attributes. As a core component, we introduce the primitive of anonymous credential-based matchmaking encryption (ACME), which enables dual-layer matching in a single step to achieve bilateral policy control, selective attribute disclosure, and multi-show unlinkability. To instantiate ACME, we design a fast anonymous credential (FAC) scheme providing constant-size credentials and efficient verification. We demonstrate PriSrv's interoperability by integrating it with popular wireless frameworks including EAP, mDNS, BLE, and AirDrop. Detailed formal security proofs and extensive performance evaluations across desktop, laptop, smartphone, and Raspberry Pi platforms demonstrate that PriSrv provides enhanced privacy guarantees with high usability, achieving secure discovery in less than one second on mainstream mobile devices.
arXiv:2606.05847v1 Announce Type: new
Abstract: Text-guided molecular generation with LLMs often yields invalid SMILES. We argue that invalid drafts should be addressed through a shift from validity-oriented repair to identity-preserving molecular recovery: the objective is not only to restore chemical validity, but also to preserve target-relevant structural cues and recover the molecular identity implied by the description. This perspective reveals the limitations of existing correction strategies. Post-hoc repair can recover validity while distorting key structures, LLM-only correction can introduce unintended global drift, and generic agentic correction remains constrained by greedy single-candidate trajectories even when equipped with executable RDKit edit tools. To address these limitations, we propose AMREC, which couples molecule-aware mismatch tracking with expanded candidate exploration and trajectory-level selection. On invalid ChEBI-20 drafts from three backbone models, AMREC achieves the strongest overall recovery profile across structural, exact-match, and string-level metrics.
arXiv:2606.05607v1 Announce Type: new
Abstract: Si-doped $\beta$-Ga$_2$O$_3$ films were heteroepitaxially grown on sapphire substrates using HVPE. The influence of sapphire offcut on growth kinetics, surface morphology, crystalline quality, and electrical transport properties was systematically investigated. Growth kinetics studies revealed a strong dependence of deposition rate on HCl flow, growth pressure, and source-to-substrate distance, with growth rates reaching up to 30 $\mu$m/hr. Increasing sapphire offcut angle from 0$^\circ$ to 8$^\circ$ promoted a transition from multidirectional growth to highly aligned terrace-dominated surfaces, reducing the surface roughness from 14.69 to 2.74 nm. The improved surface morphology was accompanied by enhanced crystalline quality, with phase-pure (-201)-oriented $\beta$-Ga$_2$O$_3$ growth and a reduction in the rocking-curve full width at half maximum from 994 to 414 arcsec as the sapphire offcut increased. Electrical characterization of films grown on 6$^\circ$ offcut substrates yielded carrier concentrations ranging from $1.0\times10^{17}$ to $3.4\times10^{18}$ cm$^{-3}$. A maximum room-temperature electron mobility of 100cm$^2$/V$\cdot$s was achieved at a carrier concentration of $1.0\times10^{17}$cm$^{-3}$, representing the highest reported room-temperature mobility for HVPE-grown $\beta$-Ga$_2$O$_3$ on a foreign substrate. Analysis of the temperature-dependent transport characteristics yielded donor activation energies of 35 and 90 meV together with a low acceptor concentration of $3\times10^{15}$ cm$^{-3}$, consistent with the improved crystalline quality achieved on the offcut sapphire substrates. These results demonstrate that HVPE is capable of producing high-quality $\beta$-Ga$_2$O$_3$ heteroepitaxial layers with good crystalline quality and carrier transport characteristics, providing a promising pathway for scalable $\beta$-Ga$_2$O$_3$ epitaxy on low-cost foreign substrates.
arXiv:2606.05855v1 Announce Type: new
Abstract: Continuous electroencephalography (EEG) emotion prediction aims to model the temporal evolution of human emotional states from EEG signals. Unlike conventional discrete emotion recognition, continuous prediction requires capturing long-range temporal dependencies and coherent emotional dynamics. However, existing methods mainly rely on point-wise regression and directly model noisy high-dimensional EEG features, limiting their ability to characterize continuous emotional evolution.To address these challenges, we propose EEGDancer, a dynamic emotional latent space learning framework for continuous EEG emotion prediction. The framework integrates vector-quantized representation learning, masked temporal modeling, and reinforcement learning-based trajectory optimization into a unified architecture.Specifically, a causal spatiotemporal Vector-Quantization Variational Autoencoder (VQ-VAE) is designed to learn structured emotional prototypes and construct a discrete-continuous emotional latent space from EEG signals. Based on the learned latent representations, a Transformer-based masked dynamic modeling strategy captures long-range emotional dependencies and temporal evolution patterns. Furthermore, continuous emotion prediction is formulated as a sequential decision-making problem, and a Soft Actor-Critic (SAC) framework is introduced to optimize emotional prediction trajectories at the sequence level instead of frame-wise local fitting.Extensive experiments on the SEED, SEED-IV, and Long-Term Naturalistic Emotion datasets demonstrate that EEGDancer consistently outperforms existing machine learning and deep learning methods. Ablation studies further verify the effectiveness of the proposed latent space and reinforcement learning-based trajectory optimization for modeling continuous EEG emotional dynamics.
arXiv:2606.05874v1 Announce Type: new
Abstract: Current evaluations for Multimodal Large Language Models (MLLMs) overwhelmingly focus on utility-driven objectives, leaving model behavior under logic-neutral scenarios largely underexplored. Stochasticity is essential in scenarios where multiple actions are equally valid, such as recommending travel itineraries or daily schedules where multiple options have similar utility. In such settings, deterministic policies may lead to repetitive behaviors and reduced coverage of valid alternatives. To bridge this gap, we propose RandomBench, a benchmark designed to evaluate whether MLLMs can maintain distributionally neutral behavior when selecting among equivalent options. We further introduce three metrics, including RI, BCI, BII, to quantify entropy and distributional bias. Experiments reveal a pervasive phenomenon termed Stochastic Collapse, where MLLMs fail to maintain uniform randomness under explicit random instructions, with top-1 probabilities reaching 97% from the ideal one quarter baseline and RI dropping to 0.068 in Claude Sonnet 4.6. Extensive ablation studies further demonstrate that these deviations persist across languages and representation formats, highlighting the robustness of distributional collapse in logic-neutral decision settings.
arXiv:2606.05600v1 Announce Type: new
Abstract: This paper investigates a rotatable antenna (RA)-enabled uplink non-orthogonal multiple access (NOMA) system, where a base station equipped with multiple independently RAs serves both ground and aerial users. Specifically, we formulate an energy efficiency (EE) maximization problem by jointly optimizing receive beamforming, user power allocation, and RA rotation. To make the problem tractable, a new block coordinate descent-based algorithm is developed, in which the receive beamforming is updated via the minimum mean square error criterion, while the power allocation and RA rotation are handled by fractional programming and successive convex approximation. Numerical results demonstrate the EE superiority of the proposed RA-NOMA scheme over several benchmarks.
arXiv:2606.05877v1 Announce Type: new
Abstract: Mirror symmetry can strongly enhance the transmission of waves through a barrier inside a complex medium. We recently showed that this phenomenon enables quantitative through-barrier sensing: by tuning programmable scatterers on one side of the barrier to maximize the broadband total transmission through the barrier, the characteristics of scatterers at mirror-symmetric positions on the other side of the barrier can be determined. Considering a sufficiently large bandwidth was crucial to ensure that no accidental narrowband asymmetric resonance can outperform the symmetry-induced transmission enhancement. Here, we overcome this scheme's vexing need for a large bandwidth by replacing the underlying frequency diversity with configurational diversity. Specifically, we introduce auxiliary tunable scatterers at mirror-symmetric positions on either side of the barrier and sweep their characteristics through a series of random mirror-symmetric configurations. We tune the programmable main scatterers on one side of the barrier to maximize the average of the total through-barrier transmission over a series of configurations of the auxiliary scatterers at a single frequency, in order to sense the characteristics of the main scatterers on the other side of the barrier. We systematically study the accuracy of our single-frequency sensing scheme based on a multiport-network system model that cascades two mirror-related wave-chaotic cavities with a weakly transmitting barrier in between. We further examine an extension to non-reciprocal chaotic cavities involving circulators. Altogether, our results establish configurational diversity as a route to single-frequency, symmetry-empowered through-barrier sensing in reconfigurable complex media.
arXiv:2606.05880v1 Announce Type: new
Abstract: Agile humanoid locomotion across diverse challenging terrain demands both wide perceptual coverage and precise local geometry understanding. Motivated by the way humans selectively look at relevant terrain during locomotion, we introduce TAGA, a Terrain-aware Active Gaze learning framework for Attention-based humanoid control. By fusing vision, proprioception, and motion commands, our framework guides the model to learn anticipatory cues and actively attend to specific areas of the height scan, selectively using these informative regions for the downstream network. This adaptively increases the information density of observations under tight onboard computational constraints, thus enabling fine-grained perceptive locomotion over larger-scale terrains. We find that such gaze behaviors can naturally emerge through reinforcement learning alone, without requiring additional supervision or explicit guidance, significantly improve training efficiency. As a result, the trained policy demonstrates robust and generalizable locomotion in simulation and on hardware, including reliable terrain-aware foothold selection, elevated-platform traversal, competitive sparse-foothold traversal, and the largest reported real-world gap traversal distance of 1.2m among perceptive humanoid locomotion systems, while maintaining stability under severe perceptual disturbances and environmental interference.
arXiv:2606.05885v1 Announce Type: new
Abstract: Long-horizon LLM agents require reinforcement learning methods that can assign credit to intermediate decisions under sparse and delayed rewards. Recent group-based methods such as GiGPO improve over GRPO by constructing step-level advantages at repeated anchor states. However, we show that such dense credit can be statistically unreliable: under limited rollouts, rare but lucky actions may receive overly large advantages, producing divergent anchor bias and late-stage training oscillation. We propose Evidence-Calibrated Policy Optimization (ECPO), a critic-free policy optimization algorithm that calibrates step-level credit before policy updates. ECPO combines Evidence-Calibrated Action Advantage, which groups rollouts by canonical actions and shrinks low-count estimates, with Variance-Gated Credit Weighting, which suppresses anchor states dominated by within-action noise. Experiments on ALFWorld and WebShop with Qwen2.5-1.5B/7B show that ECPO consistently outperforms strong baselines, improving GiGPO by +5.2/+7.3 success points on ALFWorld/WebShop with Qwen2.5-1.5B while adding only 0.1% additional advantage-computation overhead.
arXiv:2606.05888v1 Announce Type: new
Abstract: Retry-based objectives such as pass@K and max@K optimize the best return obtained from multiple sampled trajectories, and recent work has shown that they can promote exploration without explicit exploration bonuses. In discrete action spaces, ReMax was shown to do so by adapting to return uncertainty. In this work, we introduce pathwise derivative estimators for retry objectives and use them to extend ReMax to continuous action spaces. We study the resulting learning dynamics and show that, even with deterministic rewards, ReMax can encourage stochastic exploration by reshaping the policy-gradient landscape. In particular, it alters gradients both in direction, biasing updates toward higher policy entropy, and in magnitude, damping gradients and slowing convergence. We further show that Adam's adaptive normalization can mitigate this damping, depending on its numerical stabilization parameter. Empirically, we instantiate this objective as ReMax Actor-Critic (ReMAC), an off-policy actor--critic algorithm that optimizes the ReMax objective using a pathwise derivative estimator. Our experiments show that ReMAC can promote higher policy entropy without entropy regularization and achieves performance comparable to SAC.
arXiv:2606.05889v1 Announce Type: new
Abstract: We propose GLASS, a framework for composable acoustic style control in zero-shot autoregressive text-to-speech (TTS) that learns controls from post-generation rewards rather than style labels. In zero-shot TTS, a speaker prompt often entangles speaker identity with prosodic attributes such as speaking rate and pitch, making it difficult to change style without changing the prompt itself. GLASS instead treats each acoustic attribute as a reward-defined control direction. For each control axis, GLASS freezes the TTS backbone and trains one lightweight LoRA adapter with Group Relative Policy Optimization (GRPO), using speech-token length and mean F0 as style rewards and WER as an intelligibility anchor. Because each control is represented as a LoRA weight update, independently trained adapters can be swapped, interpolated, and composed through linear LoRA arithmetic without retraining the backbone. Experiments on speaking rate and pitch control show targeted style shifts while preserving naturalness, speaker similarity, and intelligibility, and demonstrate smooth interpolation and multi-axis composition across independently trained adapters.