探索全球前沿学术脉络

01.

arXiv (quant-ph) 2026-06-11 DOI: arXiv:2501.02681

Quest for quantum advantage: Monte Carlo wave-function simulations of the Coherent Ising Machine

作者:

Manushan Thenabadu↗Run Yan Teh↗Jia Wang↗Simon Kiesewetter↗Margaret D Reid↗Peter D Drummond↗

arXiv:2501.02681v2 Announce Type: replace Abstract: The Coherent Ising Machine (CIM) is a quantum network of optical parametric oscillators (OPOs) intended to find ground states of the Ising model. This is an NP-hard problem, related to several important minimization problems, including the max-cut graph problem. In order to enhance its potential performance, we analyze the coherent coupling strategy for the CIM in a highly quantum regime. To explore this limit, without assuming gaussianity, we employ accurate numerical simulations. Due to the inherent complexity of the system, the maximum network size is limited. While master equation methods can be used, their scalability diminishes rapidly for larger systems. Instead, we use Monte Carlo wave-function methods, which scale as the wave-function dimension, and use large numbers of samples. These simulations involve Hilbert spaces exceeding $10^{7}$ dimensions. To evaluate success probabilities, we use quadrature probabilities. We demonstrate the potential for quantum computational advantage by reducing the time required to reach maximum success probability in a low-dissipation regime enabled by initial quantum superpositions and entanglement. Furthermore, we demonstrate that tailored time-dependent couplings can amplify these quantum effects. Comparisons with classical CIM models give evidence that quantum tunneling effects in this strong coupling limit can overcome trapping in false minima. This can greatly increase success rates, indicating a potential for quantum advantage. Finally, we perform a coherence analysis based on the state purity to examine the role of quantum coherence in CIM performance and to determine how state purity correlates with improved optimization outcomes.

阅读与讨论 → 访问原文 →

02.

arXiv (CS.CL) 2026-06-12 DOI: arXiv:2606.12876

Multi-Bitwidth Quantization for LLMs Using Additive Codebooks

作者:

Liza Babaoglu↗Shuangyi Chen↗Ashish Khisti↗

As large language models (LLMs) are increasingly deployed across heterogeneous hardware with varying resource constraints, the ability to adaptively manage the trade-off between performance and efficiency without retraining is critical. We propose Drop-by-Drop, a novel multi-bitwidth post-training quantization framework that enables inference-time precision control over LLM weights from a single trained model. Our method is theoretically grounded in information theory and successive refinement. We establish that LLM weights, which commonly follow a Gaussian distribution, can be optimally reconstructed with increasing fidelity as additional bits are incorporated, under a weighted mean squared error distortion motivated by LLM loss functions. To realize this in practice, Drop-by-Drop incorporates Matryoshka-style supervision into the loss function, exploiting the structure of additive codebooks. Drop-by-Drop produces a single model where ordered subsets of codebooks yield accurate partial reconstructions at each precision level. This approach significantly reduces storage and memory overhead by allowing a single checkpoint to serve multiple bitwidths, while maintaining competitive perplexity and accuracy across major architectures, such as Qwen, LLaMA, Gemma, and Mistral.

阅读与讨论 → 访问原文 →

03.

arXiv (CS.CV) 2026-06-16 DOI: arXiv:2606.15409

Segmentation-based Detection for Efficient Multi-Task Spacecraft Perception

作者:

Sivaperuman Muniyasamy↗Surendar Devasundaram↗

Vision-based perception is fundamental to Space Situational Awareness and autonomous on-orbit operations such as rendezvous, docking, servicing, and navigation. However, progress in this area is limited by the scarcity of annotated space imagery and by challenging visual-domain characteristics including severe illumination changes, low signal-to-noise ratio, and high contrast. We address Stream 1 of the SPARK 2026 Challenge, which requires a single model for spacecraft classification, detection, and fine-grained component segmentation across multiple target types. We propose a compact architecture that integrates a MobileNetV3 encoder with a U-Net-style decoder, combining computational efficiency with accurate dense prediction. Detection is derived analytically from the union of predicted component masks, avoiding a separate bounding-box regression head in the single-spacecraft setting. Our method achieved an overall leaderboard score of 0.9482, with task-specific scores of 1.0000 in classification, 0.9788 in detection, and 0.8917 in segmentation. The proposed approach ranked second overall in the SPARK 2026 Challenge, demonstrating that lightweight encoder-decoder architectures can deliver strong multi-task performance for practical onboard space vision systems.

阅读与讨论 → 访问原文 →

04.

arXiv (CS.CV) 2026-06-18 DOI: arXiv:2606.18943

Physics-IQ Verified

作者:

Tim R\"adsch↗Yuki M Asano↗Hilde Kuehne↗Stefan Bauer↗Priyank Jaini↗Robert Geirhos↗Carsten T. L\"uth↗

Video generative models ( VGMs) have become a new frontier that can be used not just for video generation but for a multitude of downstream tasks, including world modeling. To advance these tasks, a good video model must understand the physical reality of the world. Evaluating this understanding is an emerging field and has led to the Physics-IQ benchmark, which quantifies this explicitly by comparing model-generated videos to real-world videos of physical experiments. In this work, we present a systematic audit of the Physics-IQ benchmark, expose shortcomings and propose three solutions that sharpen how we can measure physical understanding of VGMs. Specifically, we improve prompt and ground-truth quality to reduce the influence of confounding factors and further introduce a sample-level scoring system that weights each sample and metric equally. Our resulting benchmark, Physics-IQ Verified, refines 57.6\% of all samples and improves over 34.8\% of prompts. In a comparison study using six image-to-video generative models, we observe moderate but meaningful ranking changes (Kendall's $\tau = 0.46$). We hope Physics-IQ Verified advances the community by providing a more reliable signal toward physically accurate VGMs. The code for the benchmark can be accessed at https://github.com/google-deepmind/physics-iq-benchmark

阅读与讨论 → 访问原文 →

05.

arXiv (CS.LG) 2026-06-15 DOI: arXiv:2605.24795

Lifted Schrödinger Bridges for Gaussian Mixture Endpoints: Projection Gaps and Path-Space Obstructions

作者:

Siddhartha Ganguly↗George Rapakoulias↗Panagiotis Tsiotras↗

arXiv:2605.24795v2 Announce Type: replace-cross Abstract: We study stochastic density control between Gaussian-mixture endpoint distributions under Brownian prior dynamics. Since the direct Schrödinger bridge between Gaussian mixtures is generally not available in closed form, we introduce a lifted path-space construction in which each trajectory is augmented with a source–target component label. Consequently, the problem decomposes into Gaussian component-to-component Schrödinger bridges with explicit marginal, drift, and cost formulas, while the mixture-level assignment reduces to a finite-dimensional entropic coupling problem with a Sinkhorn scaling form. We then analyze the projection obtained by discarding or forgetting the label. By construction, the projected law satisfies the original Gaussian-mixture endpoint constraints, but its relative entropy generally differs from the lifted relative entropy by a nonnegative conditional label-information gap. This gap reveals a path-space obstruction: the lifted optimizer cannot, in general, be identified with the direct unlabeled Schrödinger bridge after projection. We also derive the posterior-averaged Markov drift associated with the projected marginal flow, prove a kinetic-energy upper bound, and identify a common path-potential condition under which the projection gap vanishes. Several numerical illustrations showing density and shape control are recorded for a self-contained exposition.

阅读与讨论 → 访问原文 →

06.

arXiv (CS.CV) 2026-06-15 DOI: arXiv:2606.13886

PhysVLA: Towards Physically-Grounded VLA for Embodied Robotic Manipulation

作者:

Namai Chandra↗Shriram Damodaran↗Lin Wang↗

Vision-Language-Action (VLA) models excel at mapping visual inputs and natural language instructions directly to robotic control policies. However, because they are trained primarily to fit behavioural demonstration data, they do not explicitly enforce fundamental physical principles such as rigid-body dynamics or contact constraints. This exposes a critical physics gap: standard temporal smoothing applied on top of single-step or chunked VLAs trades trajectory quality for added failures that short-term memory cannot resolve. To bridge this gap, we introduce PhysVLA (Physics-VLA), a plug-and-play, inference-time framework designed to wrap any frozen VLA backbone without retraining, fine-tuning, or weight access, with less than 1 ms of overhead per control step. PhysVLA intercepts the predicted control action, captures only the simulator or system state, and applies a dual-layered correction: (i) a phase-aware finite-state machine that structures discrete task segments (approach, grasp, transport, and place), and (ii) a selective Euler-Lagrange gate that activates only when a dynamics oracle detects kinodynamic inconsistency. Evaluated across OpenVLA, OpenVLA-OFT, Force-VLA, and Generalist-VLA on LIBERO-Spatial with a 7-DoF Franka Panda, the framework delivers absolute success rate increases of up to 17% and stability increases of up to 19% with no per-task regressions, improves trajectory efficiency by up to 15% across all four backbones, and shows up to a 10x improvement in trajectory jerk robustness on a Robosuite Lift cross-simulator sweep. We further validate the framework on a real Agilex Piper arm with a pick-and-place task, confirming that PhysVLA transfers to physical hardware without retraining, with success-rate improvements of up to 50%, establishing physical awareness as a composable, backbone-agnostic runtime module.

阅读与讨论 → 访问原文 →

07.

arXiv (CS.CL) 2026-06-12 DOI: arXiv:2606.13663

HyperTool: Beyond Step-Wise Tool Calls for Tool-Augmented Agents

作者:

Yaxin Du↗Yifan Zhou↗Yujie Ge↗Jiajun Wang↗Xianghe Pang↗Shuo Tang↗Tuney Zheng↗Bryan Dai↗Jian Yang↗Siheng Chen↗

Tool-augmented LLM agents commonly rely on step-wise atomic tool calls, where each invocation, observation, and value transfer is exposed in the main reasoning trace. This creates an execution-granularity mismatch: locally deterministic tool workflows are unfolded into repeated model-visible decisions, consuming context and forcing the model to manage low-level dataflow in the trace. We introduce HyperTool, a unified executable MCP-style tool interface that changes the model-visible unit of tool execution. A model invokes HyperTool with a code block that can call existing tools through their original schemas, manipulate returned values, and pass intermediate results locally, folding deterministic tool subroutines into a single outer call. To train models to use this interface, we synthesize HyperTool-format trajectories from cross-tool compositional tasks and verify them in real MCP environments. On MCP-Universe, HyperTool improves average accuracy from 15.69\% to 35.29\% on Qwen3-32B and from 9.93\% to 33.33\% on Qwen3-8B, and surpass GPT-OSS and Kimi-k2.5 on average accuracy, showing that our HyperTool can substantially improve multi-step tool use.

阅读与讨论 → 访问原文 →

08.

arXiv (math.PR) 2026-06-19 DOI: arXiv:2606.19583

Power-law hypothesis and (un)fairness of PageRank on undirected multi-type PAMs

作者:

Christian Borgs↗Florian Henning↗Remco van der Hofstad↗Nelly Litvak↗

arXiv:2606.19583v1 Announce Type: new Abstract: The preferential attachment model (PAM) describes the sequential growth of a network based on the "rich-get-richer" principle. Several versions of it have become established for modeling, e.g., citation networks, capturing a power-law degree distribution. Directed versions of the preferential attachment model where the edges are directed from the new to the old vertices have been the subject of extensive research. They have been shown to exhibit remarkable properties such as heavier tails for the limiting graph-normalized PageRank than for the in-degrees. By contrast, for the undirected version, we recently showed that PageRank has similar tails as the degree. In the present paper, we discuss the PageRank asymptotics for a multi-type version of the undirected PAM (here vertices have different colors), complementing previous results of Antunes, Bhamidi, Banerjee and Pipiras on the asymptotics of PageRank on similar directed multi-type or colored PAMs. Our studies are motivated by the aim to go beyond the rigid rule of edge orientation in directed preferential attachment models. As the main result, for the case of a finite set of colors, we show that the power-law hypothesis for PageRank is fulfilled also for the colored undirected PAM, where, by contrast to the directed case, the power-law exponent is color-dependent for some choices of the initial color distribution and the attractiveness function. For the specific case of a two-type model, we discuss implications of our results on fairness in sampling underrepresented nodes from the network.

阅读与讨论 → 访问原文 →

09.

arXiv (CS.LG) 2026-06-11 DOI: arXiv:2512.08211

MobileFineTuner: A Mobile-Native Framework for On-Device LLM Fine-Tuning in Real-World Embedded AI Applications

作者:

Jiaxiang Geng↗Lunyu Zhao↗Yiyi Lu↗Bing Luo↗

arXiv:2512.08211v2 Announce Type: replace Abstract: Large language models (LLMs) are moving from cloud-centric services toward on-device embedded AI, where models interact with private, longitudinal signals sensed from users and their physical environments. Mobile phones are a natural platform for such applications because they are continuously carried by users, connected to wearable sensors, and deeply integrated with daily mobile applications. However, practical LLM fine-tuning on commodity phones remains difficult. Existing fine-tuning frameworks are largely Python-based and server-oriented, making them hard to deploy inside mobile applications. We present MobileFineTuner, a mobile-native open-source framework for end-to-end LLM fine-tuning on commodity mobile phones. MobileFineTuner is implemented in C++ and provides a reusable training stack. To make fine-tuning feasible under mobile resource constraints, MobileFineTuner integrates a resource-aware training runtime with memory-efficient attention, activation checkpointing, gradient accumulation, parameter sharding, and energy-aware scheduling. We evaluate MobileFineTuner on real mobile phones using GPT-2, Gemma 3, and Qwen2.5 models across multiple fine-tuning tasks. The results show that MobileFineTuner reproduces standard Full-FT and LoRA fine-tuning behavior, substantially reduces memory pressure and improves executability on memory-constrained phones. We further demonstrate MobileFineTuner through a private campus health-agent application, where a local LLM is fine-tuned on user-specific wearable-sensing records to provide more personalized responses while keeping raw records on the phone. These results establish MobileFineTuner as a practical toolkit for studying and building on-device LLM fine-tuning applications in embedded AI and sensing systems.

阅读与讨论 → 访问原文 →

10.

arXiv (CS.CL) 2026-06-19 DOI: arXiv:2606.19667

CacheWeaver: Cache-Aware Evidence Ordering for Efficient Grounded RAG Inference

作者:

Kaizhen Tan↗Rong Gu↗Mingyuan Li↗

Retrieval-Augmented Generation (RAG) improves factual grounding, but it also lengthens prompts and raises prefill cost. Prefix caching in serving engines such as vLLM reduces this cost only when requests share the same token prefix. In grounded generation, however, adjacent queries may retrieve overlapping evidence in different orders, so set overlap does not become reusable prefix overlap. We present CacheWeaver, a lightweight prompt-layer method for cache-aware evidence ordering. The method keeps a prefix tree over recently served evidence sequences and uses a greedy walk to place the most reusable prefix first, while leaving the serving engine and retrieved evidence set unchanged. Across three vLLM configurations, the method lowers median time-to-first-token (TTFT) by about 20-33 percent relative to retrieval-order prefix caching, without hurting answer quality in our QA tests. The greedy policy reaches 97.5 percent of the median TTFT gain from oracle ordering, indicating that most reusable prefix locality can be recovered by a simple scheduling layer between retrieval and inference.

阅读与讨论 → 访问原文 →

11.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2606.07678

DOG-DPO:Dynamic Optimization in Geometry for Safety Alignment

作者:

Yi Nian↗Tiankai Yang↗Yudi Zhang↗Qi Pan↗Zelong Xu↗Shenzhe Zhu↗Qingqing Luan↗Yue Huang↗Xiangliang Zhang↗Yue Zhao↗

arXiv:2606.07678v2 Announce Type: replace-cross Abstract: Safety alignment for large language models relies on preference data, but current pipelines often train on large, redundant datasets. Existing data selection methods typically score each preference pair independently, collapsing directional preference information into scalar quality or diversity scores. This sample-centric view is especially limiting in multi-dataset settings, where shared safety directions coexist with dataset-specific residual risks. We propose DOG-DPO, a training-free data selection framework that treats preference pairs as structured geometric signals. DOG-DPO first represents each preference pair as a direction in model representation space. It then decomposes multi-dataset preference geometry into a global anchor subspace and dataset-specific residual subspaces. Finally, it selects subsets by maximizing diversity-based coverage, encouraging broad, non-redundant coverage of alignment directions before DPO training. Across six safety benchmarks and two model backbones, DOG-DPO achieves a strong utility-robustness trade-off using only 11% of the preference pairs. It recovers most of the safety gains of full-data training while remaining entirely teacher-free, training-free, and substantially faster than representative selection baselines.

阅读与讨论 → 访问原文 →

12.

arXiv (quant-ph) 2026-06-11 DOI: arXiv:2511.02613

Emergent Bell Phase in an Electro-Nanomechanical Quantum Simulator

作者:

David Ullrich↗Marta Cagetti↗Stefan Forstner↗Adrian Bachtold↗Anna Sanpera↗

arXiv:2511.02613v2 Announce Type: replace Abstract: Suspended carbon nanotubes hosting electrostatically defined quantum dots allow for exceptionally strong and tunable electromechanical coupling as well as mechanical modes that can reach the quantum ground state of motion simply by cryogenic cooling. This makes them a unique platform for quantum simulation of electron-phonon coupling. Here, we propose an experimentally realisable setup with two such carbon nanotubes in parallel, each hosting four quantum dots. Our system not only exhibits phonon-mediated electron-electron attraction, but also supports a robust, maximally entangled Bell phase at mesoscopic scales shared across the subsystems. These features highlight its potential as a simulator of strongly correlated quantum systems.

阅读与讨论 → 访问原文 →

13.

arXiv (CS.CL) 2026-06-19 DOI: arXiv:2606.19744

Beyond Uniform Forgetting: A Study of Sequential Direct Preference Optimization Across Preference Settings

作者:

Pranav Bhandari↗Nicolas Fay↗Amitava Datta↗Usman Naseem↗Mehwish Nasim↗

Aligning language models with human preferences often requires optimising multiple behavioural objectives. A practical approach is to apply these objectives sequentially using preference optimisation methods such as Direct Preference Optimisation (DPO), but it remains unclear whether later training uniformly degrades preferences learned earlier or whether the effect depends on the relationship between objectives. We study sequential DPO across four preference settings covering distributional conflict, multi-attribute interaction, strong safety signal, and compatible response-quality objectives. Using Llama-3.1-8B-Instruct with LoRA adapters, we evaluate all objectives after every stage with a fixed base-model reference. We find that sequential DPO does not produce a single forgetting pattern; preference change ranges from partial degradation to stability, pair-level redistribution, or positive transfer depending on objective relationship, signal strength, and training order. Pair-level analysis using length-normalised policy margins shows that aggregate metrics can mask heterogeneous changes across preference pairs, whereas quartile decomposition reveals that high-confidence pairs can either degrade or improve depending on the setting. Mechanistic diagnostics show that Stage~2 gradients and adapter updates are near-orthogonal to the previous objective across all settings, providing little evidence that direct gradient opposition is the primary driver. These findings suggest that future sequential alignment pipelines should account for objective compatibility and signal strength, rather than assuming that later objectives affect earlier preferences uniformly.

阅读与讨论 → 访问原文 →

14.

arXiv (CS.AI) 2026-06-17 DOI: arXiv:2606.17409

Discrete Autoregressive Transformer for Generative Mechanism Synthesis

作者:

Anar Nurizada↗Anurag Purwar↗

arXiv:2606.17409v1 Announce Type: cross Abstract: Planar path synthesis requires mechanisms whose coupler curves match a prescribed trajectory; the mapping from curve to linkage is inherently one-to-many across four-, six-, and eight-bar topologies. We address this design problem with simulation-grounded evaluation on a curated corpus of over one million mechanisms, reporting Chamfer distance and dynamic time warping after forward kinematics and geometric alignment. We formulate synthesis as conditional autoregressive sequence modeling: joint coordinates are uniformly quantized to tokens and generated by a decoder-only transformer with a variational-autoencoder (VAE) latent of the target curve and an explicit mechanism-type token. Training combines token cross-entropy with a Gaussian-smoothed bin auxiliary loss that respects ordinal structure among bins. At inference, a bounded latent-noise schedule decodes all mechanism types at each noise level; we retain the top five candidates by geometric error, yielding diverse accurate families without dataset lookup. On held-out tests, aggregate mean Chamfer distance is $0.0132$ and mean dynamic time warping is $0.153$; a latent $k$-nearest-neighbor baseline that conditions on training-set neighbor latents in VAE space achieves matched-topology mean Chamfer distance $0.0071$ and mean dynamic time warping $0.117$ using the same decoder.

阅读与讨论 → 访问原文 →

15.

arXiv (CS.AI) 2026-06-18 DOI: arXiv:2606.18617

AI-Driven Assessment of Human Tutors: Linking Training Performance to Real-Life Practice

作者:

Danielle R. Thomas↗Marie Cynthia Abijuru Kamikazi↗Clara Brandt↗Conrad Borchers↗Kenneth R. Koedinger↗

arXiv:2606.18617v1 Announce Type: cross Abstract: There exist numerous tutor training platforms. However, few provide AI-driven training and evaluation for human tutors based on real-life performance. We present an AI-driven system that assesses both open responses during training and authentic real-life tutoring. Unlike platforms that only assess learning through online training or simulations, our system utilizes Generative AI (Gemini-2.5-pro) to analyze transcriptions of authentic tutoring, measuring the transfer of tutor skills to real-life application. Human tutors instructing students remotely in math (N=86) completed six scenario-based lessons, averaging a significant 7.4% learning gain. Using mixed-effects models across 405 session-to-lesson pairs, we found that training performance significantly predicted real-life transcript scores with an effect size of 0.25 SD. Model comparison (AIC/BIC) indicated averaging open response and multiple choice performance during training predicted real-life tutor performance best, although open responses were comparatively more predictive. Exploratory analysis showed that after training, tutors were significantly more likely to encounter pedagogical opportunities to apply their skills (61.1% to 68.9%) and demonstrated higher execution quality within those opportunities (65.5% to 68.1%). Interrupted time series analysis suggested that these tutor improvements were part of a gradual trend over time rather than an immediate intervention effect of training. We illustrate an AI-driven method to link tutor training with real-life assessment. In doing so, we contribute open datasets, AI prompts, and scoring rubrics to support transparency and reproducibility.

阅读与讨论 → 访问原文 →

16.

Nature (Science) 2026-06-17 DOI: HASH:8f0fbd8d715b37a6817e31f5ab9e9f4e

Structure of the pre-initiation complex explains CMGE biogenesis

作者:

Thomas Pühringer↗

When cells enter S phase, bidirectional DNA replication is initiated through the kinase-regulated recruitment of three activators (Cdc45, GINS and Pol ε) to a duplex-DNA-loaded double hexamer of minichromosome maintenance (MCM) ATPases. Together, these proteins form two CMGE helicases that establish divergent replication forks as they become separated1. Here, to gain an understanding of CMGE biogenesis, we reconstituted the pre-initiation complex with purified yeast proteins. The cryo-electron-microscopy structure shows a set of firing factors caught in the act of assembling two symmetrical CMGEs. We show how stepwise complex formation reshapes MCM in preparation for DNA opening, and we explain how ATP promotes firing-factor ejection and CMGE maturation. We find that although Sld2 facilitates the recruitment of GINS to MCM, as expected, it also aids the efficient separation of the CMGE dimer, and is essential for the ejection of the lagging strand from MCM. These findings have direct implications for our understanding of the metazoan Sld2 orthologue, RECQL4, and point to a replication-fork establishment mechanism that is conserved across eukaryotes. Cryo-electron microscopy and biochemical reconstitution experiments in yeast provide insight into the assembly of the CMGE complex, a helicase that establishes bidirectional DNA replication in eukaryotic cells, and elucidate the role of the firing factor Sld2.

阅读与讨论 → 访问原文 →

17.

arXiv (CS.LG) 2026-06-17 DOI: arXiv:2603.04198

Stable and Steerable Sparse Autoencoders with Weight Regularization

作者:

Piotr Jedryszek↗Oliver M. Crook↗

arXiv:2603.04198v2 Announce Type: replace-cross Abstract: Sparse autoencoders (SAEs) are widely used to extract human-interpretable features from neural network activations, but their learned features can vary substantially across random seeds and training choices. To improve stability, we studied weight regularization by adding L1 or L2 penalties on encoder and decoder weights, and evaluate how regularization interacts with common SAE training defaults. On MNIST, we observe that L2 weight regularization produces a core of highly aligned features and, when combined with tied initialization and unit-norm decoder constraints, it dramatically increases cross-seed feature consistency. For TopK SAEs trained on language model activations (Pythia-70M-deduped), adding a small L2 weight penalty increased the fraction of features shared across three random seeds and roughly doubles steering success rates, while leaving the mean of automated interpretability scores essentially unchanged. Finally, in the regularized setting, activation steering success becomes better predicted by auto-interpretability scores, suggesting that regularization can align text-based feature explanations with functional controllability.

阅读与讨论 → 访问原文 →

18.

arXiv (CS.AI) 2026-06-11 DOI: arXiv:2604.13733

Vision-Language-Action Jump-Starting for Reinforcement Learning Robotic Agents

作者:

Angelo Moroncelli↗Roberto Zanetti↗Marco Maccarini↗Loris Roveda↗

arXiv:2604.13733v2 Announce Type: replace-cross Abstract: Reinforcement learning (RL) enables high-frequency, closed-loop control for robotic manipulation, but scaling to long-horizon tasks with sparse or imperfect rewards remains difficult due to inefficient exploration and poor credit assignment. Vision-Language-Action (VLA) models leverage large-scale multimodal pretraining to provide generalist, task-level reasoning, but current limitations hinder their direct use in fast and precise manipulation. In this paper, we propose Vision-Language-Action Jump-Starting (VLAJS), a method that bridges sparse VLA guidance with on-policy RL to improve exploration and learning efficiency. VLAJS treats VLAs as transient sources of high-level action suggestions that bias early exploration and improve credit assignment, while preserving the high-frequency, state-based control of RL. Our approach augments Proximal Policy Optimization (PPO) with a directional action-consistency regularization that softly aligns the RL agent's actions with VLA guidance during early training, without enforcing strict imitation, requiring demonstrations, or relying on continuous teacher queries. VLA guidance is applied sparsely and annealed over time, allowing the agent to adapt online and ultimately surpass the guiding policy. We evaluate VLAJS on six challenging manipulation tasks: lifting, pick-and-place, peg reorientation, peg insertion, poking, and pushing in simulation, and validate a subset on a real Franka Panda robot. VLAJS consistently outperforms PPO and distillation-style baselines in sample efficiency, reducing required environment interactions by over 50% in several tasks. Real-world experiments demonstrate zero-shot sim-to-real transfer and robust execution under clutter, object variation, and external perturbations.

阅读与讨论 → 访问原文 →

19.

arXiv (CS.CV) 2026-06-18 DOI: arXiv:2604.03156

CAMEO: A Conditional and Quality-Aware Multi-Agent Image Editing Orchestrator

作者:

Yuhan Pu↗Hao Zheng↗Ziqian Mo↗Zirui Pang↗Hill Zhang↗Tianyi Fan↗Shuhong Wu↗Jiaheng Wei↗

Conditional image editing aims to modify a source image according to textual prompts and optional reference guidance. Such editing is crucial in scenarios requiring strict structural control (i.e., anomaly insertion in driving scenes and complex human pose transformation). Despite recent advances in large-scale editing models (i.e., Seedream, Nano Banana, etc), most approaches rely on single-step generation. This paradigm often lacks explicit quality control, may introduce excessive deviation from the original image, and frequently produces structural artifacts or environment-inconsistent modifications, typically requiring manual prompt tuning to achieve acceptable results. We propose CAMEO, a structured multi-agent framework that reformulates conditional editing as a quality-aware, feedback-driven process rather than a one-shot generation task. CAMEO decomposes editing into coordinated stages of planning, structured prompting, hypothesis generation, and adaptive reference grounding, where external guidance is invoked only when task complexity requires it. To overcome the lack of intrinsic quality control in existing methods, evaluation is embedded directly within the editing loop. Intermediate results are iteratively refined through structured feedback, forming a closed-loop process that progressively corrects structural and contextual inconsistencies. We evaluate CAMEO on anomaly insertion and human pose switching tasks. Across multiple strong editing backbones and independent evaluation models, CAMEO consistently achieves 20\% more win rate on average compared to multiple state-of-the-art models, demonstrating improved robustness, controllability, and structural reliability in conditional image editing.

阅读与讨论 → 访问原文 →

20.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2502.12445

Computational Safety for Generative AI: A Hypothesis Testing Perspective

作者:

Pin-Yu Chen↗

arXiv:2502.12445v2 Announce Type: replace Abstract: AI safety is a rapidly growing area of research that seeks to prevent the harm and misuse of frontier AI technology, particularly with respect to generative AI (GenAI) tools that are capable of creating realistic and high-quality content through text prompts. Examples of such tools include large language models (LLMs) and text-to-image (T2I) diffusion models. As the performance of various leading GenAI models approaches saturation due to similar training data sources and neural network architecture designs, the development of reliable safety guardrails has become a key differentiator for responsibility and sustainability. This paper presents a formalization of the concept of computational safety, which is a mathematical framework that enables the quantitative assessment, formulation, and study of safety challenges in GenAI through the lens of signal processing theory and methods. In particular, we explore two exemplary categories of computational safety challenges in GenAI that can be formulated as hypothesis testing problems. For the safety of model input, we show how sensitivity analysis and loss landscape analysis can be used to detect malicious prompts with jailbreak attempts. For the safety of model output, we elucidate how statistical signal processing can be used to detect AI-generated content. Finally, we discuss key open research challenges, opportunities, and the essential role of signal processing in computational AI safety.

阅读与讨论 → 访问原文 →

21.

arXiv (CS.CL) 2026-06-16 DOI: arXiv:2606.15980

Do Safety Monitors Stay Reliable After an Update? Benchmarking and Predicting Activation-Monitor Staleness

作者:

Evan Duan↗

Activation monitors-lightweight probes trained on a language model's internal representations-are an increasingly common layer in deployment safety stacks. Deployed models however are rarely static: they are quantized, fine-tuned, adapted with LoRA, or served with merged adapters while the monitor remains frozen. We present the first systematic test of whether this implicit contract holds: whether activation monitors trained on a base model remain reliable after these routine model updates. Across multiple safety-relevant monitors, model depths, update families, and open-weight models, we find a sharp split: quantization-style updates largely preserve frozen probe performance, while fine-tuning-style updates frequently make probes stale. Fragility is highly monitor-dependent, with privacy/PII probes most affected and refusal-compliance probes comparatively stable, showing that retraining a behavior need not stale its corresponding monitor. QLoRA is especially damaging despite NF4 quantization alone being relatively benign, suggesting that quantization becomes riskier when combined with adaptation. We further show that degradation is predictable from pre-deployment features, enabling revalidation budgets to be triaged toward the monitors most likely to fail. These results suggest that fine-tuning should trigger activation-monitor revalidation by default, while prediction can help prioritize which monitors to check first.

阅读与讨论 → 访问原文 →

22.

arXiv (CS.CL) 2026-06-16 DOI: arXiv:2606.16240

Creative Collision: Directorial Persona Steering and Competition in Large Language Models

作者:

Subramanyam Sahoo↗Justin Shenk↗

Activation steering has emerged as a powerful tool for shaping the behaviour of large language models at inference time, yet most prior work injects a single semantic direction into the residual stream. We study the richer setting in which two semantically opposing steering vectors are superimposed – a regime we call Creative Collision. Concretely, we construct directorial persona vectors for Steven Spielberg (optimistic, redemptive moral valence) and Martin Scorsese (dark, morally ambiguous) via mean-difference activation contrast on curated screenplay-derived corpora, then interpolate between them with a scalar mixing parameter $\alpha \in [0,1]$ and a steering coefficient $\lambda$. Across five evaluation axes – moral valence, generation coherence, surface style, directional dominance, and vector geometry – three principal findings emerge: (i)~Spielberg's representational signature exhibits robust directional dominance, suppressing Scorsese's moral influence across almost the entire interpolation range; (ii)~intermediate collision points paradoxically improve generation coherence relative to pure single-director steering at high $\lambda$; and (iii)~both personas localise maximally to layer~28 of a 40-layer decoder-only transformer, revealing a shared moral-tone substrate. These results illuminate the geometry of competing semantic directions in transformer residual streams and have direct implications for controllable creative generation and value-aligned narrative synthesis.

阅读与讨论 → 访问原文 →

23.

arXiv (CS.LG) 2026-06-16 DOI: arXiv:2504.11775

Discrimination-free Insurance Pricing with Privatized Sensitive Attributes

作者:

Tianhe Zhang↗Suhan Liu↗Peng Shi↗

arXiv:2504.11775v3 Announce Type: replace-cross Abstract: Fairness has become an important concern in insurance pricing as insurers increasingly rely on machine learning models to predict expected losses. At the same time, regulatory and privacy constraints often restrict insurers' ability to access or use sensitive attributes such as gender or race. Recent actuarial research addresses fairness in this context through the concept of the discrimination-free premium, which removes both the direct and indirect effects of sensitive attributes while preserving actuarial consistency. However, implementing this approach typically requires access to the sensitive attributes themselves, which may not be available in practice. This paper studies the estimation of discrimination-free insurance premiums when sensitive attributes are observed only in privatized or noise-perturbed form. We consider a multi-party data setting in which insurers observe non-sensitive attributes and outcomes, while a trusted third party holds privatized sensitive attributes generated through a privacy mechanism. Within this framework, we develop statistical methods for estimating discrimination-free premiums using only the privatized attributes. We study two settings of practical relevance: when the privacy mechanism is known and when its noise level is unknown. For both cases, we establish theoretical guarantees for the proposed estimators. Numerical experiments and empirical applications demonstrate that the proposed approach enables fair insurance pricing while respecting privacy and regulatory constraints.

阅读与讨论 → 访问原文 →

24.

arXiv (CS.LG) 2026-06-19 DOI: arXiv:2606.19481

Insulin4RL: Real-Time Insulin Management in the Intensive Care Unit for Offline Reinforcement Learning

作者:

Thomas Frost↗Steve Harris↗

arXiv:2606.19481v1 Announce Type: new Abstract: Offline reinforcement learning (ORL) offers the potential to improve the quality of clinical decision-making using historical electronic health record (EHR) data. Current training and evaluative practices in this field rely heavily on EHR datasets that have been temporally discretised into fixed, regular time intervals. Discretisation creates fictional representations of complex clinical scenarios and compromises the generalisability of retrospective model evaluations. In this paper, we introduce Insulin4RL, a healthcare ORL dataset featuring naturally irregular inputs and actions from real clinical trajectories. Derived from MIMIC-IV, Insulin4RL comprises over 375,000 labelled decisions across 12,209 patients requiring insulin infusion titration in the Intensive Care Unit. The dataset can thus be used for research into ORL model performance under realistic clinical sampling assumptions. We provide a description of the dataset's structure and characteristics, baseline performance metrics using model-free offline reinforcement learning, and a standardised evaluation protocol using fitted Q-evaluation. We conclude with suggested areas for future research that could be addressed using this resource.

阅读与讨论 → 访问原文 →

25.

arXiv (CS.LG) 2026-06-19 DOI: arXiv:2510.06048

BLISS: A Lightweight Bilevel Influence Scoring Method for Data Selection in Language Model Pretraining

作者:

Jie Hao↗Rui Yu↗Wei Zhang↗Huixia Wang↗Jie Xu↗Mingrui Liu↗

arXiv:2510.06048v5 Announce Type: replace Abstract: Effective data selection is essential for pretraining large language models (LLMs), enhancing efficiency and improving generalization to downstream tasks. However, existing approaches often require leveraging external pretrained models, making it difficult to disentangle the effects of data selection from those of the external pretrained models. In addition, they often overlook the long-term impact of selected data if the model is trained to convergence, primarily due to the prohibitive cost of full-scale LLM pretraining. In this paper, we introduce BLISS (BileveL Influence Scoring method for data Selection): a lightweight data selection method that operates entirely from scratch, without relying on any external pretrained oracle models, while explicitly accounting for the long-term impact of selected data. BLISS leverages a small proxy model as a surrogate for the LLM and employs a score model to estimate the long-term influence of training samples if the proxy model is trained to convergence. We formulate data selection as a bilevel optimization problem, where the upper-level objective optimizes the score model to assign importance weights to training samples, ensuring that minimizing the lower-level objective (i.e., training the proxy model over the weighted training loss until convergence) leads to best validation performance. Once optimized, the trained score model predicts influence scores for the dataset, enabling efficient selection of high-quality samples for LLM pretraining. We validate BLISS by pretraining 410M/1B/2.8B Pythia and LLaMA-0.5B models on selected subsets of the C4 dataset. Notably, under the 1B model setting, BLISS achieves $1.7\times$ speedup in reaching the same performance as the state-of-the-art method, demonstrating superior performance across multiple downstream tasks.

阅读与讨论 → 访问原文 →