Estimating the ground energy of interacting fermions is central to quantum many-body physics, and recent quantum algorithms provide polynomial-time guarantees for some structured families. Classical approaches commonly rely on quantum Monte Carlo, where the sign problem is a major obstacle. However, even when a nonnegative expansion without sign problem is available, rigorous guarantees for rapid mixing are still far from well understood. In this paper, we give an efficient classical randomized algorithm for approximating the ground energy of fermionic Hamiltonians with conjugate hopping matrices and conjugate-paired orbital interactions, which includes the attractive Hubbard models. This guarantee not only holds without a physical spectral gap or a weak-coupling assumption, but also covers interactions beyond onsite density attractions. Technically, we first reduce ground-energy estimation to partition-function estimation and then use a discretization with controlled energy error to obtain a nonnegative weighted sum of squared determinant moduli and then apply the Chen--Liu log-concave counting theorem. In addition, we obtain a sharper arithmetic bound by controlling how many independent sampling runs and matrix updates are needed, and using sparse matrices to reduce the cost of each update.
Agentic workloads have emerged as a major workload for LLM inference. They differ significantly from chat-only workloads, requiring long-context processing, the ability to handle multimodal inputs, and structured multi-turn interactions with tool calling capabilities. As a result, their context exhibits structure that can carry different importance along three key axes: temporal recency to the current turn, modality such as text or image tokens, and semantic role such as user queries, tool calls, observations, or reasoning. These axes capture distinct token behaviors and lead to different sensitivities to KV-cache compression. However, existing KV-cache quantization methods are typically homogeneous or exploit only heterogeneity on a single dimension, such as temporal proximity or modality, overlooking the interactions among them. To this end, we introduce TriAxialKV, a novel mixed-precision KV-cache quantization scheme that assigns each token a triaxial tag, calibrates per-tag sensitivity, and allocates INT2/INT4 bit-widths under a fixed memory budget. We implement TriAxialKV as an end-to-end serving system, comprising calibration, mixed-precision quantization and memory management, and custom fused Triton decode kernels. When using Qwen3-VL-32B-Thinking as a computer-use agent operating the OSWorld, TriAxialKV matches the accuracy of SGLang with BF16 KV cache while supporting 4.5$\times$ KV cache capacity and achieving 30% higher end-to-end throughput, when running on real GPU systems.
Egocentric video captioning requires fine-grained detail and accurate timing, yet current multimodal large language models (MLLMs) often miss brief actions, confuse their order, or misinterpret the wearer's motion. Existing caption formats also offer limited support for combining detailed narration with precise timing. We present EgoCaptioner, which pairs a video-level overview of the setting, participants, and the wearer's overall activity with timed steps whose narration adapts to varying perceptual demands and marks key moments at specific timestamps. We develop a coarse-to-fine annotation pipeline that establishes event structure before expanding step descriptions, yielding 50,000 training examples from eight egocentric video sources. Complementary rewards assess factual accuracy and coverage, measure step-level temporal alignment, and penalize fragmented or redundant narration, guiding caption refinement through Group Relative Policy Optimization (GRPO). To improve the reliability of factual supervision, we supplement caption references with pose-derived motion facts when sensor data are available. Across five egocentric benchmarks, EgoCaptioner achieves leading performance among evaluated open-source models in caption quality, caption-conditioned question answering, and temporal alignment. It also surpasses Gemini 3.1 Pro on PhysCapBench and all evaluated proprietary baselines on Minerva-Ego under the caption-to-QA protocol. Ablations support the complementary roles of factual and temporal rewards and show that coarse-to-fine annotation provides more effective supervision than single-pass annotation.
We investigate the functional differentiation learned by RoPE Transformers to derive principles for autoregressive language modeling. We introduce two intervention-based metrics: RoPE Frequency Importance Score (RFIS) measures how removing a frequency contribution changes a head's full attention distribution, and RoPE Positional Dependence (RPD) isolates the effect of rotary modulation. In Qwen3 and Llama3.1, these metrics distinguish retrieval from positional modeling, while individual heads can exhibit both functions. A mid-low-frequency Global Positional Band (GPBand) separates their frequency dependencies and tracks the training sequence length in controlled experiments, suggesting that global positional fitting can hinder zero-shot length extrapolation. These findings motivate two principles. (1) Functional Organization: retrieval and positional functions should be assigned separately and mixed within layers with layer-specific allocation; and (2) Positional Dependence: positional modeling should operate only locally, while global access should rely on position-independent retrieval. Matched pretraining from scratch and targeted architectural ablations at 380M and 1.4B parameters support both principles and highlight the importance of mechanism choice in realizing each function. Together, these findings suggest a promising route toward architectural innovation grounded in the functional requirements of autoregressive language modeling.
Once a status update enters service, age-optimal policies must wait for it, even when heavy-tailed delays inflate the Age of Information (AoI). We develop a framework in which the sender may preempt the update in service and sample a fresh one, under general service-time distributions and explicit action costs. Our framework covers sources that generate updates at will as well as random update arrivals. Under nondecreasing age costs, we prove that, with preemption, a finite mean delay suffices for a finite average age even under infinite delay variance. We show the optimality of constant service-age thresholds from prior work for exponential service, and exhibit numerically that for general delays the optimal preemption rule depends on both the AoI and the elapsed service time. Under an action-cost budget, we further minimize the Value at Risk (VaR), which bounds the AoI except for a prescribed fraction of time, and the Conditional Value at Risk (CVaR), the mean AoI over that fraction. We reduce both to the same optimality equations without enlarging the state space. Under Pareto and log-normal delays, our algorithm TAILOR keeps the average cost bounded as delay variance grows, while optimal non-preemptive sampling diverges. With optimal preemption, a more variable network can allow fresher data than a steadier one of equal mean delay.
Vision-language models (VLMs) demonstrate broad generalization across tasks and domains. However, comparable generalization in robot manipulation remains challenging. Heterogeneity in robot embodiments, task formulations, geometric conventions, and action interfaces fragments interaction experience across benchmarks and hardware. We rethink robot skill learning through alignment: Can heterogeneous interactions across benchmarks and embodiments jointly improve one shared skillset for robot manipulation? We propose UniCraft, a framework for learning robot manipulation skills by aligning heterogeneous settings. Specifically, UniCraft represents observed scene geometry using consistent coordinate conventions and aligns sensing and control interfaces across robots. It organizes tasks from different benchmarks in a canonical format, allowing interaction feedback to jointly refine shared skills and the orchestration policy that guides their selection and composition, without any demonstrations or model weight updates. We interleave tasks from 3 benchmarks and 6 embodiments, collecting 2,700 learning rollouts per agent model. During training, UniCraft improves pooled task success on unseen scenarios and tasks by 9-28 percentage points on LIBERO-PRO, 10.8-13.3 on RoboTwin 2.0, and 3.1-14.4 on RoboCasa365 (Epochs 1-3). After training, with the learned skills frozen, UniCraft demonstrates zero-shot transfer across 4 new robot embodiments and 5 additional agent models on robosuite and RoboDojo.
Random coloured digraphs defined by a Markov logic network
Yasmin Tousinejad, Vera Koponen
AI 速读 基于摘要
研究马尔可夫逻辑网络定义的随机有向图概率随域大小变化的极限行为,并给出判定所属情形的有效方法。
math.LO · cs.AI
查看原文摘要
A Markov Logic Network (MLN) $M$ is a probabilistic relational model used in Statistical Relational Artificial Intelligence for defining a probability distribution $P_D^M$ on the set $W_D$ of possible worlds with domain $D$ for an arbitrary finite domain $D$. An MLN consists of first-order formulas, which we call soft constraints, with associated weights which are nonnegative real numbers. The behaviour of the probability distribution, including the effect of the weights, on different domain sizes is poorly understood. We approach the problem in a systematic way by studying sequences $(P_n^M(ψ) : n = 1, 2, 3, \ldots)$ where $ψ$ is a first-order sentence and $P_n^M(ψ)$ is the probability that $ψ$ holds in $W_D$ where $D = [n] := \{1, \ldots, n\}$. We consider a first-order language that can express a property $P(x)$ and relation $R(x, y)$ and show that all of the following cases can be realized for an MLN $M$ such that each of its soft constraints is a boolean combination of $P(x)$, $P(y)$ and $R(x, y)$ in which both $x$ and $y$ appears: For every first-order sentence $ψ$, $P_n^M(ψ)$ converges to 0 or 1 as $n\to\infty$. For every first-order sentence $ψ$, $P_n^M(ψ)$ converges as $n\to\infty$ but not necessarily to 0 or 1. There is a first-order sentence $ψ$ such that $P_n^M(ψ)$ does not converge as $n\to\infty$. In fact, all three cases can be realized even after fixing the soft constraints and, for any $\varepsilon > 0$, by varying their weights by less than $\varepsilon$. We also show that, given any MLN as described above, we can efficiently determine which case it belongs to by inspecting the MLN. Then we show that if the weights are scaled by $1/n$ then we end up in case~(1) no matter how the weights are chosen.
Identifying active molecules for known targets and potential targets for existing drugs is central to drug discovery and repurposing. Drug-target interaction (DTI) models can rank vast numbers of drug-target pairs, but limited experimental capacity restricts testing to a small number of candidates. We compared the bidirectional recommendations of 13 released systems across 711 molecular structures of FDA-approved drugs and 614 human proteins. For the same target or drug, pairs of models produced completely disjoint Top-10 molecule and protein lists in 77.8% and 70.1% of comparisons, respectively. Even where recommendations overlapped, agreement between some systems largely reflected recurring preferences for the same candidates. Published inhibition measurements for 79 molecules and 65 kinases revealed that some systems enriched active candidates, yet their hit yields changed little when recommendations were matched to different drugs or targets, whereas other systems lost substantially more hits. Some of the former systems also rarely ranked both molecules and kinases correctly when two drugs favoured different kinases. We further retrained the same architecture on the same data, with the training objective fixed and random seeds varied. Despite similar test performance, mean pairwise Top-10 overlap on a separate screening catalogue was about two candidates. Our study moves DTI model evaluation beyond benchmark performance to assess whether model predictions can guide meaningful experimental decisions in drug discovery and target identification.
Fast Self-Supervised Generative CT Denoising via Few-Step Sampling
Xingyi Yao, Zhengtao Ma, Zhennong Chen
AI 速读 基于摘要
N2N-iMF 通过少步采样加速自监督生成式 CT 去噪,在保持图像质量的同时大幅缩短生成时间。
cs.CV
查看原文摘要
Low-dose and thin-slice computed tomography (CT) images suffer from elevated noise that obscures anatomical detail, while clean references for supervised denoising are often unavailable. Although Noise2Noise permits self-supervised training from noisy pairs, direct regression tends to oversmooth fine structure. Averaging multiple independently generated samples mitigates this blur, though repeating multi-step diffusion sampling is costly. We introduce Noise2Noise with Improved Mean Flow (N2N-iMF) to accelerate this process via few-step sampling. N2N-iMF learns to generate one noisy image conditioned on its paired counterpart; at inference, independent outputs are sampled and averaged to suppress noise. We evaluated image quality and sampling efficiency on simulated thin-slice brain and low-dose abdominal CT across four sampling methods and averaging counts from 1 to 20, and validated the approach on real-world photon-counting CT (PCCT) brain scans. With ten samples per conditioning input and three network evaluations per sample, N2N-iMF achieved a mean absolute error of 2.60 HU on brain CT and 14.02 HU on abdominal CT, outperforming 50-evaluation denoising diffusion implicit model (DDIM) sampling (2.91 and 14.53 HU). On the same GPU, per-output generation time dropped from 7.29 to 0.43 s. Across eight PCCT cases, N2N-iMF reached a mean gray-white matter contrast-to-noise ratio of 1.341, versus 0.916 for Noise2Noise and 1.176 for DDIM. N2N-iMF accelerates self-supervised generative CT denoising without distillation while keeping the number of averaged outputs adjustable at inference.
In the global attention layers of Gemma 4 31B, the value is the normalized key projection, and the key is the value scaled by a fixed weight, with rotary position encoding on one quarter of its dimensions. The Gemma 4 report states that this design reduces the global KV cache by up to 37.5%, but the open serving engines that we checked still store a full key and a full value: 1,024 values per KV head and token. We store one 640-value row instead, exact in real arithmetic, and an unchanged paged attention kernel reads it through two overlapping strided views as its key cache and its value cache. In a vLLM prototype with the 16-bit 31B model, the outputs are bit-identical to an emulation of the same arithmetic on the stock cache, 2 of 432 final answers differ from the stock engine (within its run-to-run noise), and the steady decode KV memory per request falls by a factor of 1.34 to 1.40 at 32K to 64K tokens. On a prefill/decode pair of two A100 GPUs with NIXL transfer and CPU offload, a decode engine with a full 15 GiB pool ran 1.38 times the stock decode tokens per step for 16K-token prompts (repeated to 0.06%) and 1.24 times for 8K-token prompts; a model of the vLLM block pool predicted both ratios to within 1% (the engine was 0.7% to 0.9% higher) and matched the change between them. The eager decode throughput rose by 1.30 and 1.11 times, and on this path the compact engine gave the same output tokens as the emulation for 8 prompts at batch 1. The run-time prototype makes an eager step 4% to 9% slower at equal batch, and the method does not yet stack with an 8-bit KV cache at the default scale.
Contactless Estimation of Respiratory Frequency Using 3D-CNN on Thermal Images
Federica Gioia, Filippo Pura, Alberto Greco, Dario Piga, Arcangelo Merla, Marco Forgione
AI 速读 基于摘要
作者提出一种直接在原始热视频上运行的3D卷积神经网络,无需复杂预处理和ROI跟踪即可估计呼吸频率。
cs.CV
查看原文摘要
Monitoring physiological parameters such as respiratory rate ($f_R$) is essential for diagnosing and managing various pathological conditions. Thermal imaging offers a promising contactless alternative to traditional methods, which often rely on partially invasive sensors or obtrusive wearable systems. However, existing approaches for $f_R$ estimation from thermal signals typically require extensive pre-processing and manual or semi-automatic region-of-interest (ROI) tracking, limiting their practical applicability. This study proposes a deep learning-based method for estimating $f_R$ directly from thermal videos, eliminating the need for complex pre-processing and ROI tracking. A 3D Convolutional Neural Network (3D-CNN) is developed to operate on raw thermal video data. To address challenges related to small datasets, the model is trained using data augmentation and transfer learning from synthetic datasets. Experimental results demonstrate that the proposed approach achieves a validation $R^2$ score of approximately 0.61 on both pre-processed and raw thermal videos. By simplifying the workflow, this method holds promise for enhancing the feasibility of thermal imaging in real-world applications, such as remote healthcare and driver monitoring in automotive applications.
World models are increasingly used to simulate physical interactions, but training a separate model for each environment limits scalability. Cross-environment prediction requires identifying the current physical rule, which appearance alone may not reveal. We therefore study visual calibration: using object motion and interactions in a short video as direct evidence to infer the current rule and predict motion under that rule from a different initial state. However, differences in data availability across physical conditions can yield a long-tailed training distribution, biasing models toward common rules even when evidence supports rare ones. To study this challenge, we introduce TailWorld, which evaluates the transfer of demonstrated rules across initial states while independently controlling training imbalance and calibration informativeness. The benchmark compares predictions with two reference futures from the same prediction initial state, one following the demonstrated rule and the other a common rule, to measure prediction bias toward common dynamics. We further propose Rule Inference for Transfer (RIFT), which learns how different rules affect object motion, infers the current rule from calibration interactions, and applies it to prediction from a new initial state. RIFT explicitly controls the influence of training frequency on rule inference, allowing sufficiently informative observational evidence to support rare rules. Experiments on TailWorld show that RIFT improves prediction accuracy and physical consistency across varying training imbalances and distribution shifts, with gains in predicting rare dynamics.
This paper tackles visible challenges in deep ensemble learning, where deep neural networks serve as ensemble members: training and storage burdens, and robustness of cautious (set-valued) predictions targeting multiple utilities, which may involve reward-sensitivity. To mitigate the training and storage burdens, we propose to employ compact ensembles, such as Bayesian Neural Networks and Convolutional Neural Networks with the Monte-Carlo dropout prediction option, to produce probabilistic predictions. For each query instance, these probabilistic predictions are then used to define a representative distribution optimizing some statistical distance. The representative distribution is then employed to define the Bayes-optimal prediction (BOP) of any utility. To address the potential unrobustness of singleton prediction making, we propose a family of set-utilities satisfying some desirable properties and whose set-valued BOPs can be found efficiently. Empirical evidence is then given to illustrate the potential (dis)advantages of the proposed ensemble learning framework.
Attention in large language models (LLMs) scores every query against every key with a full dot product. We investigate whether this dot product must be computed completely for attention to be effective. We propose Hidden Sparse Attention (HiSA), which keeps the query and key projections intact but only computes a partial dot product, following a static, parameter-free rule set by the token positions: attention becomes sparse in the hidden (feature) dimension, an axis much less explored than token sparsity, rather than over tokens as in almost all sparse attention methods. Based on the structure of rotary position embeddings (RoPE), we decompose keys and queries into parts which can be omitted when computing attention. By training LLMs from scratch with HiSA, we find that this partial dot product does not hurt but improves performance compared with the complete product while computing only about a third of the query-key products. During inference HiSA improves decoding speed and shrinks memory footprint. We also extend HiSA to FlashAttention-3 kernels for training and inference. Finally, we motivate our approach by showing that dense attention learns to suppress the attention score parts that HiSA drops. Code is available at https://github.com/ambroise-b/hisa.
In the field of precision manufacturing in complex constrained environments, the role of soft robots is increasingly prominent, and the realization of anti-winding control based on multi-intelligent body reinforcement learning has become a research hotspot. One of the core problems at present is to coordinate multiple robots to complete the unwinding operation in a highly constrained environment. The existing distributed training framework faces some observability challenges in high-density barrier and unstable environments, resulting in poor learning results. This paper proposes a topology-driven Multi-Agent Reinforcement Learning (TD-MARL) framework to coordinate multi-robot systems to avoid entanglement. Specifically, the critical network adopts centralized learning, so that each intelligent body can perceive the strategies of other intelligent bodies by sharing the topological state, thus alleviating the training instability caused by complex interactions; eliminating the demand for communication resources between robots through distributed execution, Upgrade system reliability; the integrated topological security layer uses topological invariants to accurately assess and mitigate the risk of entanglement to avoid the strategy from falling into local difficulties. Finally, the full simulation experiments carried out in the real simulation environment show that the method is better than the current advanced deep reinforcement learning (DRL) method in terms of convergence and anti-winding effect.
LLM agents on long-horizon tasks fail in recurring, procedural ways. They mis-paginate an API, resolve the wrong entity, or answer when no answer was asked. We built a pipeline that lets an agent learn from its own history. It first mines short \emph{lessons} from the agent's own trajectories, one to three per episode. It then consolidates those lessons into a smaller set of \emph{guidelines}, where each guideline records how many lessons it merged. At inference the agent receives either all guidelines or a selected subset. The pipeline ships in the open-source altk-evolve library and is deployed in IBM's CUGA agent and watsonx Orchestrate platform. We report what we learned building, measuring, and productizing it. Our primary instrument is the AppWorld benchmark, across models spanning 40--92\% baseline task completion. The central finding is that the right amount of memory is a property of the model, not the memory. A strong model keeps improving as guidelines are added. DeepSeek-V3.2 gains +9.5 points of task completion and +16.1 points of scenario completion with its full 191-guideline set. A weak model gains less from the same full set than from a small selected subset. gpt-oss-120b starts at 39.9\% task completion. It reaches 51.2\% with all 199 of its guidelines but 55.4\% (mean of two runs) with a selected subset of about 29. Around that result we report what went wrong: a high-support guideline that backfired, provenance silently dropped during consolidation, a swallowed API error that failed every task while producing a plausible log, and a baseline confound that nearly produced a misleading comparison with a prior method. We distill these into recommendations for teams deploying agent memory.
Continuous diffusion models generate discrete data through stochastic dynamics in an embedding space. We consider jump diffusions that additionally reset coordinates to embeddings of categorical values. A coordinate's own corrupted state can make prediction easy, potentially reducing the pressure to learn dependencies between coordinates. We study leave-one-out (LOO) prediction, in which each coordinate is predicted from the corrupted states of the other coordinates. We show that, under stated assumptions, jump diffusions with complete coordinate resets and state-independent clocks preserve the prescribed marginals when reset destinations use exact LOO probabilities and the drift uses the full endpoint posterior. We further show that this drift can be constructed from LOO probabilities by analytic likelihood completion without additional network evaluations. We evaluate unrestricted, LOO, and completed predictions on Text8 and LM1B text generation and ZINC250k molecular graph generation. LOO-based variants improve key generation metrics over matched unrestricted controls on all three datasets. Controlled interventions assess reliance on local inputs and context, and we analyze the effect of analytic likelihood completion.
Beam is a sparse Mixture-of-Experts (MoE) model with 501.3B total parameters and 23.4B active parameters, specifically built and optimized for coding, reasoning, and agentic workloads. Beam delivers frontier level reasoning, while requiring 3-4x less inference compute than many models of its class. Pretrained on 23.8T diverse, curated, high-quality tokens, Beam features a 1M-token context window achieved through a novel two-stage midtraining phase, built on verifiable, correctness-anchored data. Post-training leveraged a large-scale, fully asynchronous reinforcement learning (RL) campaign across 10.5K GB300 GPUs, generating over 100M rollouts over 4 weeks of training. We believe this is one of the largest scale RL runs conducted by any open lab to date. Across our evaluation suite, capabilities continued to improve as we increased RL compute, with no sign of a plateau. Supported by deep hardware-model co-design, FP8/NVFP4 quantization, and a 1.93B-parameter speculative decoding drafter (DFlash 2), Beam maximizes serving throughput with efficient, flexible deployment. Averaged across conversation turns and harm categories, Beam refuses or pivots on 88.9% of harmful requests while complying with 81.5% of harmless requests, achieving the highest refusal/pivot rate on harmful requests among open-weight models on the Pareto frontier.
Insider threats pose security risks when employees or other trusted users misuse their access to organizational systems. Detecting these threats is difficult because harmful activity can resemble routine behavior. Honeypots, or decoy resources monitored for suspicious interactions, can provide evidence of potentially malicious activity. We develop a decision framework that models evidence accumulation through honeypot interactions using a continuous-time Markov chain and jointly selects honeypot counts and monitoring horizons to minimize expected losses from malicious activity and false detections. For thresholds of two and three interactions, we derive closed-form true- and false-detection probabilities and expected detection times conditional on detection within the monitoring horizon. We characterize boundary cases and establish stochastic-ordering results. Our numerical results highlight the value of the proposed framework across different settings.
We study translational packing of convex polyhedra in three dimensions. Given $n$ convex polyhedra in $\mathbb{R}^3$, we seek a placement with pairwise disjoint interiors that minimizes the volume of an axis-parallel bounding box. Constant-factor approximation algorithms are known when rotations are allowed, but no approximation guarantee was previously known for arbitrary convex polyhedra with fixed orientations. A central difficulty is that, unlike in the plane, the total volume and the coordinate extents of the objects do not provide a useful lower bound on the optimum: even objects with small total volume and small diameter may require a container of large volume. We overcome this difficulty using a better geometric lower bound based on a surprising connection to Wang and Zahl's celebrated recent proof of the three-dimensional Kakeya conjecture. Using this, we obtain a polynomial-time $n^{o(1)}$-approximation algorithm. A recent strengthening of Wang and Zahl's theorem claimed by OpenAI gives a polylogarithmic approximation ratio for the same algorithm. Our ideas extend to bin and strip packing: for any fixed $δ>0$, we obtain the same approximation guarantees for translational packing of convex polyhedra whose coordinate widths are at most a $1-δ$ fraction of the corresponding container widths.
Irregular multivariate time series forecasting supports healthcare monitoring, environmental analysis, and activity recognition. Channel-independent (CI) approaches capture each variable's temporal patterns but do not explicitly use observations from other variables. Channel-dependent (CD) modeling exploits relationships among variables, yet asynchronous measurements make these relationships difficult to learn. Incorporating cross-variable information while preserving each variable's irregular history therefore remains a key challenge. We propose Temporal-Reference Complementary Forecasting (TRCF), a framework that augments a compact CI forecaster with a complementary CD branch. ANCHOR aggregates observed values and timestamps at shared temporal locations, retaining observation support to form compact CI features. BRIDGE constructs complementary features through history completion. It first estimates each variable's historical values from its own observations, then uses differences between other variables' observed and estimated values at nearby times to correct missing entries. The completed history preserves observed values and is encoded to supply context alongside the direct CI representation. Finally, the two representations are fused, combined with the query-time embedding, and passed to a shallow multilayer perceptron for forecasting. Experiments on four real-world datasets show that TRCF achieves the lowest mean squared error and mean absolute error in the main benchmark comparison, while reducing measured peak inference memory by 17.28% to 87.98% relative to APN.
PUPPET: Physical-Unified Platform for Presence, Embodiment, and Teleoperation
Mahika Vats, Milan Groshev, Carlos Gomez Cubero, Eduardo Castelló Ferrer
AI 速读 基于摘要
PUPPET 通过持续同步的 VR 副本和试验验证流程,提升远程操作任务表现与实验记录可复用性。
cs.RO
查看原文摘要
Remote access to robotics hardware does not by itself guarantee usable experimental data. A teleoperated trial can complete the task and still produce an incomplete or inconsistent record. PUPPET aims to solve this problem by presenting a cyber-physical framework that opens a physical robotics laboratory through a continuously synchronized VR replica of the robot, stranding environment, and task objects. In addition, PUPPET integrates a trial validation pipeline built into the experimental workflow. We evaluated PUPPET in a within-subject study with 30 participants, who performed the same table-to-shelf object-transfer task in person, through 2D teleoperation, and through PUPPET. VR reduced post-learning completion time by 27.7% over 2D (68.7,s vs. 95.0,s), statistically indistinguishable from in-person execution (72.1,s). It also produced more validation-ready (85.6% vs. 66.0%) and intervention-free (89.7% vs. 67.2%) trials. With PUPPET we show that modality shapes the remote task performance and the reusability of the resulting experimental record.
Perfumer: A Large Language Model-based Agentic Framework Accelerating Structural Relaxation and Electronic Band Prediction for First-Principle Calculation
First-principles structural relaxation and band calculations require substantial computation and expertise. Here we present Perfumer, a large language model-based agentic framework that couples Quantum ESPRESSO with two predictive models. Across tested structures, pre-relaxation with Perfumer-relax model followed by QE refinement yielded a 2.75-fold overall speedup over QE alone. Perfumer-band model combines a periodic graph neural network with a latent Hermitian spectral head returns 12 band energies at any requested wavevector, preserving reciprocal lattice periodicity and time-reversal symmetry without constructing Hamiltonian. Trained on 125,153 self-curated band structures, Perfumer-band model reaches a median per-material mean absolute error of 0.157 eV on 2,607 test materials, including lanthanide, actinide and magnetic compounds, from held-out composition families spanning 85 elements, taking about 97 ms per material on two consumer CPU cores. The agent turns natural-language requests into relaxation-to-band workflows, from a single uploaded structure to automatically constructed new candidates and returns predicted properties.
For 4D reconstruction, a synchronized multi-camera rig records all views at the same timestamps and misses the motion between them. Staggered triggers spread the same frames over more timestamps. However, when the cameras along each row fire one after another, as in 4DSloMo's schedule on the N3DV rig, the row acts as one moving camera that confounds depth with motion. We introduce OffBeat, which models a staggered rig as a spatially distributed multicoset sampler: shifted copies of one sampling grid. From the calibration alone, OffBeat derives camera groups that fire together and their firing order. Since one group records each timestamp, a short-lived primitive can fit an error its cameras share, such as a color offset, and flicker in novel views. We show that a primitive's opacity over time weights all groups equally, for every number of groups, only if it lasts at least one camera frame interval. On Spacetime Gaussians, a minimum lifetime approximates this bound, and a closed-form smoothed box of that length satisfies it exactly, with less flicker than a Gaussian in its place at nearly equal PSNR. In simulated staggered capture on N3DV, OffBeat outperforms the full 4DSloMo pipeline with video diffusion on every test frame, by 3.04 dB in mean dynamic-region PSNR. Between the first and last synchronized frame, it outperforms the best synchronized method of each scene, with or without frame interpolation, by 1.98 dB in mean dynamic-region PSNR, while synchronized capture is better at its own frames. On PanopticSports, in this frame range, OffBeat is comparable in PSNR to synchronized capture with interpolation. On every N3DV scene, replacing 4DSloMo's schedule with ours raises the dynamic-region PSNR of Spacetime Gaussians, 4DSloMo without diffusion, and our reconstruction. Multi-camera rigs could follow faster motion at the same frame count by changing only when each camera fires.
Can AI systems be moral agents? That is, can they undertake actions that can be correctly assessed as morally right or wrong? There is no settled normative criterion of moral agency that applies neatly outside paradigm human cases. Because of the difficulties with a normative account, we instead propose a research program to explore descriptive moral agency. Most basically, we ask, what are the capabilities we humans possess that underlie our moral agency? To do so, we draw on historical scholarship in philosophy and moral psychology to isolate candidate capabilities. Then we suggest refining candidate capabilities so that they are testable in the context of AI systems. Finally, we offer specific empirical recommendations to test AI systems, especially large language models (LLMs), for those refined capabilities. Our approach offers several benefits. Firstly, it may demonstrate whether the behaviors of AI systems are aligned, or in a broad sense desirable. Secondly, the debate may inform policy questions (e.g., what models should be deployed in which contexts?). Thirdly, our approach may complement engineering and design efforts: when capabilities can be precisely characterized, they can inform the development of models. Fourthly, applying our moral reasoning concepts in the new context of AI systems should improve our understanding of those concepts.
What does a recursive definition mean in a grounded logic, where a sentence is true only if a terminating process makes it so? This paper takes Kripke's answer for truth, the least fixed point, as a rule. Reflective Grounded Arithmetic (RGA) writes any recursive definition as one self-referential term; Inductive Grounded Arithmetic (IGA) adds the elimination rule for every positive one, making each an inductive definition. IGA is omega-sound and proof-grounded: sound and complete for a semantics whose certificates are its own proofs. IGA proves Goodstein's theorem, the Kirby-Paris hydra game for primitive recursive strategies, and, along Gentzen's route, the consistency of Peano Arithmetic. IGA certifies the Veblen notation below every $\varphi(n,0)$: its accessibility ordinal there is $\varphi(ω,0)$, that of the paracomplete truth theory PKF. Every theorem is machine-checked in Isabelle/HOL; order types and PKF's ordinal are cited. IGA proves the Veblen notation well-founded exactly when it admits bivalence of accessibility for it, which it does not, and that bivalence is sound exactly when the order is well-founded. So $Γ_0$ lies exactly one instance of excluded middle away, and that instance, the well-foundedness sentence and one membership claim in an internal form of Kleene's $\mathcal{O}$ are interderivable over IGA. Bivalence at the Bachmann-Howard order keeps proof-groundedness. One such claim buys well-foundedness of the countable part of Buchholz's notation at every finite level, of type $ψ_0(Ω_ω)$, and $k$ stratified bivalence rules give a proof-grounded system proving the countable part up to level $k$ well-founded; $k$ iterated closings-off of truth do too, sound only for a closed-off model. Unrestricted conditional proof is inconsistent (Curry's paradox); the purchase is implication at a hypothesis not provably decided.
DeepQuest 是一个以知识缺口驱动的 AI 科学家框架,用于阿尔茨海默病研究,将领域综述中的未解问题转化为可追溯、可检验的假设,并通过数据可行性筛选和分组 Elo 排序进行优先级排序,结合迭代实验与两级审计来评估科学发现率。
cs.AI
查看原文摘要
Automating experiments does not necessarily advance scientific understanding. Research agents must identify meaningful knowledge gaps, determine what available data can resolve, and assess whether their findings advance existing knowledge. We introduce DeepQuest, a knowledge-gap-driven AI scientist framework for Alzheimer's disease research. DeepQuest transforms unresolved questions from domain reviews into traceable, testable hypotheses and prioritizes them through data feasibility screening and groupwise Elo ranking. It couples iterative experimentation with two-level auditing that separates experimental reliability from knowledge contribution relative to two frozen literature corpora. On OASIS-3, DeepQuest produces 42 feasible hypotheses and outperforms the strongest baselines in 14 of 15 evaluator-dimension comparisons across three language-model reviewers. The full framework achieves an audit-assessed scientific finding rate of 95.24%, outperforming both general-purpose executors and domain-specific research agents. This scientific finding rate remains stable across literature corpora and evaluator model families. Representative findings concern neuropsychiatric symptoms and dementia risk, longitudinal amyloid spatial stability, and neocortical tau-associated clinical impairment, with key associations supported in independent ADNI and AIBL cohorts. These results show how aligning knowledge gaps, data feasibility, and evidence auditing can turn automated research into traceable, evidence-supported scientific discovery. Our source code is available at https://anonymous.4open.science/r/DeepQuest-3066/
Large language models increasingly rely on extended reasoning traces, yet the training signals used to shape this behavior remain coarse. Supervised fine-tuning on full trajectories forces the model to imitate an entire demonstration, overconstraining its output distribution, while trajectory-level reinforcement learning provides only weak credit assignment, rewarding or penalizing whole completions without identifying where they succeed or fail. We propose edit-based training, a general framework for learning from a model's own trajectories via localized textual corrections. Given a sampled completion, an editor identifies the first point of failure under a specified criterion and proposes a minimal replacement span. The model is then supervised only on these edited spans. By design, this concentrates supervision where the model's reasoning first breaks down and aims to preserve its existing reasoning patterns elsewhere. The method applies both offline and online, with the latter continually adapting supervision to the model's evolving failure modes. We instantiate this approach on reasoning instruction following with Qwen3-32B on the ReasonIF benchmark. Compared to full-trajectory imitation and trajectory-level reinforcement learning baselines, edit-based training substantially improves compliance with the reasoning instruction while largely preserving the model's skills on untargeted tasks. These results suggest that localized correction is an effective complement to these approaches for shaping how LLMs reason.
Large multimodal models can support long-horizon robot manipulation, but frequent model calls incur substantial latency and token costs. The central challenge is determining when agent assistance or intervention is needed and whether errors require correcting robot execution, subtask planning, or both. To address this problem, we propose RoC (Reasoning on Call), an efficient framework that coordinates a low-level policy and a high-level reasoning agent through an online task-state monitor. The monitor combines temporal visual observations with ordered interaction memory to jointly estimate progress within the active subtask and predict its successor. These predictions guide event-triggered agent review for coordinated execution repair and plan revision. Verified outcomes govern memory updates and subtask transitions, allowing routine execution to remain with the policy while reasoning focuses on decisions requiring additional assessment. On RMBench, RoC improves task success rate by 24.0 percentage points and reduces token consumption by 65.3% compared with the strongest agent-based baseline. It also outperforms the strongest policy-based baseline by 15.1 percentage points in task success rate. Real-world experiments further validate its effectiveness across three long-horizon manipulation tasks. Overall, RoC supports efficient agent intervention and coordinated recovery across execution and planning for long-horizon robot manipulation.
A one-pass decision encoder reads a text, a question and several named options, and returns a probability for each option in a single forward pass. We ask how far Laya, a 421M-parameter encoder of this kind, can be compressed before its decisions change. We count decisions that differ from Laya's own full-precision (fp32) answer, and track calibration (how well confidence matches accuracy, measured by expected calibration error, ECE), on 1,221 synthetic items with the training schemas, 273 with an unseen schema, and 800 out-of-family questions from ARC, MMLU and HellaSwag. Layer pruning of a related model family is known; we add flips, calibration, held-out data and retraining. bf16 changes 0.2% of in-family decisions but 2.2% out of family; 6-bit grouped rounding 1.3% and 5.9%; deleting 6 layers 0.9% and 9.1%, while ECE rises from 5.9% to 39.2%. One fitted temperature, needing no labels, restores it to 3.6%. One scale per tensor fails below six bits, and CPU dynamic int8 changes 30% of decisions, nearly all through one layer. Total variation tracks flips for rounding but not for deletion, and an early exit saves layers over a static cut only if its confidence gate is first recalibrated. Distilling a 16-layer student cuts its in-family flips from 44% to 9% (45% out of family). At equal stored size, rounding flips fewer decisions than deletion. Data are synthetic and LLM-labelled, for one architecture.