Science & Mathematics

Telecom language models must combine broad networking knowledge with specialist behavior across routing, broadband services, access technologies, and operational diagnostics while remaining practical to deploy as a single model. We present an end-to-end framework that builds source-grounded synthetic training data from structured networking taxonomies and expert graphs, materializes 212,559 routed training records across fourteen networking routes, trains six frozen specialist LoRA teachers over one shared continued-pretrained backbone, and consolidates their behavior into one deployable student adapter using Routed Specialist On-Policy Distillation (RSOD). In RSOD, the committed student generates each trajectory, the route-assigned specialist teacher-forces the exact native token sequence, compact teacher hidden states are reconstructed through the shared frozen output head, and only the student LoRA is updated. A controlled reconstruction validation shows mean absolute logit error of 7.26 x 10^-6 with 100% top-1 agreement between native and reconstructed teacher logits. On four RTX 6000 Ada GPUs, the completed 2,000-step run produced 16,000 trajectories with zero observed student-version or token- ID mismatches. Under the primary route-conditioned evaluation over 2,850 held-out questions, the raw Meta-Llama-3.1-8B-Instruct model scores 12.8661/25 micro (13.0172/25 macro), the combined routed six-teacher system scores 21.7641/25 micro, and the consolidated RSOD student, released as Llama-CalixCLM v1, reaches 21.7112/25 micro (21.8006/25 macro) on the same five dimension LLM-judge rubric. Under a separate neutral deployment-like prompt, the same student reaches 20.0028/25 micro (20.0778/25 macro). The consolidated student remains within 0.0529 micro-score points of the combined routed-teacher system while retaining a numerically validated and auditable train-serve execution path.

Tactile representation learning for dexterous manipulation often requires substantial robot-side data collection, limiting its scalability across different hands and sensors. We explore whether large-scale human hand-object interaction (HOI) data can provide a useful prior for robotic tactile learning. To this end, we propose HuTact, a framework that pretrains tactile-aware hand representations from human HOI data without requiring real tactile signals during pretraining. HuTact transfers the interaction structure learned from human hand motions to robotic tactile observations and can be further adapted with a small amount of unlabeled robot interaction data. We evaluate HuTact on tactile perception and manipulation tasks, including grasp prediction, slip detection, object classification, and tactile-conditioned policy learning. Across these tasks, human HOI pretraining improves data efficiency and provides a strong initialization for downstream learning. Notably, with HOI pretraining followed by midtraining on unlabeled robot play data, HuTact achieves 88.26% Macro-F1 on grasp prediction using only 25\% of the labeled downstream data, comparable to training from scratch on the full labeled dataset (88.23%). These results suggest that human interaction data can serve as a scalable complementary source for robotic tactile representation learning.

Learning loco-manipulation policies requires demonstrations that capture coordinated base and arm motion on the target embodiment. Robot-free interfaces allow demonstrations to be collected independently of the target robot, increasing flexibility but without directly capturing platform-specific commands, proprioception, and body-mounted observations. On-robot teleoperation allows to capture these signals directly, but requires an interface capable of commanding the complete system concurrently. This paper presents LocoPilot, a single-operator bilateral teleoperation and data-collection framework implemented on a 12-DoF Unitree Go2 quadruped equipped with a 6-DoF AgileX Piper L manipulator, an Intel RealSense D435 wrist camera, and an Intel RealSense D405 head camera. A second 6-DoF Piper L arm, used as the leader and two handheld controllers provide simultaneous control of the base, body posture, manipulator, and gripper. The follower arm interaction forces are estimated without a dedicated force/torque sensor and rendered to the operator through a passivity-constrained haptic channel. A ROS~2 pipeline records synchronized multimodal observations together with a unified 11D action shared by teleoperation and policy deployment. Compared to prior work, LocoPilot uniquely unifies robot-in-the-loop legged data collection, force feedback without motion-capture instrumentation, and the same action interface at deployment. We evaluate the interface across multiple loco-manipulation tasks and validate the complete collection-to-deployment pipeline by fine-tuning MolmoAct2 \cite{MOLMO} on pick-and-place demonstrations and deploying on the physical robot without action retargeting or additional inverse kinematics. Code available at: https://github.com/iit-DLSLab/LocoPilot

VLMs for medical image interpretation are typically evaluated on clean inputs, despite the progressive evidence loss that can occur in high stakes clinical settings. Existing robustness studies show that performance can degrade while confidence remains high, but do not establish when a previously reliable model-case pair ceases to behave reliably. We introduce a baseline-anchored framework that traces baseline-correct cases through progressive image degradation to identify an empirical evidence boundary, and define mirage as continued commitment at this boundary, regardless of correctness. Across three VLMs on three medical datasets, only MedGemma provides a sufficiently large baseline-correct pool for trajectory analysis. On VQA-RAD, the only dataset in which the question text varies, language-only answerability is associated with baseline correctness (AUROC 0.677) and with commitment at the boundary (AUROC 0.653 to 0.670). MedGemma's post-boundary behavior is heterogeneous, including direct mirage, correct recovery, mirage dissolution, consistent abstention, and oscillation, with mirage rates of 16.0 to 100.0% across dataset-modality pairs. Despite this behavioral heterogeneity, residual-stream probes decode commit-versus-abstain behavior at the boundary with AUROC of 0.895 to 0.994 under nested cross-validation. These results show that progressive evidence loss can produce persistent commitment that output entropy and abstention behavior do not consistently reflect, motivating boundary-aware evaluation, option-order controls, and improved abstention mechanisms for reliable medical foundation models operating under partial observation.

Diffusion models generate samples through a multi-step denoising trajectory, yet condition every step on the same instruction, so the trajectory cannot be addressed at the level of individual operations. We introduce Dancing Diffusion, which assigns each span of timesteps a dedicated continuous condition block and thereby turns conditioning into a latent program. A compiler maps text features into this program, and a fixed pretrained generator executes its blocks over successive sampling intervals. The trajectory can then dance, keeping its executed history while its continuation is redirected. A shared prefix is executed once and reused, and a later operation is rewritten while content fixed by earlier operations is preserved. The compiler is trained through the images it produces, with visual feedback on both the requested change and the retained content. On SANA, reusing an executed prefix preserves markedly more of the scene than regenerating from noise at a similar strength of edit, learned conditions preserve more background than native prompt switching from the same saved state, and the learned program improves generation quality over native SANA. Later branches weaken the edit, which leaves semantic control over individual operations limited. By converting a single-instruction trajectory into a program of addressable operations, Dancing Diffusion lays a foundation for structured intermediate computation in diffusion models.

Pathology foundation models (PFMs) support a wide range of histopathology tasks on whole-slide images through multiple instance learning (MIL). Despite recent progress, these models remain sensitive to cross-hospital distribution shifts: performance often degrades when a task head (e.g., classification) is trained on data from one hospital and then used on data from another hospital. To address this challenge, we propose AdaptMIL, a training strategy for integrating domain adaptation into MIL training. AdaptMIL is motivated by our finding that directly applying domain adaptation objectives during MIL training can interfere with MIL aggregation learning, often resulting in performance degradation over standard training. AdaptMIL resolves this by decoupling representation-level domain alignment learning from MIL aggregation learning through a two-stage training procedure. We show that AdaptMIL improves cross-hospital performance and outperforms alternative adaptation methods across experiments spanning multiple WSI datasets, tasks, PFMs, MIL architectures, and domain adaptation objectives. AdaptMIL's code will be released upon publication.

We present lower bounds for the maximum minimal hitting set problem and for the problem of enumeration of minimal transversals in bounded-dimension hypergraphs. Our starting point is the recent reduction of Kobayashi, Kurita, and Wasa (2026), developed to establish ETH-based hardness of hypergraph dualization. Adapting their construction, we show that, under ETH, deciding whether a hypergraph $\mathcal{G}$ with $n$ vertices and $m$ hyperedges has a minimal transversal of size at least $k$ cannot be done in time $n^{f(k)}m^{o(k)}$, for an arbitrary computable $f$. We transfer this bound to the following decision problem: given a dimension-at-most-$r$ hypergraph and a family of $s$ of its minimal transversals, decide whether another exists. Under ETH, this problem admits no uniform algorithm with running time $n^{f(r)}(s+1)^{o(r)}$. Under SETH, for every fixed $t \geq 2$, it admits no algorithm with running time $\poly(|\mathcal{G}|)(s+1)^c$ for dimension at most $2t$ and any $c

Echo chambers and filter bubbles provide a framework to view communication and connection in a social network age. Such systems are characterised by isolated communities regurgitating information which reinforces prior beliefs. Theoretical analysis of echo chambers is suited for a network science approach to understand the mechanics of information influences between users. Here, we propose an evolutionary network model that extends the Quoter Model, rewiring edges according to cross-entropy between text-generating nodes. Parameters characterise binary reinforcement vs exploration node intents, determining whether a node will seek high or low novelty neighbours, measured by cross-entropy. Through an empirical simulation study, we observe emergent model attributes. Specifically, we compare to a Watts--Strogatz model, highlighting parameter regions of high clustering and hyper-exploration. Further, we identify a quality of disjoint communities akin to rudimentary echo chamber structures. These findings underscore user behaviours that drive echo chamber entrenchment, contextualising future social network analysis.

Examining existing benchmarks for grounding in 3D VLMs reveals a pervasive "semantic leakage" in the evaluation and prompting protocols: a heavy reliance on bounding boxes as spatial prompts or representation primitives. Because bounding boxes are highly informative of object categories, they introduce a shortcut for the models to neglect visual information during 3D VLM training. We demonstrate this with a "blind" LLM that, prompted only with the box coordinates, can perform surprisingly well. To further analyze this shortcut, we propose point-based grounding prompting - querying models with a single 3D point instead of a box. Under this protocol, the accuracy of the blind LLM drops from 36.4% to 18.0% on ScanNet's Nr3D, and similarly degrades for 3D VLMs (e.g. LL3DA). Building on this insight, we introduce P$^3$-VLM, a novel tri-pathway 3D VLM that supports point-based prompts and achieves 52.1% state-of-the-art accuracy on location-grounded captioning under this setting. To achieve this, we redesign the architecture while keeping it detector-free, scene-centric, and suitable for 3D GS inputs. P$^3$-VLM utilizes three complementary pathways (location-, task-, and context-aware), enabling strong reasoning about scenes without any bounding box dependencies. The key benefit of point-based learning is a consistent improvement across settings: P$^3$-VLM achieves state-of-the-art performance under point-based grounding while also improving robustness, OOD generalization, and performance on ungrounded scene-centric 3D VQA benchmarks, reflecting its reduced reliance on bounding-box shortcuts and resulting stronger visual-spatial reasoning.

Reinforcement learning (RL) enables adaptive and autonomous decision-making in embedded, edge, and heterogeneous computing systems. Field-programmable gate arrays (FPGAs) and integrated circuits (ICs) offer low latency, energy efficiency, and architectural flexibility for RL deployment, while introducing challenges in resource constraints, implementation complexity, and hardware-algorithm co-design. This review surveys more than 100 studies on RL implementations across FPGA and IC platforms, emphasizing architectures, implementation strategies, performance evaluation, and applications. The literature is organized by RL family, including model-based Markov decision processes, classical value-based methods, deep RL, actor-critic, policy-gradient, and multi-agent approaches. We further classify studies by abstraction level, hardware realization strategy, and application domain, examining how RL algorithms map onto reconfigurable and fixed hardware. Unlike surveys restricted to particular algorithms or platforms, this work provides a unified analysis of FPGA and IC implementations, covering acceleration, reconfigurable architectures, energy-aware control, and system-level optimization. Comparative assessments consider resource utilization, latency, throughput, power consumption, and reported performance gains. Finally, we identify recurring architectural patterns, practical limitations, open challenges, and research directions for efficient RL-driven hardware systems. The survey offers a consolidated reference for researchers developing and evaluating RL solutions on specialized hardware.

Practitioners use reinforcement learning methods for language models through training frameworks, so a method that a framework includes is one configuration change away for its users. We measure how often this happens. We trace seven open-source frameworks for RL post-training through their commit histories and link them to 6,939 screened arXiv papers from January 2025 to June 2026. Of 1,598 papers that propose a general RL training method, 37 reached at least one framework. Across five quarterly cohorts of papers, the number included within 180 days stayed between 5 and 7 while the cohorts grew from 72 to 386 papers. Compared with papers from the same quarters, included papers more often trained policies of 30B parameters or more, at 41 against 10 percent, and more often changed a coefficient, clip, mask or weight in the existing loss, at 29 against 9 percent. Both gaps hold without papers from the companies that maintain the frameworks, while code release was equally common in both groups. The frameworks' issues and pull requests show that few methods are turned away in public. More than nine in ten of the general methods no framework included have no thread there that cites or names them, and many proposals that do appear stay open without a reply. Two further records, the frameworks' issues and the technical reports of 13 large open-weight models, show that users rarely report training failures in public and that the reports state 8 of the 13 widely included methods. We release the data as a tracker that links training failures to framework implementations and lists each framework's contribution rules and earlier proposals.

For a planar graph $G$ with at least one edge, let $ρ(G)$ be the infimum of the ratio of the longest to the shortest edge length over all planar straight-line drawings of $G$. We prove that every $n$-vertex $2$-tree $G$ satisfies $ρ(G)\le4+5\lfloor\log_2(2n-3)\rfloor$. This improves the known $O(\sqrt n)$ upper bound and, together with the lower bound of Blažej, Fiala and Liotta, shows that the maximum of $ρ(G)$ over $n$-vertex $2$-trees is $Θ(\log n)$. The proof adapts Biedl's recursive bar construction using nonintegral heights and applies a height-preserving straightening theorem.

High-quality training data are a major bottleneck in machine learning, motivating increasing use of artificially generated data. We study when a binary classifier trained on synthetic data transfers successfully to a real distribution, under a shared conditional label distribution. First, we derive finite-sample excess-error bounds when the real covariate distribution is absolutely continuous with respect to the synthetic distribution under a controlled density ratio. Second, we study the partially and fully singular case, where the real distribution has mass outside the synthetic one. If the hypothesis class is sufficiently expressive, synthetic risk alone cannot uniformly control real-world error: a classifier can be nearly optimal on synthetic data while approaching worst-case error on the uncovered real region, even when the distributions are arbitrarily close in Wasserstein distance. For function classes where hypotheses and the log-odds of the conditional label probability are Lipschitz, we recover a bound in terms of the probability that a real sample lies outside the reach of the synthetic data, restoring a transfer guarantee even in the fully singular case. Experiments confirm this in controlled low-dimensional settings and on 32,200 images generated as counterparts to real ImageNet images by four state-of-the-art generators: an adversarial classifier reaches 0.5% error on the synthetic domain but 99.25% error on the real domain, while ordinarily trained architectures show positively correlated synthetic and real test errors. Synthetic data should be budgeted at high sample counts and models validated on real data.

Adiabatic demagnetization refrigeration (ADR) based on frustrated rare-earth magnets provides a promising route toward helium-free sub-Kelvin cooling, while balancing magnetic entropy and interaction scales is essential for optimizing the ADR functionalities. Here we investigate the low-temperature magnetic and ADR properties of the alkali rare-earth double phosphates K$_3$$Ln$(PO$_4$)$_2$ ($Ln$ = Yb, Gd), which crystallize in a monoclinic $P2_1/m$ structure containing distorted isosceles triangular lattices of rare-earth ions. Magnetization and specific heat measurements confirm the absence of long-range magnetic ordering down to 0.1 K in both compounds. Direct quasi-ADR measurements demonstrate strongly contrasting refrigeration behaviors: K$_3$Yb(PO$_4$)$_2$ reaches a minimum temperature of 29.8 mK from an initial state of 2 K and 6 T, whereas K$_3$Gd(PO$_4$)$_2$ exhibits extended hold times below 150 mK associated with its larger magnetic entropy reservoir. Crystal-electric-field calculations confirm a $J_{\mathrm{eff}} = 1/2$ ground-state doublet in K$_3$Yb(PO$_4$)$_2$, while thermodynamic analysis based on magnetic entropy and interaction scales accounts for the distinct ADR performance of the two compounds. These results identify K$_3$$Ln$(PO$_4$)$_2$ as a chemically robust platform for rare-earth-tunable solid-state ADR.

The magnetic ground states of rare-earth frustrated magnets can be governed by competing energy scales arising from crystal-field anisotropy, exchange interactions, and long-range dipolar interactions. Here we investigate this interplay in the triangular-lattice antiferromagnet NaBaEr(BO$_3$)$_2$ by combining magnetization, specific heat, and neutron scattering measurements. Neutron powder diffraction resolves rotations of the ErO$_6$ octahedra and supports the lower-symmetry $R\bar{3}$ structure over the previously proposed $R\bar{3}m$ model. Magnetization measurements and inelastic neutron scattering establish an effective spin-1/2 description of its low-temperature magnetism with pronounced easy-axis anisotropy, arising from a well-isolated Kramers ground-state doublet separated from the first excited crystal-electric-field level by approximately 4 meV. Below $T_{\rm N}=0.1$ K, single-crystal neutron diffraction reveals A-type antiferromagnetic order with the ordered moment predominantly oriented within the $ab$ plane, despite the pronounced paramagnetic easy-axis anisotropy. The estimated dipolar interaction scale is comparable to $T_{\rm N}$, and classical dipolar-energy calculations for the observed stacking pattern favor an in-plane moment orientation. These results reveal contrasting crystal-field and dipolar anisotropies in determining the magnetic ground state of NaBaEr(BO$_3$)$_2$.