Papers

A rainbow $k$-dominating function assigns the empty set or exactly one of the colors $\{1\},\{2\},\ldots,\{k\}$ to each vertex of a simple graph in such a way that every vertex receiving $\emptyset$ sees all $k$ colors in its neighborhood. This model has several facility-location interpretations where colored vertices represent sites hosting different types of services, and uncolored vertices represent the beneficiaries of those services who must have access to all of them. In this paper we study this domination variant in regular graphs. First, we develop lower and upper bounds for this class of graphs, and explore their structural consequences. Then, we use those results to prove NP-completeness of a decision formulation of this problem. Additionally, we derive exact formulas for the rainbow $k$-domination number of cycles, Möbius ladders and prisms. Lastly, we develop a dynamic-programming algorithm for generalized Petersen graphs of the form $P(6t, t)$ and state two conjectures based on the output of its implementation and related integer linear programming computations.

Large language models (LLMs) progressively support embedded hardware development, but a plausible recommendation may pose safety risks when acting on the physical hardware. Because electrical safety can be different between every microcontroller board, a practical question is raised: can LLMs have the capability to make the right decision on whether a proposed user operation is safe for the board? Although benchmarks for embedded development exist, they primarily target code generation and hardware design tasks. This highlights the lack of tested board-specific judgment regarding electrical safety. To address this gap, we introduce the ElecSafety benchmark to evaluate whether an LLM can respond with the correct label and derive that label from the proper manufacturer constraints and rules. We extract these constraints from vendor datasheets and hardware manuals. Each scenario is defined by a decisive condition and a gold label (safe, hazardous, or cannot determine) before translating the scenario into natural language. We categorize the scenarios into violation and compliant cases, board-swap, near-boundary, and missing-information. We then evaluated the scenario on six open-weight LLMs under various conditions of rule access and token limitations. The finding shows board rules can improve accuracy, whereas limiting the token has a smaller impact. Even the best configuration remains unreliable on near-boundary and missing-information scenarios, and a correct label often rests on the inaccurate constraints.

We develop a sparse graph method for transporting probability mass toward multimodal target distributions through damped nonlocal Hamiltonian dynamics. The formulation combines logarithmic-mean mobility with symmetric Lévy-type interaction weights, coupling the evolving density to an edge momentum field. A graph constructed from nearest-neighbor connections and sampled long-range edges provides direct mass exchange between spatially separated regions. Once the graph is constructed, the density evolution is deterministic, and each update costs linear in the number of nodes and the long-range sampling budget. Linearization around the target distribution yields a damped oscillator governed by a weighted graph Laplacian. Its spectrum characterizes the interaction between nonlocal connectivity and inertia, with the Lévy exponent alpha tuning the nonlocal connectivity: the spectral gap determines the optimal asymptotic damping, while the largest eigenvalue governs the time-step stability. Experiments on synthetic multimodal distributions demonstrate improved mode balance and more stable mode coverage relative to first-order and MCMC baselines. The resulting framework provides a sparse implementation of nonlocal inertial density transport for sampling problems with low-dimensional spatial structure.

Purpose: To evaluate the performance-feasibility tradeoffs of foundation models (FMs) and task-specific convolutional neural networks (CNNs) trained from scratch for breast implant classification in 2D mammography, with emphasis on suitability for near real-time clinical deployment. Methods: We evaluated four models: two FMs (RAD-DINO and MammoCLIP) and two CNNs trained from scratch for implant prediction (ResNet18 and our lightweight ResNetLite). Using the Emory Breast Imaging Dataset, 5,000 unilateral screening mammograms were used for training/validation and 1,000 manually reviewed unilateral images were held out for testing. For the FMs, global image embeddings from the pretrained encoder were classified using a support vector machine (SVM). The CNNs were trained end-to-end on 2D mammograms, with ResNetLite optimized via grid search over depth and width to balance accuracy and efficiency. Performance was evaluated using AUROC, sensitivity, specificity, accuracy, embedding visualization, and inference-latency. Results: All models demonstrated strong performance on held-out test data (n = 1,000). MammoCLIP achieved the highest AUROC (0.999) with the quickest training time of 493 seconds. RAD-DINO achieved the highest sensitivity (0.980; accuracy 0.989) but had the slowest inference and training times. ResNet18 and MammoCLIP achieved comparable accuracy (0.985). ResNetLite showed no statistically significant difference from ResNet18 (AUROC 0.993; accuracy 0.976) despite using only 1.4% of ResNet18's parameters, and had the fastest inference time. Conclusion: FMs and task-specific CNN models reliably detect breast implants on 2D mammography. Model selection is best guided by deployment context: MammoCLIP for GPU-equipped hospital settings requiring scalable integration, and lightweight CNNs such as ResNetLite for resource-constrained or edge deployments.

The Capacitated Vehicle Routing Problem (CVRP) remains a central concern in logistics research, as route optimisation under rigid capacity and fleet restrictions directly affects operational performance. Genetic Algorithms are widely used for this problem, yet their effectiveness depends heavily on the choice of solution encoding and on the mechanisms used to preserve feasibility during the search. This paper proposes a sequential framework that separates the evolutionary process from route construction. We adapt Node Shift Encoding (NSE) to the CVRP giant-tour setting and benchmark it against Path Representation and Double Chromosome. We further propose an unsupervised machine-learning repair heuristic that restores fleet feasibility after an unconstrained linear split, and compare this pipeline against fleet-constrained split taken from the literature. Computational tests on standard benchmark instances show that NSE consistently outperforms the other encodings in solution quality and feasibility. The repair heuristic restores feasibility in approximately 72% of cases, and the Linear-Split-ML-Heuristic pipeline proves more robust than the constrained split, delivering lower costs and higher feasibility rates.

A time-series forecast is scored on a future value of the series. A representation loss that regresses the next latent, as in LeNEPA, is a different least-squares problem on the same bottleneck. We write both programs down. The forecast program minimizes the error of a decoded latent on the coordinate that will be reported. For a scalar target and a linear decoder, every latent rank of at least one matches ordinary least squares, and an isotropy constraint is only a rescaling: after the decoder is refit, the forecast does not move. The other program fits the whole next vector at a fixed rank, then freezes the encoder and attaches a head. On a four-dimensional series whose last three coordinates are the same autoregression, that rank-1 fit puts mass $0.9998$ on the repeated coordinate and forecasts the remaining signal at the marginal variance $2.794$. The forecast program puts mass $1$ on the signal and matches the innovation variance $0.992$. Rank $2$ gives the vector fit a second direction, and the two programs agree. Iterating the fitted one-step coefficient $0.803$ raises the open-loop error from $0.992$ at one step to $2.700$ at eight steps.

A wiring diagram is an arrangement of $m$ pseudolines, every two of which cross exactly once. Reading its crossings from left to right encodes it as a reduced word of the permutation $w_0\in S_m$ that reverses $1,\dots,m$, unique up to commutations, and a braid move $\sigma_i\sigma_{i+1}\sigma_i\leftrightarrow\sigma_{i+1}\sigma_i\sigma_{i+1}$ flips one of its triangles. We prove that deciding whether two wiring diagrams are within a given braid-move distance is NP-complete. Equivalently, deciding flip distance is NP-complete in the higher Bruhat order $B(m,2)$, for triangle flips of marked arrangements of pseudolines, and for rhombic tilings of a $2m$-gon; this answers an open question on flip graphs of pseudoline arrangements (SODA 2024). The reduction is from Vertex Cover: for every graph $G$ we construct two wiring diagrams at braid-move distance $|D|+2\operatorname{VC}(G)$, where $\operatorname{VC}(G)$ is the vertex cover number of $G$ and the Hamming distance $|D|$ is the number of triples of wires whose triangles are oriented differently. The construction glues copies of a seven-wire gadget whose braid-move distance exceeds its Hamming distance by $2$. Adding a line at infinity transfers the construction to rank-$3$ uniform oriented matroids, where two oriented matroids are adjacent in the mutation graph if their chirotopes differ in one basis; so deciding distance in the mutation graph is NP-complete in rank $3$. An iterate of Rambau's expansion of signotopes amplifies the excess and gives NP-hardness in every fixed rank $r\ge4$, and by duality in every fixed corank at least $3$. The same instances show that flip distance in the higher Bruhat orders $B(n,k)$ is NP-complete for every fixed $k\ge2$.

Lattice constellations can provide significant coding and shaping gains over traditional quadrature amplitude modulations (QAMs), but have not been widely adopted due to the high implementation complexity of modulation, demodulation, and decoding. Voronoi constellations (VCs) enable an efficient modulation map that achieves shaping gain using the Voronoi cell of a lattice. Multilevel coset codes (MLCCs) are a recently proposed coded modulation scheme for VCs that realizes more efficient decoding. However, MLCCs still suffer from a soft demodulation complexity of $O(d2^{sd})$ for a $d$-dimensional lattice with spectral efficiency $s$ bits per real dimension. In this paper, we propose fast quantize-walk demodulation (FQWD), an efficient demodulation scheme for MLCCs that reduces the complexity to $O(sd2^d)$. Our strategy utilizes novel radial projections onto the Voronoi cells of the $D_4$ and $E_8$ lattices. Experimental results demonstrate that FQWD has only negligible loss in block error rate (BLER) relative to maximum likelihood demodulation under the additive white Gaussian noise (AWGN) channel. Using FQWD, we simulate MLCCs on higher-dimensional lattices with $s>2$ bits per dimension under AGWN for the first time, achieving BLER gains over multilevel polar codes on QAMs of up to 0.6 dB for the $D_4$ lattice and 1.2 dB for $E_8$.

We propose a two-level filtered subspace iterative (TLFSI) method for computing all eigenpairs whose eigenvalues in the interval of interest $[a,b]$ after finite element discretization of self-adjoint PDE eigenvalue problems. The method combines a two-level approach with a modern filtered subspace iteration framework. In this paper, we use contour integral filtering as an example. The two-level method constructs a coarse space according to the spectral location of the eigenvalues of interest and obtains from it an approximation to the target invariant subspace. We prove that the distance between the approximate and target subspaces depends only on the separation between the spectral region of interest and the remainder of the spectrum, together with the coarse mesh size \(H\). Numerical experiments on benchmark elasticity problems confirm the theoretical predictions.

Online students often face emotional challenges such as frustration, isolation, and fluctuating motivation, which can hinder sustained engagement in online learning. While affective chatbots have shown promise in healthcare and wellness, most learning support chatbots focus primarily on cognitive assistance with limited attention to learners' emotional and motivational experiences. In this work, we investigate how adaptive affective chatbot support can be designed for online learning through an iterative design process. Through a mixed-methods needfinding study (n=38), we identified substantial variation in how learners respond to affective support, with no single strategy universally preferred. This led us to develop three adaptive design alternatives: user-controlled, AI-driven, and hybrid co-controlled adaptation, and examine key trade-offs between automation, transparency, and user control. A final evaluation (n=25) showed that 88% of participants favored the adaptive hybrid design over a non-adaptive alternative, highlighting the value of balancing transparency, user control, and interaction simplicity. We discuss design implications for adaptive affective chatbot systems in emotionally challenging online learning contexts.

Context: Fine-grained emotion classification of mobile app reviews enables requirements engineering activities that go beyond polarity-based opinion mining, including emotionally informed issue prioritisation and feature-oriented feedback analysis. However, automatic fine-grained emotion extraction from app reviews remains understudied. Objectives: Building on a previously published annotation framework and human-labelled ground truth adapted from Plutchik's taxonomy, this paper investigates how large language models can be leveraged for automatic multi-label emotion classification under severe class imbalance. Methods: We compare encoder-only fine-tuning under multi-label and binary-ensemble formulations, decoder-only zero- and few-shot prompting across open-source and proprietary models, and a catalogue of imbalance mitigation strategies (loss reweighting, resampling, generative data augmentation), with the synthetic-review generator and prompting strategy selected via an intrinsic augmentation-utility ranking. Results: Fine-tuned encoders trail the best decoder-only few-shot prompting (macro-F1 0.642) by a wide margin at baseline (multi-label: 0.387; binary ensemble: 0.450); pairing the best multi-label encoder with generative data augmentation and positive-weighted loss closes most of this gap (+0.204) at up to three orders of magnitude lower inference latency than the decoders, with the largest gains on the rarest emotions, from undetected to gains of up to +0.501 F1. Conclusion: Large language models make fine-grained, multi-label emotion classification of app reviews feasible for requirements engineering pipelines, with modest macro-F1, and the best formulation and mitigation strategy are backbone- and formulation-dependent. We release the experimental pipeline, synthetic corpora, and fine-tuned checkpoints for replication and reuse.

Recovering executable CAD programs from 3D meshes is challenging due to the compositional nature of CAD construction and the interaction between discrete modeling choices and continuous parameters. Many learning-based methods predict complete programs in a single pass and rely predominantly on sketch-extrude representations, limiting operation diversity and opportunities to correct geometric errors during reconstruction. We introduce StepCAD, a generative optimization approach that combines a state-conditioned CAD policy with geometry-guided search. Given an input mesh, the policy predicts construction actions conditioned on both target and intermediate geometry, and an IoU-guided tree search refines the resulting program through local edits. We also introduce ARCADE-1.5M, a large-scale dataset of 1.5M executable CAD programs spanning diverse operations, sequences with a maximum length of 150+ counted operations, and 12.5M intermediate state-action transitions. Experiments across multiple CAD reconstruction benchmarks show that StepCAD achieves state-of-the-art geometric reconstruction accuracy with consistently high validity, yielding up to 87.2% relative IoU improvement over the strongest evaluated baseline, with particularly large gains on complex shapes. Project page: https://ghadinehme.com/stepcad.github.io/

World Action Models (WAMs) leverage pretrained video foundation models for robot manipulation, but their large backbones and video-action co-prediction are expensive. Although existing acceleration techniques offer many ways to reduce this cost, selecting and composing them requires substantial engineering for each model and hardware platform. To tackle this bottleneck, we present WAMJET, an agentic harness that accelerates WAM inference by equipping coding agents with reusable optimization guidance and measurement and validation tools. WAMJET follows a bottleneck-driven workflow where the agent profiles inference, modifies targeted code, validates effects, and iteratively refines the acceleration stack as bottlenecks shift, while preserving action quality. Experiments span six WAMs, three coding agents, and two GPU architectures. WAMJET achieves up to 9.95x lossless speedup over upstream implementations. Approximation and hardware-aware optimization yield additional latency reductions, with comparable success rates. The results show that WAMJET can produce effective acceleration stacks for WAM deployment.

Rigorous schedulability analysis is essential for the design of hard real-time systems, yet errors in pen-and-paper proofs threaten the safety of critical applications. The Prosa initiative addresses this by offering a foundation for building machine-checkable schedulability analysis proofs in the Rocq proof assistant. However, the substantial time and expertise required to construct such proofs remain a major barrier for wider adoption of Prosa. This work presents ProsaBuddy, an LLM?based agent system designed to lower the effort needed to develop mechanized real-time schedulability proofs. ProsaBuddy employs a ReAct loop with retrieval over the Prosa codebase, access to Rocq tools and optional human-written hints. It uses a subgoal?delegation architecture, decomposing a lemma into subgoals and dispatches them to subagents for proof. We evaluate ProsaBuddy on a mini benchmark drawn from real-time scheduling literature. Experiment results show that ProsaBuddy significantly outper?forms state-of-the-art LLM-based Rocq automated proving agent systems and a general coding agent OpenCode

Formal verification of quantum circuits becomes difficult as circuit size grows because exhaustive reasoning over the Hilbert space is computationally expensive. In previous work [1], we introduced a bit-vector abstraction that separates superposition tracking from basis-state behavior, enabling SMT-based verification of fixed-size error-detection and entanglement circuits built from H, X, and C-NOT gates. This paper extends the framework in five directions. First, we generalise correctness properties to parametrised repetition and concatenated phase-flip codes with up to 101 logical elements. Second, we verify syndrome extraction for CSS codes specified by arbitrary parity-check matrices, including the Steane, [[15,7,3]] Hamming, [[15,1,3]] Reed-Muller, and rotated surface codes with distances up to 101. We consider arbitrary error patterns, detection up to a specified weight, and recovery using a classical decoder, with counterexamples beyond each code's correction or detection capability. Third, we characterize the {H,X,C-NOT} fragment for which the abstraction faithfully represents Hilbert-space semantics. We prove soundness, demonstrate necessity with counterexamples, validate the characterisation against state-vector simulation, and prove the six lemmas from the original work. Fourth, a mutation study using 13 operators identifies equivalent mutants and detects all 622 non-equivalent mutants within the evaluated fault model. Fifth, a direct Z3 implementation verifies GHZ circuits with up to 262,144 qubits across three topologies on a 16GB laptop. Random H+C-NOT networks show verification cost rises sharply with depth when outputs depend on many inputs. The approach does not target general stabilizer circuits, Y-type stabilizers, fault-tolerant protocols, or physical noise; we identify extensions needed to address these cases.

Vehicle-to-everything (V2X) sidelink and uplink broadcast must deliver safety-critical and application-layer data reliably under fast time-varying fading, intermittent packet erasures, and exposure to passive eavesdroppers. Fixed-rate physical-layer codes are poorly matched to these conditions because link quality and the number of successfully received symbols vary across vehicles and over time. This paper proposes a unified vehicular communication framework that combines rateless Construction-D' Raptor lattices, nested-lattice physical-layer security (PLS) with direct-sequence spread spectrum (DSSS), and blockchain-assisted trust management for decode-and-forward (DF) cooperative relaying. At the physical layer, Raptor codes provide incremental redundancy over unreliable V2X links; multilevel lattice decoding enables power-efficient joint coding and modulation; nested shaping and spread spectrum limit fine-lattice recovery by eavesdroppers; and a distributed ledger records decoding success at relays to support reputation-aware relay selection. We model the vehicular air interface as a block-Rayleigh fading channel with additive Gaussian noise and random symbol erasures, representative of urban V2X sidelink behavior. The proposed code concatenates Luby-transform (LT) codes with quasi-cyclic low-density parity-check (QC-LDPC) precodes within Construction-D', with girth-4 cycle removal on the Tanner graph. AWGN benchmarks show competitive symbol error rate (SER) and improved frame error rate (FER) versus matched-length LDPC and LDLC lattices. Under block-Rayleigh fading with erasures, for n=995 at average VNR of 1.5 dB and erasure probability 0.15, rateless Raptor reception achieves 0.3% outage with about 2.3 packets on average, outperforming LDLC and LDPC under identical channels, while an eavesdropper with 8 dB combined SNR and spreading loss fails in 62% of sessions.

A cardinality constraint written on an edge of an entity-relationship diagram admits two opposite readings. Under the reading used by UML and by Chen's original model, the label is read with the entity set on the same side of the relationship; under the reading used by standard database textbooks, it is read with the entity set on the opposite side. The two readings are exact opposites, so a reader who assumes the wrong one takes away the opposite of what the designer meant. For binary relationships the difficulty is confined to interpretation, because the two constraints a binary relationship set admits can both be drawn. For relationships among three or more entity sets the picture is worse: a ternary relationship set admits twelve cardinality constraints, and a diagram with three edges can carry at most three of them. This paper presents a notation, Card(R; p; q) = (lower, upper), that removes the ambiguity without taking a side in it, and that is not limited to one constraint per edge. We give the number of constraints an n-ary relationship set admits, show which of them a design determines without ever writing them down, and state two inference rules, decomposition and augmentation, that derive one constraint from another, together with the side conditions under which each is sound. We close with a worked case study that turns three business requirements into three explicit constraints and nine more that the design decides on its own.

Event cameras have microsecond-level temporal resolution and high dynamic range which make them more resilient to motion blur and poor illumination than standard frame-based cameras. Event-camera visual odometry (VO) pipelines maximize these benefits when they process the asynchronous event stream at the native temporal resolution. Continuous-time Gaussian process (GP) regression and a white-noise-on-acceleration (WNOA) prior can handle asynchronous measurements but result in a prohibitively large estimation state when applied naively. This paper presents a continuous-time event-camera stereo VO pipeline that maintains the native measurement times of asynchronous events while also running in real time. It reduces the estimation states to keytimes while maintaining full temporal resolution by interpolating measurements to their exact timestamps with a physically founded WNOA prior. This decouples the state size from the dense number of measurements without discarding their asynchronous nature. The real-time continuous-time VO pipeline is evaluated on the MVSEC and DSEC datasets. It provides estimates in real time that are more accurate than ES-PTAM, a state-of-the-art discrete estimator, in all but one of the tested sequences. The pipeline respectively provides estimates at 22 Hz and 6 Hz on MVSEC and DSEC and RMS relative errors of 0.46 cm and 0.038 degrees across all valid sequences, which were 11 and 15 times better than ES-PTAM, respectively.

In instruction-guided generative recommendation, LLM-based recommenders need to balance two goals: responding to the user's current request and aligning with the preferences in their interaction history. When the two conflict, history events can override the request. We show that turning the effect of individual history events into supervision faces two obstacles. First, the events that most influence a recommendation are not necessarily the ones that support the target item. Second, removing a misleading event can raise the target's score but a competing item's score even more, so a higher target score alone does not guarantee a better ranking. We propose Asymmetric Intervention-Guided Margin Supervision (AIMS), which converts the effect of removing individual history events into ranking supervision. For training requests already ranked correctly, a frozen reference model identifies request-specific deletions that improve both the target's score and its margin over a competitor near the recommendation cutoff. These margins serve as training targets, while the complete history is retained as input. Training combines cross-entropy with an asymmetric auxiliary loss that penalizes margin shortfalls and routes its gradient only through the competitor score. Inference is unchanged, requiring no history editing or deletion search. Across six LLM backbones on an industrial dataset and two public benchmarks, AIMS improves Recall and NDCG over strong baselines. Ablations support request-specific margins and asymmetric supervision, and the selected deletions preferentially remove constraint-violating history.

Sustainable protein discovery lacks the fast computational proxies, analogous to molecular docking or density functional theory, that accelerate drug and materials discovery. Evaluating whether a novel food tastes like its animal-based target requires expensive human sensory panels, bottlenecking the design-build-test loop. We introduce TasteBench, a multimodal benchmark and privacy-preserving competition for sensory prediction, spanning two tasks: a food-level ranking task built on 21K+ human evaluations across 215 plant-based foods in 24 product categories, yielding 935 within-category ranking pairs, and a supporting molecular-level taste classification task over 15K flavor molecules. To enable rigorous interpretation of model performance, we characterize the ground truth: inter-rater agreement among panelists is low (Krippendorff's $\alpha = .077$), and the split-half reliability ceiling of panel-aggregated rankings is .825, establishing the range within which ML systems on this benchmark should be assessed. We evaluate baselines across four input modalities; on the same pairs panelists rated, the best model achieves .661 pairwise accuracy, competitive with the median individual panelist (.650), and .683 across all within-category pairs. TasteBench provides the evaluation infrastructure and baselines for measuring progress on computational screening for sustainable protein discovery.

Deploying LLMs on consumer-grade GPUs with insufficient memory to hold their weights can result in prohibitively slow inference, because decoding repeatedly transfers offloaded weights from system RAM or flash storage into GPU at much lower bandwidth than local GPU-memory access. Activation sparsity reduces these transfers by skipping weights associated with zero or near-zero activations. However, as more activation contributions are omitted, model quality eventually degrades rapidly, indicating that weights associated with small-magnitude activations collectively influence model quality sharply. In this work, we improve the trade-off between model quality and decoding performance when exploiting activation sparsity. Our key idea is to replace the binary choice of whether or not to read a weight with three options: fully retain it, approximate it using a compressed weight representation, or omit it entirely. SpAx skips weights associated with activations closest to zero, reads approximate weights for smaller-magnitude activations, and reads original weights for the largest-magnitude activations. Smaller-magnitude activations attenuate the errors introduced by approximate weights, while compressed weight representations require fewer bytes to be transferred. With weights offloaded to CPU memory, SpAx speeds up decoding by 3.86X on average (up to 5.57X) with 16-bit weights and 2.06X (up to 2.74X) with 4-bit weights, at a WikiText-2 perplexity increase of at most 10%. With weights offloaded to flash storage, the speedups are 3.31X on average (up to 4.81X) and 1.54X (up to 2.03X).

Vision foundation models such as DINOv2, SigLIP2, and MASt3R develop complementary capabilities from different pretraining objectives, yet their knowledge remains distributed across separate, specialized models. Multi-teacher knowledge distillation offers a path toward consolidating these capabilities into a single agglomerative backbone, but existing approaches assume a fixed set of teachers, and incorporating a new teacher requires repeating expensive joint distillation over the entire teacher set. We introduce GRAFT, a continual multi-teacher distillation framework that enables a unified backbone to progressively acquire capabilities from an open-ended sequence of foundation models. When a new teacher arrives, GRAFT treats the previously distilled model as a teacher for preserving learned capabilities, while the current student jointly learns from both the previous model and the incoming teacher. Furthermore, to reconcile the incompatible representation geometries of heterogeneous teachers, we introduce Teacher Specific Readout Tokens, which grant each teacher an independent read-out of the shared encoder, together with Geometry Agnostic Relational Loss that aligns a vision-language teacher by matching image-text similarity structures rather than raw feature values. We provide GRAFT model, which is a single, continually extensible backbone that unifies five domains, including image understanding, 2D dense prediction, 3D human pose estimation, 3D vision, and vision-language, delivering strong performance across all of them while acquiring each new capability at the cost of a single distillation rather than a full re-distillation.

Online education has enabled learners worldwide to take up courses from reputed institutions. However, a self-paced online course cannot guarantee the motivation and engagement that a learner experiences in a real-time classroom. Self-paced access also makes platform load unpredictable, which drives up the compute cost. We propose a Commitment Scheme for course access that aligns learner commitment with platform capacity. Learners book their study slots in advance; the instructor sets the budget of learning hours available for the course; and learners who use a full window earn additional watch hours. A booked window records an intention to study at a stated time, and a learner who appears in that window implements it. The byproduct is a platform load that can be forecast and bounded. The study involved two courses taken in sequence by the same learners, with the slot booking system activated only in the second. In-window study was observed on 86.8% of booked windows, and the median committer placed 95.5% of all study time inside self-booked windows. Among learners who studied across the launch, study regularity improved from 1.01 to 1.33 active days per week, with a supporting difference of +1.37 days per week against the same learners' preceding course. Commitments made on the same day as the study slot were honoured more often than advance bookings (88.8% against 62.5% two days ahead), which is consistent with the classic intention-behaviour gap.

Numerical measurements capture how a system behaves, but often leave the meanings of its variables unspecified. Some variables are measured but never labeled, and others are never measured at all. Existing methods assign semantics to such variables by consulting general human knowledge, but this inherits its biases where that knowledge exists and offers nothing where it does not. We bridge this gap between measurements and their meanings with causal structure instead, reading a variable's semantics from how it acts on other variables. We formalize this as structure-constrained semantic alignment, in which the embedding of each unnamed variable is solved under the dependence relations implied by the causal graph, with the embeddings of a few known names as anchors. Accordingly, we build CausalBridge, a framework that discovers the causal graph from the measurements, latent variables included, solves for the embeddings under those relations, and expresses them as names through a language model. The causal structure reflects the mechanism that generated the measurements and is recovered from the measurements alone, which may make it the one source of information free of bias from human knowledge. We evaluate CausalBridge on five questionnaires and three robotics scenarios, with 20 to 90% of the variable names masked. It recovers the semantics of observed and latent variables more accurately than existing methods that rely on association, and its lead widens as less of the system is documented. The graph it discovers names variables as accurately as the documented one, and a new system is named in minutes and at a fraction of the cost of sampling methods. Once the semantic gap is bridged faithfully, machines can understand the world and take actions causally.

Data selection is already a central bottleneck in large-language-model training, where web-scale corpora are noisy and token budgets are finite. In continual pre-training (CPT), it becomes a forgetting-control problem: a poorly chosen target-domain corpus can overwrite capabilities encoded in the pretrained checkpoint. Existing CPT practice either scores candidates with parameter-agnostic scalars such as perplexity, or mitigates forgetting by spending many extra general-domain replay tokens. Neither strategy directly asks how training on a candidate will move the model parameters. We show that loss-based selection causes the post-CPT Fisher diagonal to drift downward on exactly the high-Fisher coordinates the pretrained model had committed to, while leaving low-Fisher coordinates largely untouched. This asymmetry exposes a parameter-space mechanism for catastrophic forgetting. Motivated by this observation, we propose a Fisher-aware CPT selector that decomposes each candidate's gradient into an anchor component, which measures perturbation along committed parameter directions, and a frontier component, which measures update capacity in unconstrained low-Fisher subspaces. We aggregate these signals with a log-determinant submodular objective and optimize it in a single pass using a scalable streaming data selection pipeline. On TinyLlama-1.1B and Llama-3.1-8B CPT over medical data, our selector improves target-domain quality while bounding forgetting on held-out pretraining benchmarks. Most importantly, it is substantially more token-efficient than forgetting-aware replay. 1B selected tokens already outperform the replay strategy trained with 10B tokens on both adaptation and forgetting, giving a 10x token-efficiency advantage.

Research on information operations has focused primarily on the coordinated accounts involved in such campaigns and in the content they promote. Bad actors manage the activity of inauthentic accounts through programmatic interfaces (APIs), but the role of this underlying infrastructure in enabling and uncovering such coordinated behaviors remains largely unexplored. This study investigates how the Twitter (now X) API was used to coordinate influence operations on the platform. Since API access is managed through developer apps, we extract app metadata from 43 operations taken down between 2018 and 2021 and apply three complementary computational approaches to study diverse forms of manipulation. Our analysis reveals several recurring patterns: API-as-a-service infrastructure reused across campaigns, gaming apps performing unauthorized engagement without user consent, spoofed app names to hide activity origins, and synchronized app networks indicating centralized control. Adding app usage as a coordination indicator provides a modest but consistent improvement in detection. These findings suggest that social media APIs play an important role in enabling online manipulation, and that the removal of app-level metadata by Twitter/X has eliminated an important tool for platform transparency and moderation.

Agents are increasingly given open-ended research tasks: discovering an empirical law from self-designed experiments, improving a heuristic whose optimum nobody knows, or beating a standing record. Their execution logs record every step of this research, yet the runs are still judged by their outcome score. That score alone does not establish whether an agent's claims follow from executed experiments, and a reference answer may be unavailable. We evaluate the agent's epistemic process: how it forms hypotheses, tests them, and revises them in response to evidence. We introduce OEB (Open-Endedness Bench), a benchmark-agnostic methodology that reads only the agent's execution record and never a reference answer or an outcome score. OEB compiles the record into a unified epistemic event graph whose edges connect the propositions the agent states to the executed actions that test them; each node carries an exact excerpt that code verifies against the record. One principle governs scoring: prose can state a proposition, but only evidence returned by an executed action can support or refute it, so OEB checks what the agent writes against what it actually ran. From the graph, OEB scores four competence axes (evidence, experiment, revision, and no reward hacking), mostly as the share of opportunities for sound research that the agent took, and profiles six subjective persona traits that describe the agent's research habits. We score 119 existing runs over 12 tasks from three benchmarks: LLM post-training, chip design, and a training-speed record. Against logged results, only 16-29% of the improvements agents claim are real. On 9 of 10 tasks, the best run tries more new ideas in its second half than the worst run. The persona readings follow the model: for every trait, the model that ran explains more of its variance across runs than the task (a median of 43% against 7%).

A typed decision model answers a fixed question about an input by returning a probability for each of several caller-defined options. Each option carries a short label and a written definition, which is where a developer states the rule the model should apply. Jev introduced this interface for routing, moderation and triage, open implementations followed, and the same operation occurs whenever a language model is used as a classifier by scoring label strings. We study the open implementations, whose weights we can inspect and patch, and ask whether the probability follows the definitions or the labels. A preference for the label we call option-label bias. Across four open-weight typed decision models, three ways of reading an answer from a Qwen2.5 backbone, eleven classification tasks and PolicyBench, a synthetic routing suite we introduce in which the rule appears only in the definitions, the answer is mostly the labels. Deleting every definition leaves accuracy unchanged (laya-td: 0.8559 against 0.8487), although those definitions support 0.7971 on their own, and renaming the options to A and B raises accuracy by +0.1511 [+0.1377, +0.1646]. One system, von, is unaffected, and the two code bases differ in one expression: laya writes each option as "{label}: {definition}", while von writes only the definition. Changing that expression in both directions, with no weight changed, makes all three laya checkpoints exactly invariant (+0.0000 [+0.0000, +0.0000]) and creates the effect in von, whose accuracy falls from 0.8511 to 0.2281 when a label contradicts its definition. Earlier work attributed this failure to the constrained decision head these models use in place of a text decoder; our results locate it in the prompt rendering. We give a two-call test that tells a practitioner which case applies to their model, and measure what four mitigations are worth.

Sparse Mixture-of-Experts (MoE) models combine learned routing with selection from a large expert pool. We isolate the contribution of dynamic expert combination using a dense analogue: $K$ SwiGLU experts, all active for every token and combined by a softmax gate, at fixed total FFN width. With no larger pool or discrete selection, the dense baseline is the $K=1$ case. Validation loss varies non-monotonically with $K$: $K=2$ improves over the baseline by $0.0048$, whereas $K=4$ and $K=6$ worsen it by $0.0053$ and $0.0197$, respectively. Routing generally remains soft and expert usage balanced, except in the first layer of the $K=4$ model, where routing is nearly one-hot. Forcing this gate to uniform increases loss by $2.4$ nats on a diagnostic subset, indicating that its concentrated routing is functionally important. We further show that the architecture is exactly a dense SwiGLU with token-dependent, simplex-constrained group scaling and contains the dense baseline in its function class. These single-run results suggest a granularity sweet spot for always-active, softly gated FFNs at fixed width, with $K=2$ performing best among the configurations tested.

Metropolitan concentration has motivated policies for more balanced regional development. We use agent-based simulation to examine trajectory diversity, the variety of people's recurring activity orientations toward neighborhoods, urban centers, and wider networks. The model maps 2010 population shares to 2026 distributions in five country cases, China, Russia, Japan, the United Kingdom, and the United States, and simulates change to 2036. Trajectory diversity declines by 52.6-79.0% over 2026-2036. Russia and Japan show the largest increases in the share of the most common trajectory, while US diversity contracts despite a slight decline in this share. Cost burdens rise throughout, whereas welfare falls in the Russian and Japanese cases but rises in the other three. These findings suggest that regional development should be assessed through activity diversity, cost burdens, and welfare together.