Science & Mathematics

In data-rich enterprises the hard question is not whether a candidate dataset is good in isolation, but whether it adds information beyond what the organization already holds, and whether that increment is worth its price. We present the Data Differentiator Framework (DDF), a composite scoring method for candidate data assets that combines (i) a mutual-information-based redundancy penalty against the existing portfolio, (ii) a cost-normalized, significance-gated marginal utility score derived from an ablation study, and (iii) explicit governance inputs (trust, identity resolution quality, strategic weights) set by the business. The framework is executed by a staged pipeline, the Strategic Data Sourcing Agent Ecosystem (SDSAE), that short-circuits expensive model-retraining steps for assets failing governance or redundancy gates. Conceptually, the DDF extends the relevance-minus-redundancy principle of mRMR feature selection from the feature level to the asset level, and adds cost normalization, a statistical significance gate, and governance terms that per-sample valuation methods such as Data Shapley do not model. We validate the framework on the public credit-card fraud benchmark (284,807 transactions) with five constructed candidate assets spanning the redundancy spectrum. Only the genuinely differentiated asset passes the significance gate ($Δ$AUC = +0.0029, DeLong $p = 1.4\times 10^{-9}$, Holm-corrected $p = 7.2\times 10^{-9}$) and receives an acquire recommendation (DDS = +0.663); the fully redundant asset scores lowest (DDS = -0.135). The recommendation is stable across 19 of 22 admissible strategic-weight configurations and across all decision thresholds in [0.15, 0.66]. We state clearly which quantities in the case study are computed from data and which are simulated governance inputs, and we discuss the design limits of a single-dataset evaluation.

Coding agents now complete tasks autonomously, modifying real-world state such as artifacts, software environments, and configurations along the way. Yet they show limited backup awareness: recognizing that an action will modify state that may later be needed, and preserving it beforehand. To evaluate backup awareness, we introduce RegretBench, a benchmark of 205 executable tasks grounded in regret moments reported on platforms such as X.com, GitHub, and Stack Overflow, where users explicitly regretted not preserving state they later needed. Across six frontier models, agents complete 86.1% of tasks on average yet back up before these moments in only 15.2%, showing that task completion ability does not imply backup awareness. Because it is hard to predict which changes will be regretted, we propose BackUpper, a harness plug-in backup layer that checks every tool call and preserves affected state before modification. It combines rule-based strategies for common operations with a lightweight subagent for complex cases, and its rule table can expand with subagent decisions over time. On RegretBench, BackUpper covers every regret moment action and restores states with 92.2% average similarity to the original. Achieving this costs only $0.014 per task, with 7x fewer tokens than the main-agent self-backup baseline. When users have preferences on what to back up, BackUpper can also narrow its scope accordingly, matching human annotations on whether an action is worth preserving with an 84.5% F1 score. These results suggest that agents can keep actions reversible at low cost while respecting users' preferences.

Conformal prediction (CP) for language models is typically evaluated under a matched protocol, in which calibration and deployment use the same post-training inference configuration, such as quantization mode and inference temperature. In deployment, these choices often differ from those used at calibration to meet memory or throughput constraints or serving policies. We argue that this exposes a limitation in current CP evaluation protocols for language models. We formalize the underlying phenomenon as inference-time model shift: a change in the conformal score distribution induced by inference configuration rather than by the input--label distribution, and theoretically characterize how it breaks split-conformal exchangeability even when the task distribution and trained model are unchanged. We then evaluate its consequences across 8 language models, 4 finite-label benchmarks, and 12 inference configurations. We find that matched evaluation can substantially overstate deployment-time coverage, and that the effect is heterogeneous across models and datasets. We argue that inference configuration should be reported as a first-class experimental axis in CP evaluations for language-model systems.

Stateful dataflow systems must decide where state lives relative to the tasks that process it. For long-running applications that require frequent reconfiguration due to changes in resource allocation, partitioning, or task placement, this design choice presents a core trade-off between adaptability and performance. Co-locating state with computation ensures low latency at steady state but makes reconfiguration disruptive, as any change forces state migration. On the other hand, keeping state in an external store makes reconfiguration seamless, but incurs overhead on every access. We present Koala, a reconfiguration protocol for distributed dataflow systems that retains the operational performance of embedded-state architectures while matching the disruption-free adaptability of external-state designs. Unlike prior protocols which relocate all state regardless of whether it will soon be accessed, Koala migrates state in a fine-grained, lazy manner. To achieve that, it employs a lightweight ownership transfer mechanism that safely decouples event routing from state location. We implement Koala in a new streaming dataflow system and evaluate it on real-world and synthetic benchmarks. When compared with the state-of-the-art protocol DRRS, Koala consumes backlog up to 23x faster and transfers 8.7x less state on average during reconfiguration.

Voice style conversion (VSC) aims to transform a source utterance to match the timbre, accent, and emotion of a target voice while preserving linguistic content. StyleStream 1.0 introduced the first streamable zero-shot VSC system with state-of-the-art conversion quality, but remained limited in two ways: approximately 1s end-to-end latency on consumer hardware, and the lack of a text-based interface for designing or editing target voices. We present StyleStream 2.0, a fast and controllable streaming VSC framework that addresses these limitations. To reduce latency, we train a few-step pixel MeanFlow model and further fine-tune it with autoregressive feedback inspired by Self Forcing, improving robustness under streaming inference. This reduces end-to-end latency to 520~ms on a consumer GPU, a 2x speedup over StyleStream 1.0, while maintaining conversion quality. To enable controllable style manipulation, we remove the mel-spectrogram context and instead route target style information through a compact style embedding. In this embedding space, we train a unified text-conditioned flow matching model that supports both text-based voice design, which maps natural language prompts to voice styles, and instruction-based voice editing, which modifies a specific attribute of a source utterance while preserving the rest. Experiments show that StyleStream 2.0 achieves the strongest target style fidelity on VSC, voice design, and voice editing, while remaining competitive on intelligibility and non-edited attribute preservation.

Clothes-Changing Person Re-Identification (CC-ReID) remains challenging due to drastic appearance shifts across outfits. Most prior works devise biometric signal-extraction heuristics to enhance inter-person variance, e.g., using segmentation masks, while leaving intra-person variance uncontrolled. We present MINER, a simple model-agnostic framework that suppresses garment bias by contrasting hard examples and adaptively balancing dataset-dependent intra- and inter-person variance. Extensive experiments show that MINER improves retrieval across CC-ReID benchmarks, achieving state-of-the-art performance on four datasets and ranking among the leaders in all three tracks of the open TVRID RGB-Depth ReID challenge.

Protein engineering often requires improving several properties simultaneously, motivating the search for high-performing sequences that represent useful fitness tradeoffs. Discovering these sequences is challenging because the design space is combinatorially large and experimental campaigns often permit only a few rounds of batched measurements. We introduce \textit{MODS}, a modular framework for multi-objective protein optimization via inference-time diffusion steering. For each proposal, MODS samples augmented Chebyshev scalarization weights and a fitness-surrogate ensemble member to define a scalar steering reward. Varying the weights directs generation toward different objective tradeoffs, while varying ensemble members diversifies the predictions used for guidance. Because MODS expresses multi-objective acquisition through scalar rewards, it can be paired with different diffusion models and compatible steering procedures. After each evaluation batch, MODS updates the surrogates without retraining the generative model. We evaluate MODS on six protein optimization tasks using predictive oracles trained on experimental assay data, with two diffusion models and two steering procedures. MODS achieves competitive final hypervolume across all six tasks with both diffusion models, and substantially outperforms the best evaluated baseline on several tasks. Our results demonstrate a flexible and effective approach to multi-objective protein optimization with limited feedback.

Semantic leakage occurs when a concept in the prompt steers model generation about an entity, even though the concept is irrelevant to it (e.g., a person who likes yellow is said to work as a school bus driver). It has mostly been studied in settings where the concept is directly attributed to the entity the model is asked about. This leaves open whether leakage is a flawed heuristic, where the model builds a profile of that entity, or a much broader associative tendency, e.g., working across seemingly unrelated entities. We argue for the latter: even after removing the attribution of the concept to the specific entity, leakage persists. Next, we extend this idea to gender bias. While work on gender bias in LMs typically looks at cases where the stereotyped concept is attributed to an entity, our findings suggest that a concept may steer outputs with its mere presence in the scene. We show this using gendered concepts under two conditions: when the subject engages with the concept, and when the concept just sits in the scene. We find that both shift model generation toward the stereotype to a similar extent. Together, our results suggest that semantic leakage drives a form of bias that is overlooked by current evaluations.

Large language model (LLM) agents are increasingly expected to interact with environments and improve their decisions during deployment. However, how anticipating future outcomes and drawing on accumulated interaction experience can jointly support decision-making remains underexplored. We propose a test-time framework, inspired by Model Predictive Path Integral (MPPI) control, that combines multi-branch predictive analysis with experience-based adaptation. At each step, the agent proposes multiple candidate actions and imagines their outcomes with a world model, so that alternatives can be compared without being executed in the environment. Since language environments provide neither explicit dynamics nor costs, the predicted outcomes are evaluated for task progress and risk, and MPPI-style weights over the resulting costs guide the final decision. A dual-branch experience memory enables self-improvement: the planner branch refines candidate generation and action selection, while the world model branch improves next-observation prediction. Experiments on ALFWorld and ScienceWorld across three backbones demonstrate the strong overall performance of our method and the effectiveness of experience-guided predictive control. Our code will be made publicly available.

NVIDIA tensor cores have supported 2:4 semi-structured sparsity since Ampere, and the pattern composes with quantization. When the two are combined, the practical question is which step to take next: from a given precision, add 2:4 sparsity or drop to the next lower precision? We compare the two at equal bits, with sparsity metadata counted, and at equal nominal tensor-core tiers, over all six GEMM operands of a transformer layer, with kernels measured on three GPU generations and online pruning and quantization included. Our results show that as long as a lower precision is available, adding 2:4 is the worse step, both less accurate and slower, because the energy that 2:4 discards from each group of four along the hardware-forced axis does not shrink with bits, while quantization error does. Once quantization reaches the lowest precision the hardware supports, 2:4 still speeds up linear layers by 1.3--1.7\x{}, and on an H100 a retrained 2:4 model served in FP8 by vLLM reaches 27\% higher prefill throughput than the dense FP8 model, with GSM8K accuracy on par. Its quality has to be won back by retraining, however, which restores perplexity long before reasoning; on Blackwell, whose sparse FP4 uses a pairwise 4:8 pattern, the loss is larger and our retraining budget did not recover it.

Mobile manipulators must adapt not only to objects, but also to the spatial context in which they appear: where to position the base, how to manipulate, and where to look while navigating surrounding scene geometry. To address this, we present COWBOY, a scalable sim-to-real framework for learning such contextual whole-body manipulation. On whole-body dexterous grasping, COWBOY transfers zero-shot to hundreds of objects and diverse in-the-wild scenes, spanning cluttered tabletops, shelves, drawers, and other unstructured settings. Rather than learning scene adaptation directly with RL in simulation, a vectorized environment-aware whole-body controller (WBC) handles reaching in thousands of scenes by coordinating base-arm motion, collision avoidance, and active vision. Then, local RL experts are trained in local scenes for object interaction and can be reused across thousands of full-scene layouts. We distill these composed WBC-RL teachers into a single vision policy that controls the mobile base, arm, dexterous hand, and actuated camera from point-cloud observations. The resulting policy exhibits contextual whole-body behavior, automatically adapting base placement, camera viewpoint, arm trajectory, and grasp strategy to observed scene and object geometry. Videos are available at https://cowboy-manip.com

Safety guards check content before an LLM application acts on it, and most paste that content into a plain-text prompt. Content that imitates the prompt can append a second task, and the guard answers that one instead. We call this target rebinding. Our explanation is that the guard grades the last complete task in its input, here a harmless decoy, so the harmful request goes through unchanged. In a pre-registered test, an appended forged query and decoy clear all 75 held-out AdvBench requests that Shieldstral-1.0-3B flags. The guard is not persuaded: a directive to ignore the text above clears none of them, and in a follow-up run a harmful decoy stops the attack. On our GuardRebind benchmark, each of five guards has an attack family, chosen per guard, that clears at least 93% of its flagged items. Rebinding explains this best on Shieldstral and gpt-oss-safeguard; on ShieldGemma and Llama Guard 3, generic evasion and trailing text respectively do part of the work, and Qwen3Guard falls mainly to a forgery of its whole task. Training the guard to score content no lower than its most harmful part cuts attack success from 75% to 0.8-1.6% on attack families, decoys and forged queries held out of training. All of these attacks append the forged structure after the request, and we added the held-out-decoy evaluation after a first round of test runs. Clean AUROC falls by at most 0.01, but our test of non-inferiority is inconclusive. A score-guided search still beats this trained guard on 47 of 50 items, usually by placing forged structure before the request. On about half of the items where the search wins, the same search also finds benign text, with no forged structure, that clears the guard.

Conventional wafer defect analysis relies on offline, closed-world classifiers that fail to detect novel, unseen defect signatures emerging during continuous semiconductor manufacturing. To address this, we propose WaferHDC, an online open-world wafer defect discovery framework based on hyperdimensional computing (HDC). The memory-centric architecture of WaferHDC enables rapid anomaly detection and grouping of recurring failures into compact hypervector representations. This allows the memory space to dynamically scale with new defect modes without retraining a static classifier or archiving entire raw wafer history. WaferHDC first models normal spatial distributions, then dynamically groups recurring failures into persistent defect memory without labels or predefined class counts. It outperforms the strongest online baseline by 13.3 p.p. in adjusted Rand index and 10.5 p.p. in normalized mutual information on the industry-standard WM-811K dataset. These results demonstrate the effectiveness of WaferHDC for continuous and open-world wafer defect analysis.

We present HyperDrone, an extremely fast quadrotor simulator focused on high-performance batch rendering that is more than 20x faster than existing state-of-the-art batch quadrotor simulators. HyperDrone combines highly optimized multirotor dynamics with HyperLux, a new ray-tracing engine that supports Nvidia OptiX, Vulkan, Apple Metal and WebGPU backends that takes full advantage of modern ray-tracing hardware, capable of running both components entirely on the GPU. HyperLux renders RGB, depth, semantic/instance segmentation, surface normals and optical flow observations at configurable levels of fidelity, including reflections, realistic lighting, motion blur, anti-aliasing, and Physically Based Rendering (PBR). Even in the highest-fidelity settings, HyperLux achieves millions of Frames Per Second (FPS) on consumer-grade GPUs while rendering realistic 3D scenes. We demonstrate the merits of HyperDrone by training a vision-based yaw-control policy purely in simulation and transferring it to a real small-scale quadrotor (Crazyflie) with a camera, running completely onboard.

Autonomous mobile robots require spatial representations that capture geometry and semantics while staying sparse enough for planning in large, unknown environments. Graphs provide a natural representation for this purpose and underpin both classical planning methods and a growing class of reinforcement-learning (RL) policies. However, RL training needs hundreds to thousands of parallel environments, each constructing and updating its own graph. Existing graph libraries are typically designed for either single-robot deployment or fixed graphs, and CPU-based graph construction limits scalable training to only a few dozen environments. We introduce GGraph, an open-source GPU-accelerated graph library for both robot deployment and large-scale learning. GGraph supports real robots through ROS 2 interfaces and vectorized GPU-based policy training, with built-in support for tasks including 3-D navigation, exploration, and loop-closure integration. For single-robot operation, GGraph updates graphs in milliseconds. At scale, it enables graph-based exploration policies to train across hundreds of parallel robots, reducing wall-clock training time by one order of magnitude. Across simulation and real-world experiments, these results establish GGraph as a unified graph infrastructure for both large-scale robot learning and deployment.