arXiv:2605.16208v1 Announce Type: cross
Abstract: Flexible continuous-time survival modeling is critical for capturing complex time-varying hazard dynamics in high-dimensional data; however, training such models remains challenging due to the intractable integral required for likelihood estimation. We introduce QSurv, a scalable deep learning framework that enables nonparametric continuous-time modeling without relying on time discretization or restrictive distributional assumptions. We propose a training objective based on Gauss-Legendre numerical quadrature, which approximates the cumulative hazard with high-order accuracy while facilitating efficient end-to-end training via standard backpropagation. Furthermore, to effectively capture non-stationary hazard dynamics in complex architectures, we introduce time-conditioned low-rank adaptation, a mechanism that conditions general neural backbones on time by dynamically modulating weights via low-rank updates. We provide theoretical analysis establishing approximation error bounds for cumulative-hazard evaluation. Comprehensive experiments across synthetic benchmarks, large-scale real-world tabular datasets, and high-dimensional medical imaging tasks demonstrate that QSurv achieves competitive predictive performance with advantages in instantaneous hazard function estimation, enabling more interpretable characterization of time-varying risk patterns.
Science Journals
arXiv:2605.15579v1 Announce Type: cross
Abstract: To fit diverse display and bandwidth constraints, high-frame-rate videos are temporally downscaled to low-frame-rate (LFR) and later upscaled, requiring joint optimization for effective frame-rate rescaling. However, existing methods typically link the two operations via training objectives, without fully exploiting their reciprocal nature, which may cause high-frequency information loss. Moreover, they overlook the impact of lossy codecs on LFR videos, limiting real-world applicability. In this work, we propose an end-to-end framework for compression-aware frame-rate rescaling, named TVRN. To regularize high-frequency information lost during frame-rate downscaling, TVRN adopts an invertible architecture that combines a Multi-Input Multi-Output Temporal Wavelet Transform with a high-frequency reconstruction module. To enable end-to-end training through non-differentiable lossy codecs, we design a surrogate network that approximates their gradients. Finally, to improve robustness under various compression levels, we extend TVRN to an asymmetric architecture by incorporating compression-aware features learned via a learning-to-rank strategy. Extensive experiments show that TVRN outperforms existing methods in reconstruction quality under industrial video compression settings. Source code is publicly available at https://github.com/fengxinmin/TVRN_public.
arXiv:2605.15656v1 Announce Type: cross
Abstract: Under the trend of multi-waveform coexistence in 6G IoT, intelligent receivers must first identify physical-layer waveform types before performing correct demodulation and resource scheduling. However, existing signal identification research largely focuses on symbol-level modulation classification. Research directly targeting physical-layer waveform types (e.g., OFDM, OTFS, LoRa) is not only extremely scarce but also heavily reliant on deep neural networks and complex time-frequency transforms, making deployment on resource-constrained terminals difficult. Symbol modulation classification methods themselves cannot circumvent the prerequisite of ``waveform identification first.''
To address this dual gap, we propose an ultra-lightweight waveform classification framework based on time-frequency multidimensional features with a cooperative Z-test tree (ZTree). The framework employs low-complexity time-domain feature extraction, and the classification backend adopts a ZTree optimized by Z-statistical testing, which uses hypothesis testing confidence to automatically control decision tree splitting and size, ensuring efficient execution on resource-limited processors. Tested on ten 6G candidate waveforms including OFDM, OTFS, DSSS, LoRa, and NB-IoT, the method achieves 99.5\% average accuracy under AWGN and 87.4\% under TDL-C multipath channels, with main confusion between OTFS and LoRa. Implemented in C on an x86 platform, single inference latency is under 4~ms. To the best of our knowledge, this is the first work achieving real-time recognition of ten IoT waveform types. Future work will target deployment acceleration on embedded MCUs. Code and dataset are open-sourced at: https://github.com/Einstein-sworder/IoT-wave.
arXiv:2605.15642v1 Announce Type: cross
Abstract: Nuclear-powered submarines are difficult to track with conventional methods in congested waterways. We revisit antineutrino-based detection as a barrier concept, analogous to a neutrino-enabled SOSUS-style fence in strategic straits. Using analytic scaling relations and numerical estimates, we show that detectability depends primarily on closest approach, detector depth, and deployed mass. For representative assumptions, a 20\,kt detector in the Strait of Gibraltar reaches a local benchmark score $Z_A\simeq2.54$ for an assumed 100\,MW thermal-power sensitivity-study case in a conservative worst-case transit (with Poisson operating point $(P_\mathrm{FA},P_\mathrm{det})\simeq(5.5\times10^{-3},0.51)$ at threshold $k=2$), while a three-detector line raises the mapped score to $Z_A\simeq4.66$. For broad ocean passages such as GIUK, required detector counts are substantially larger; in the baseline maximum passing distance $\mathrm{PDD}_{\max}=5$\,km geometry, about 80 detectors yield only $Z_A\sim1.6$. The paper outlines detector technology choices, statistical assumptions, and deployment constraints for a first-generation feasibility assessment.
arXiv:2602.06239v3 Announce Type: replace
Abstract: We introduce PEPO (Pessimistic Ensemble based Preference Optimization), a single-step Direct Preference Optimization (DPO)-like algorithm to mitigate the well-known over-optimization issue in preference learning without requiring the knowledge of the data-generating distribution or learning an explicit reward model. PEPO achieves pessimism via an ensemble of preference-optimized policies trained on disjoint data subsets and then aggregates them through a worst case construction that favors the agreement across models. In the tabular setting, PEPO achieves sample complexity guarantees depending only on a single-policy concentrability coefficient, thus avoiding the all-policy concentrability which affects the guarantees of algorithms prone to over-optimization, such as DPO. The theoretical findings are corroborated by a convincing practical performance, while retaining the simplicity and the practicality of DPO-style training.
arXiv:2507.14200v2 Announce Type: replace
Abstract: Existing multi-LLM collaboration systems often encounter scalability challenges when integrating new LLMs and tasks, leading to suboptimal performance. To address this, we propose SMCS, a Scalable Multi-LLM Collaboration System designed to effectively coordinate multiple open-source LLMs. The system consists of two core components: a Retrieval-based Prior Selection (RPS) module, which dynamically selects the most suitable LLMs for each input, and an Exploration-Exploitation-Driven Posterior Enhancement (EPE) module, which fosters response diversity and selects high-quality outputs through a hybrid scoring mechanism. Experiments on eight mainstream benchmarks validate the effectiveness of our system: by integrating fifteen open-source LLMs, SMCS outperforms prevailing closed-source LLMs, e.g., GPT-4.1(+5.36%) and GPT-o3-mini(+5.28%) across multiple tasks. Remarkably, it even exceeds the average of best results on different datasets with open-source LLMs (+2.86%), significantly advancing the empirical performance frontier of open-source collaboration. The code is released at https://github.com/magent4aci/SMCS.
arXiv:2605.15381v1 Announce Type: cross
Abstract: Non-bonded interactions govern structure, stability, and function across a wide range of solid-state materials, yet their chemical origins are often difficult to resolve from total energies alone. Here we generalize absolutely localized molecular orbital energy decomposition analysis to quantify and interpret non-bonded interactions within and between solids at the density functional theory level. Across molecular crystals, moir\'e heterobilayers, and layered perovskite heterostructures, this framework separates lattice-formation energies, interlayer binding energies, and band-structure changes into chemically intuitive contributions from frozen interactions, polarization, and charge transfer. The analysis reveals how dispersion controls polymorph stability in pharmaceutical crystals, how local stacking modulates interlayer coupling in MoS2/WSe2, and how alkali-cation substitution switches the quantum-well character of layered perovskite heterostructures. By connecting emergent solid-state properties to microscopic interaction mechanisms, this framework provides a chemically transparent basis for understanding and designing complex materials.
arXiv:2605.15808v1 Announce Type: cross
Abstract: Integrated sensing and communication (ISAC) relies on monostatic sensing (MS) and bistatic positioning (BP) to enable comprehensive environmental awareness and user localization. However, existing frameworks predominantly assume static geometries and optimize these modalities independently, neglecting user mobility and sequential information sharing. In this paper, we propose a velocity-aware sequential beamforming framework that dynamically couples MS and BP in time. We derive the Cramer-Rao bounds (CRBs) in the position domain to formulate a non-convex resource allocation problem. Instead of relying on static weighted-sum tradeoffs, we introduce a sequential Bayesian optimization strategy where MS is executed first to construct a reliable structural prior on the UE and passive targets (PTs). This covariance prior is subsequently passed to the UE to regularize the BP estimation stage. We demonstrate that optimizing a single shared beamformer globally across both phases yields superior synergistic gains compared to a two-stage greedy approach. Simulation results validate that the shared sequential design efficiently balances limited symbol resources, achieving centimeter-level positioning accuracy for both the UE and PTs, robust velocity estimation, and a significantly reduced computational runtime.
arXiv:2605.13246v2 Announce Type: replace
Abstract: Reference counting bugs in Linux kernel drivers can lead to severe resource mismanagement and security vulnerabilities. We introduce DrvHorn, a novel automated tool to detect these bugs by reducing reference counting verification to an assertion checking problem leveraging the Linux driver interface. Through efficient modeling of the Linux kernel and aggressive program slicing, DrvHorn discovered 545 bugs, of which 424 were previously unknown, across all platform drivers in v6.6 Linux kernel, with a lower false positive rate of 29.9% compared to prior studies. To address the root causes of these newly discovered bugs, we submitted patches to the Linux kernel, and 45 of them were merged.
arXiv:2605.13788v2 Announce Type: replace
Abstract: Active learning for machine-learning interatomic potentials (MLIPs) must address several challenges to be practical: scaling to large candidate pools, leveraging energy-force supervision, and maintaining robustness when candidate pools are biased relative to the target distribution. In this work, we jointly address these challenges. We first introduce a linearly scaling acquisition framework based on chunked feature-space posterior-variance shortlisting. By avoiding materialisation of the candidate and train set kernels, this approach enables screening of ~200k structures within hours and applies broadly to acquisition strategies that score candidates based on molecular similarity metrics. We then extend the Neural Tangent Kernel (NTK) to a force-aware setting via mixed parameter-coordinate derivatives, yielding a force NTK and a joint energy-force NTK that provide natural similarity metrics for vector-field prediction. We demonstrate the effectiveness of the joint energy-force NTK on the OC20 dataset, where force-aware acquisition is crucial: it achieves the lowest energy and force MAE and RMSE across all metrics and distribution splits. Across T1x, PMechDB, and RGD benchmarks, our force NTK methods remain competitive with established baselines while being significantly more efficient than committee-based approaches. Under a controlled candidate-pool shift case study on T1x, acquisition based on pretrained MLIP embeddings and NTKs remains robust, whereas committee-based methods exhibit higher variance. Overall, these results show that a single pretrained MLIP can enable scalable, force-aware, and distribution-robust active learning for foundation-model fine-tuning.
arXiv:2605.15823v1 Announce Type: cross
Abstract: Researchers and practitioners in the field of reliability engineering and optimization frequently use active redundancy techniques to intensify the performance of systems. In this article, we study allocation strategies of non-matching active redundancies (spares) in coherent systems consisting of possibly dependent and identical components for achieving better system reliability. The dependence of the components is modeled through copulas using the distortion function. Sufficient conditions are derived to establish optimal allocation strategies for two heterogeneous active redundancies at the component or system levels. Moreover, the results are true for the component lifetimes following a general family of parametric distributions. The results guarantee the likelihood ratio (reversed hazard) ordering between the coherent systems at the component level (system level) active redundancies. Some aging properties are also established in this endeavor. Several examples are provided to demonstrate the theoretical results.
arXiv:2605.15839v1 Announce Type: cross
Abstract: Turing patterns are a cornerstone of biological self-organization, yet their emergence typically requires finely tuned parameters occupying narrow regions of high-dimensional space. This poses a fundamental challenge: how can evolving biological systems reliably find and exploit such rare conditions? In this work, we propose that common biochemical limit cycles, such as those arising from genetic feedback loops, can act as natural explorers of Turing space. By coupling a reaction-diffusion system to an orbit that modulates some of its parameters, we show that the system can dynamically sweep through Turing-permissive regimes and generate transient spatial patterns. We use an entropy-based measure in Fourier space to quantify pattern formation and demonstrate how cycles enhance the detectability and robustness of Turing islands. We further explore how coupling to positional gradients increases reproducibility, suggesting a route from oscillatory dynamics to stable developmental programs. Our results highlight a powerful mechanism by which nature might bootstrap complex spatial structure from simple temporal motifs.
arXiv:2605.15869v1 Announce Type: cross
Abstract: The quantum Internet relies on the ability to distribute entangled quantum bits (ebits) between quantum memories at the end nodes, to perform applications like blind or distributed quantum computing that are impossible if end nodes are connected via a classical, i.e., non-quantum network. This need creates new challenges due to the fragile nature of entanglement, which decoheres over short timescales and cannot be amplified, buffered, or retransmitted. Two broad categories of approaches have been proposed in the scientific literature to realize such an entanglement distribution in a given path: one relying on a synchronous time-slotted model, and another one where intermediate nodes interact asynchronously. However, both of them implicitly assume a serial operation, where one ebit is established and made available to the application on end nodes before creating a new one. This is inefficient in long-range networks, with high transmission latencies, if the intermediate nodes have multiple memory qubits that could be used in parallel. To overcome this limitation, in this paper, we study the implications of multiplexing concurrent ebit requests on the same quantum, for both synchronous and asynchronous operation. Furthermore, for the latter, we define a novel distribution protocol, called HOPPER, where the intermediate nodes make autonomous and hop-by-hop decisions on the use of their local resources when establishing an ebit. With numerical simulations, we show that HOPPER is effective in handling multiple ebit requests in parallel, and it exhibits significantly better performance than a synchronous alternative in different scenarios.
arXiv:2512.19053v2 Announce Type: replace-cross
Abstract: A detection method for high-frequency gravitational waves using two-dimensional ion crystals is investigated. Gravitational waves can resonantly excite the drumhead modes of the ion crystal, particularly the parity-odd modes. In the optical dipole force protocol, entanglement between the drumhead modes and the collective spins transfers the excitation of the drumhead modes to the rotation of the total spin. Furthermore, gravitational wave detection beyond the standard quantum limit becomes possible as a squeezed spin state is generated through this entanglement. The sensitivity gets better with a larger ions crystals as well as a larger number of the ions. Future realization of large ion crystals can significantly improve the sensitivity to gravitational waves in the 10 kHz to 10 MHz region.
arXiv:2403.13805v2 Announce Type: replace
Abstract: CLIP (Contrastive Language-Image Pre-training) uses contrastive learning from noise image-text pairs to excel at recognizing a wide array of candidates, yet its focus on broad associations hinders the precision in distinguishing subtle differences among fine-grained items. Conversely, Multimodal Large Language Models (MLLMs) excel at classifying fine-grained categories, thanks to their substantial knowledge from pre-training on web-level corpora. However, the performance of MLLMs declines with an increase in category numbers, primarily due to growing complexity and constraints of limited context window size. To synergize the strengths of both approaches and enhance the few-shot/zero-shot recognition abilities for datasets characterized by extensive and fine-grained vocabularies, this paper introduces RAR, a Retrieving And Ranking augmented method for MLLMs. We initially establish a multi-modal retriever based on CLIP to create and store explicit memory for different categories beyond the immediate context window. During inference, RAR retrieves the top-k similar results from the memory and uses MLLMs to rank and make the final predictions. Our proposed approach not only addresses the inherent limitations in fine-grained recognition but also preserves the model's comprehensive knowledge base, significantly boosting accuracy across a range of vision-language recognition tasks. Notably, our approach demonstrates a significant improvement in performance on 5 fine-grained visual recognition benchmarks, 11 few-shot image recognition datasets, and the 2 object detection datasets under the zero-shot recognition setting.
arXiv:2410.01990v3 Announce Type: replace
Abstract: This paper explores alternative formulations of the Kolmogorov Superposition Theorem (KST) as a foundation for neural network design. The original KST formulation, while mathematically elegant, presents practical challenges due to its limited insight into the structure of inner and outer functions and the large number of unknown variables it introduces. Kolmogorov-Arnold Networks (KANs) leverage KST for function approximation, but they have faced scrutiny due to mixed results compared to traditional multilayer perceptrons (MLPs) and practical limitations imposed by the original KST formulation. To address these issues, we introduce ActNet, a scalable deep learning model that builds on the KST and overcomes many of the drawbacks of Kolmogorov's original formulation. We evaluate ActNet in the context of Physics-Informed Neural Networks (PINNs), a framework well-suited for leveraging KST's strengths in low-dimensional function approximation, particularly for simulating partial differential equations (PDEs). In this challenging setting, where models must learn latent functions without direct measurements, ActNet consistently outperforms KANs across multiple benchmarks and is competitive against the current best MLP-based approaches. These results present ActNet as a promising new direction for KST-based deep learning applications, particularly in scientific computing and PDE simulation tasks.
arXiv:2411.10037v3 Announce Type: replace
Abstract: Effective usage of approximate circuits for various performance trade-offs requires accurate computation of error. MCAC is a novel model counting framework for exact computation of several average and worst-case error metrics that are used to evaluate approximate circuits. Unlike other methods in the literature, our framework uses the same error miter for all metrics. It requires a single synthesis of the system consisting of the exact and approximate circuits followed by a subtractor that finds the difference of the two outputs. Existing miter-based methods require multiple calls to the model counter, one for each output of the miter. MCAC uses the CNF formula of the system to compute all metrics. Our algorithm converts the formula to a tree and uses message passing to compute all metrics. We propose data structures to efficiently store and perform sparse computations required for conversion to a tree and message passing. Results for all the error metrics for several benchmark instances show a significant speedup over using off-the-shelf model counters along with specialized miters for each metric.
arXiv:2605.03258v2 Announce Type: replace
Abstract: Large language models often fail at simple counting tasks, even when items to count are in the prompt. We investigate whether this failure occurs because transformers do not represent counts internally, or because they cannot convert representations to the correct output tokens. Across three model families: Pythia, Qwen3, and Mistral, ranging from 0.4B to 14B parameters, we find evidence for the second explanation. Linear probes recover the correct count from intermediate layers with $R^2>0.99$, showing that the information is present. However, the internal directions that encode counts are nearly orthogonal to digit-token output-head rows ($|\cos| \leq 0.032$). In other words, the model stores the count in a form that the digit logits do not naturally read out. We localize this failure with two interventions. Updating only the digit rows of the output head (36,864 parameters) substantially improves constrained digit prediction (60.7--100.0% on four tasks), but it does not fix unconstrained generation (0%); we do not claim that digit-row repair fixes open-ended text. By contrast, small LoRA on attention Q/V (7.67M parameters) improves upstream routing and achieves 83.1%$\pm$7.2% in true greedy autoregressive generation (deployable fix). Logit-lens at layer 35 (entity counting; correct-digit rank): (i) median over 3 seeds drops from order-$10^4$ to 1; (ii) seed 42 shows $54{,}332 \to 838$ (median top-1 while one seed stays far below). Norm, logit-lens, and cross-task analyses generalize the bottleneck to counting, addition, and list length; nulls on MMLU and GSM8K and limited DROP transfer. These results identify counting failure as a geometric readout bottleneck, not an internal-representation failure: the model knows the count but the output pathway is misaligned with tokens needed to express it.
arXiv:2501.19128v5 Announce Type: replace
Abstract: In many real-world scenarios, reward signal for agents are exceedingly sparse, making it challenging to learn an effective reward function for reward shaping. To address this issue, the proposed approach in this paper performs reward shaping not only by utilizing non-zero-reward transitions but also by employing the \emph{Semi-Supervised Learning} (SSL) technique combined with a novel data augmentation to learn trajectory space representations from the majority of transitions, {i.e}., zero-reward transitions, thereby improving the efficacy of reward shaping. Experimental results in Atari and robotic manipulation demonstrate that our method outperforms supervised-based approaches in reward inference, leading to higher agent scores. Notably, in more sparse-reward environments, our method achieves up to twice the peak scores compared to supervised baselines. The proposed double entropy data augmentation enhances performance, showcasing a 15.8\% increase in best score over other augmentation methods
arXiv:2603.15273v2 Announce Type: replace
Abstract: Noble liquids, notably argon and xenon, are utilised as both detector media and as the detector target for dark matter and neutrino physics experiments. When the noble liquid is excited by particles, it scintillates vacuum ultraviolet light, which sensors then detect. A major focus of the detector development community is on producing precision light sensors for noble liquid detectors. We introduce a flash lamp to test VUV-sensitive light sensors with light at wavelengths observed at noble liquid detectors. This paper discusses the design and presents results from a flash lamp prototype operated at room temperature.
arXiv:2502.04871v2 Announce Type: replace
Abstract: The Landau-Lifshitz equation describes the dynamics of magnetization in ferromagnetic materials. Due to the essential nonlinearity and nonconvex constraint, it is typically solved numerically. In this paper, we developed a finite volume element method (FVEM) with the Gauss-Seidel projection method (GSPM) for the micromagnetics simulations. We provide the approximation error in space and depict the energy law when the FVEM is adopted. Owing to the GSPM for time-marching, the discrete system is decoupled component by component, making the computational complexity comparable to that of solving the scalar heat equation implicitly. This significantly accelerates real simulations. We present several numerical experiments to validate the theoretical analysis and the efficiency gain. Additionally, we study the blow-up solution and efficiently simulate the 2D magnetic textures using the proposed method.
arXiv:2503.02597v3 Announce Type: replace
Abstract: Recent Multimodal Large Language Models (MLLMs) have demonstrated significant progress in perceiving and reasoning over multimodal inquiries, ushering in a new research era for foundation models. However, vision-language misalignment in MLLMs has emerged as a critical challenge, where the textual responses generated by these models are not factually aligned with the given text-image inputs. Existing efforts to address vision-language misalignment have focused on developing specialized vision-language connectors or leveraging visual instruction tuning from diverse domains. In this paper, we tackle this issue from a fundamental yet unexplored perspective by revisiting the core architecture of MLLMs. Most MLLMs are typically built on decoder-only LLMs consisting of a causal attention mechanism, which limits the ability of the earlier modalities (e.g., images) to incorporate information from the latter modalities (e.g., text). To address this problem a MLLM that unlocks causal attention into our proposed modality-mutual attention (MMA) to enable image tokens to attend to text tokens. This simple yet effective design allows MMA to achieve state-of-the-art performance in 12 multimodal understanding benchmarks (+6.2% on average across 3 LLMs backbones) without introducing additional parameters. Our MMA design is intended to be generic, allowing for applications across various modalities, and scalable to accommodate diverse multimodal scenarios.
arXiv:2503.07518v2 Announce Type: replace
Abstract: Large Language Models (LLMs) rely on the Key-Value (KV) Cache to store token history, enabling efficient decoding of tokens. As the KV-Cache grows, it becomes a major memory and computation bottleneck. However, there is an opportunity to alleviate this bottleneck, prior research has shown that only a small subset of tokens contribute meaningfully to each decoding step. A key challenge in finding these critical tokens is that they are dynamic, and heavily input query-dependent. Existing methods either risk quality by evicting tokens permanently, or retain the full KV-Cache but rely on retrieving chunks of tokens and many existing KV-Cache sparsity methods rely on inaccurate proxies for token importance. To address these limitations, we introduce TokenButler, a high-granularity, query-aware predictor that learns to identify these critical tokens. TokenButler predicts low-dimensional importance queries at a fixed depth stride, and combines them with a learned projection of the real KV-cache keys to score tokens cheaply, enabling dynamic per-token selection under a fixed budget while preserving the full KV cache. We train TokenButler by distilling the model's masked causal attention distributions, optimizing a lightweight predictor with minimal parameter overhead. We evaluate TokenButler on a novel synthetic small-context co-referential retrieval task, demonstrating near-oracle accuracy where existing methods fail. Furthermore, TokenButler achieves competitive or superior performance on long-context benchmarks (RULER, LongBench), up to $\approx1.6\times$ on-GPU speedup using our proposed *prediction interval with neighbor fetching* that amortizes predictor cost while maintaining accuracy within $\approx$1.1\%, and up to 7.6$\times$ reduction in latency compared to Dense Attention with CPU offloading. Code is available: https://github.com/abdelfattah-lab/TokenButler
arXiv:2605.16070v1 Announce Type: cross
Abstract: Antibody-based therapeutics-including antibody-drug conjugates (ADCs), bispecific antibodies, and novel formats-are reshaping oncology, yet key determinants of efficacy, safety, and manufacturability frequently emerge after conjugation and formulation. We argue that computational biophysics provides an underexploited framework to address this gap by connecting molecular interactions to biological outcomes. We highlight how molecular dynamics, coarse-grained simulations, and free energy calculations reveal how conjugation site, linker chemistry, and drug-antibody ratio reshape conformational landscapes. We emphasize structural coupling between antibody, linker, and payload, with implications for antigen binding, internalization, and developability. We propose that integrating physics-based modeling into development pipelines-alongside experimental validation-can reduce empirical iteration and de-risk translation. As force fields, and hybrid physics-machine-learning methods improve, this field is poised to become a central driver of next-generation ADC design.
arXiv:2504.08300v5 Announce Type: replace
Abstract: Benchmark-based evaluation, e.g., multiple-choice questions (MCQs) and open-ended questions (OEQs), is widely used for evaluating Large Language Models (LLMs), yet their reliability is undermined by benchmark contamination. When pre-exposed to the testing benchmark during training, less capable LLMs have been found to achieve inflated performance, thereby yielding erroneous results in LLM evaluation. In this study, we reframe contamination as an inherent aspect of learning and seek to disentangle and expose genuine capability acquisition from superficial memorization in LLM evaluation. Following this, firstly, by analyzing model performance under different memorization conditions of MCQs, we uncover a counterintuitive trend: LLMs perform worse on memorized benchmarks than on non-memorized ones, indicating the coexistence of two learning phenomena, i.e., rote memorization and genuine capability learning. To disentangle them, we propose TrinEval, a novel evaluation framework that reformulates MCQs into an alternative knowledge-centric trinity format, reducing memorization while preserving inherent knowledge, enabling the evaluation of genuine capability in the presence of memorization. Extensive experiments validate the effectiveness and robustness of TrinEval in reformulating benchmarks, and the evaluation results further reveal that mainstream LLMs rely on rote memorization for an average of 19.6% of knowledge points across the MMLU and the GSM8K dataset.