https://arxiv.org/api/XOFDS+9O6V4QmR6dKj2Qb4Mh5ZU 2026-06-21T21:49:35Z 273759 1005 15 http://arxiv.org/abs/2601.22642v2 2026-06-15T09:42:06Z

Large Language Models (LLMs) show remarkable capabilities, yet their stochastic next-token prediction creates logical inconsistencies and reward hacking that formal symbolic systems avoid. To bridge this gap, we introduce a formal logic verification-guided framework that dynamically interleaves formal symbolic verification with the natural language generation process, providing real-time feedback to detect and rectify errors as they occur. Distinguished from previous neuro-symbolic methods limited by passive post-hoc validation, our approach actively penalizes intermediate fallacies during the reasoning chain. We operationalize this framework via a novel two-stage training pipeline that synergizes formal logic verification-guided supervised fine-tuning and policy optimization. Extensive evaluation on six benchmarks spanning mathematical, logical, and general reasoning demonstrates that our 7B and 14B models outperform state-of-the-art baselines by average margins of 10.4% and 14.2%, respectively. These results validate that formal verification can serve as a scalable mechanism to significantly push the performance boundaries of advanced LLM reasoning.

2026-01-30T07:01:25Z Accepted by ICML 26 Chuxue Cao Jinluan Yang Haoran Li Kunhao Pan Zijian Zhao Zhengyu Chen Yuchen Tian Lijun Wu Conghui He Sirui Han Yike Guo http://arxiv.org/abs/2603.02668v2 2026-06-15T09:37:11Z

We present SorryDB, a dynamically-updating benchmark of open Lean tasks drawn from 78 real world formalization projects on GitHub. Unlike existing static benchmarks, often composed of competition problems, hillclimbing the SorryDB benchmark will yield tools that are aligned to the community needs, more usable by mathematicians, and more capable of understanding complex dependencies. Moreover, by providing a continuously updated stream of tasks, SorryDB mitigates test-set contamination and offers a robust metric for an agent's ability to contribute to novel formal mathematics projects. We evaluate a collection of approaches, including generalist large language models, agentic approaches, and specialized symbolic provers, over a selected snapshot of 1000 tasks from SorryDB. We show that current approaches are complementary: even though an agentic approach based on Gemini Flash is the most performant, it is not strictly better than other off-the-shelf large-language models, specialized provers, or even a curated list of Lean tactics.

2026-03-03T06:55:15Z Austin Letson Leopoldo Sarra Auguste Poiroux Oliver Dressler Paul Lezeau Dhyan Aranha Frederick Pu Aaron Hill Miguel Corredera Hidalgo Julian Berman George Tsoukalas Lenny Taelman http://arxiv.org/abs/2502.06178v6 2026-06-15T09:35:39Z

Bayesian optimization is highly effective for optimizing expensive-to-evaluate black-box functions, but it faces significant computational challenges due to the cubic per-iteration cost of Gaussian processes, which results in a total time complexity that is quartic with respect to the number of iterations. To address this limitation, we propose a novel algorithm, Bayesian optimization by kernel regression and density-based exploration (BOKE). BOKE uses kernel regression for efficient function approximation, kernel density for exploration, and integrates them into the confidence bound criteria to guide the optimization process, thus reducing computational costs to quadratic. Our theoretical analysis rigorously establishes the global convergence of BOKE under noisy evaluations. Through extensive numerical experiments on both synthetic and real-world optimization tasks, we demonstrate that BOKE not only performs competitively compared to Gaussian process-based methods and several other baseline methods but also exhibits superior computational efficiency. These results highlight BOKE's effectiveness in resource-constrained environments, providing a practical approach for optimization problems in engineering applications.

2025-02-10T06:16:51Z Tansheng Zhu Hongyu Zhou Ke Jin Xusheng Xu Qiufan Yuan Lijie Ji http://arxiv.org/abs/2606.16462v1 2026-06-15T09:31:56Z

Cross-subject EEG decoding promises more training data, but it also exposes neural networks to strong inter-subject distribution shifts. We study whether task supervision and architecture alone can learn subject-aligned representations. We replace a shared EEG encoder with subject-specific encoders followed by a common classifier, and compare this hybrid model with standard EEGNet, AttentionBaseNet, and CTNet baselines with Euclidean Alignment (EA) on four motor-imagery datasets. EA improves shared encoders by recentering subject covariances, but the hybrid encoder largely internalises this role: validation-loss curves and latent-distance analyses change little when EA is removed. Subject-specific heads increase class distinctiveness and place each subject close to its own latent manifold, improving most subjects while leaving a method-sensitive subset. These results support subject-specific encoders as a learned alignment mechanism for EEG decoding and identify head selection for unseen subjects as the remaining bottleneck.

2026-06-15T09:31:56Z Bruna J. Lopes Gabriel Schwartz Sylvain Chevallier Raphael Y. de Camargo Bruno Aristimunha http://arxiv.org/abs/2606.16461v1 2026-06-15T09:31:24Z

Running large language models locally is often impractical, pushing inference on sensitive text to third-party providers. Split inference partially mitigates this by keeping tokens on the client and sending only hidden representations, but these representations can still be recovered via nearest-neighbor search against the public embedding table. We propose an orthogonal obfuscation procedure in which the client multiplies embeddings by a secret orthogonal matrix before transmission. To enable correct inference under arbitrary rotations, we introduce ConjFormer, a transformer variant that is exactly $\mathrm{O}(d)$-equivariant via a lightweight normalization change (scalar RMSNorm) together with blockwise orthogonal conjugation of all linear weights. As a result, the server performs the full forward pass entirely in the rotated basis and never observes unrotated hidden states. Experiments on GPT-2 and Llama 3.2 1B models fine-tuned on PubMed show that orthogonal obfuscation eliminates direct cosine nearest-neighbor inversion and reduces token recovery from over 35% top-10 to at most 1.3%, while increasing perplexity by only 0.4% after fine-tuning. These results indicate that enforcing symmetry at the architectural level can provide a practical defense for privacy-preserving LLM inference without noise injection or heavy cryptographic machinery.

2026-06-15T09:31:24Z Alexander Yukhimchuk Andrey Shulga Mladen Kolar Martin Takáč http://arxiv.org/abs/2606.16456v1 2026-06-15T09:28:24Z

Mixture-of-Experts (MoE) models enable efficient scaling, but training them from scratch remains prohibitively expensive. MoE upcycling mitigates this cost by converting pretrained dense models into sparse MoE models. However, existing upcycling methods typically rely on large-scale continued training and often perform poorly under data-constrained supervised adaptation, due to either homogeneous experts or overly disruptive perturbations to pretrained parameters. In this setting, effective upcycling must leverage pretrained weight structure while introducing sufficient diversity among routed experts. To this end, we propose SVD-Partitioned Residual Initialization (SPRI), which distributes SVD-partitioned residuals derived from pretrained feed-forward network (FFN) weights across routed experts, introducing controlled expert diversity grounded in pretrained spectral structure. We further introduce a two-stage training strategy to improve adaptation stability. We evaluate SPRI on multilingual speech-to-text translation, where limited supervised data challenges MoE upcycling and multiple target languages provide natural routing heterogeneity. On CoVoST2 across 15 En-to-XX directions, SPRI improves average BLEU and COMET over fully fine-tuned dense models by 2.58 and 3.32 points, respectively, and outperforms the prior best MoE upcycling baseline by 3.39 BLEU and 4.34 COMET points.

2026-06-15T09:28:24Z 8pages, 12 tables, 3 figures Weiqiao Shan Ruixiang Mao Yuang Li Yuhao Zhang Yingfeng Luo Tong Zheng Chen Xu Yucheng Qiao Chunxiang Jin Yi Yuan Jingdong Chen Tong Xiao Jingbo Zhu http://arxiv.org/abs/2606.16454v1 2026-06-15T09:27:50Z

Low-Rank Adaptation (LoRA) enables efficient adaptation of large pre-trained models to downstream tasks by parameterizing weight updates with low-rank matrices. In this paper, we investigate the limitations of the LoRA parameterization from a geometric perspective. Specifically, we show that when a full fine-tuning gradient is backpropagated to the low-rank matrices, it undergoes anisotropic scaling driven by their singular values. We argue that this phenomenon is undesirable because it distorts the full fine-tuning gradient by skewing it toward dominant singular directions while suppressing others. Our analyses demonstrate that anisotropic gradient scaling reduces the effective rank of the low-rank matrices' gradients and results in suboptimal alignment between the full fine-tuning gradient and its low-rank approximation in LoRA, thereby exacerbating the gap to full fine-tuning. To address these limitations, we propose a new low-rank parameterization, SDS-LoRA, which structurally decouples singular values from the backward pass. Our method ensures that the full fine-tuning gradient backpropagates only through the orthonormal bases of the low-rank matrices' subspaces, independent of their scales. Convergence analysis demonstrates that while LoRA's convergence rate degrades with the condition number of the low-rank matrices, SDS-LoRA remains independent of it. Experimental results across natural language and vision benchmarks show that SDS-LoRA improves loss convergence and reduces the gap to full fine-tuning, significantly enhancing adaptation performance.

2026-06-15T09:27:50Z Junghun Oh Sungyong Baik Kyoung Mu Lee http://arxiv.org/abs/2606.16440v1 2026-06-15T09:11:26Z

Publicly documented accelerator architectures generally separate training computation from optimizer-state updates or rely on external memory and host orchestration. This paper presents NeuronFabric, a software reference architecture intended for future FPGA and ASIC implementations of transformer training with local Adam updates. A complete C# prototype implements forward pass, backpropagation, and Adam optimization without external machine-learning frameworks. The goal is to validate numerical correctness and memory requirements before hardware implementation. The evaluated model is a 334K-parameter autoregressive transformer (d=88, H=4, f=264, L=4, vocab=256) trained on the Shakespeare corpus. The BF16W configuration achieves evaluation loss 1.5426 after 80K samples, compared with 1.5224 for an FP32 GPU reference, while producing coherent character-level text. The paper introduces BF16W, which stores weights in BF16 while retaining Adam optimizer moments in FP32. This reduces memory requirements for on-chip training. A 334K-parameter FP32 model with Adam moments requires approximately 4.0 MB, matching the BRAM capacity of a Xilinx ZCU102 device. The BF16W variant requires approximately 3.34 MB, leaving memory available for activation storage. We describe the vocabulary-budget constraint observed during earlier experiments, quantify BF16W memory savings, and outline FPGA training as the next stage of development. No FPGA measurements are included in this paper. This publication serves as a public architectural disclosure and software reference implementation for future FPGA and ASIC exploration of the NeuronFabric architecture.

2026-06-15T09:11:26Z Evgeny Ukladchikov http://arxiv.org/abs/2409.01062v5 2026-06-15T09:08:44Z

Model Inversion (MI) attacks pose a significant privacy threat by reconstructing private training data from machine learning models. While existing defenses primarily concentrate on model-centric approaches, the impact of data on MI robustness remains largely unexplored. In this work, we explore Random Erasing (RE), a technique traditionally used for improving model generalization under occlusion, and uncover its surprising effectiveness as a defense against MI attacks. Specifically, our novel feature space analysis shows that models trained with RE-images introduce a significant discrepancy between the features of MI-reconstructed images and those of the private data. At the same time, features of private images remain distinct from other classes and well-separated from different classification regions. These effects collectively degrade MI reconstruction quality and attack accuracy while maintaining reasonable natural accuracy. Furthermore, we explore two critical properties of RE including Partial Erasure and Random Location. Partial Erasure prevents the model from observing entire objects during training. We find this has a significant impact on MI, which aims to reconstruct the entire objects. Random Location of erasure plays a crucial role in achieving a strong privacy-utility trade-off. Our findings highlight RE as a simple yet effective defense mechanism that can be easily integrated with existing privacy-preserving techniques. Extensive experiments across 37 setups demonstrate that our method achieves state-of-the-art (SOTA) performance in the privacy-utility trade-off. The results consistently demonstrate the superiority of our defense over existing methods across different MI attacks, network architectures, and attack configurations. For the first time, we achieve a significant degradation in attack accuracy without a decrease in utility for some configurations.

2024-09-02T08:37:17Z Accepted in Transactions on Machine Learning Research (TMLR). First two authors contributed equally Viet-Hung Tran Ngoc-Bao Nguyen Son T. Mai Hans Vandierendonck Ira Assent Alex Kot Ngai-Man Cheung http://arxiv.org/abs/2606.16434v1 2026-06-15T09:06:27Z

Accurate state of health (SOH) estimation is a critical diagnostic service for lithium-ion battery management. However, reliance on labor-intensive manual feature engineering and opaque black-box models hinders scalable industrial deployment. To address this, we introduce TC-SOH: a modular, plug-and-play service architecture for autonomous, end-to-end SOH prediction. TC-SOH employs a temporal-contrastive mechanism and a cross-window prediction pretext task to extract degradation-relevant representations directly from raw operational data. To improve transparency, we connect model efficacy with representation diagnostics: visualization, sensitivity analysis, redundancy analysis, bidirectional probing, future-SOH probing, and temporal shuffling show that learned features overlap with selected expert descriptors while retaining additional SOH-relevant variation, and that ordered temporal context improves subsequent-SOH prediction. Across four public datasets, TC-SOH outperforms the considered physics-informed and data-driven baselines, reducing MAPE by 1.91 times and RMSE by 2.13 times.

2026-06-15T09:06:27Z Junting Wen Dan Li Qihao Quan Xiwen Wang Hang Yang Zhaohong Meng Zigui Jiang Changlin Yang Tianle Liu Diego Muñoz-Carpintero Jian Lou http://arxiv.org/abs/2505.13553v3 2026-06-15T09:04:45Z

The hallucination of code generation models hinders their applicability to systems requiring higher safety standards. One critical bottleneck in addressing code hallucination is the difficulty of identifying the functional correctness of generated code, due to its unnatural form. We address this core bottleneck by automatically generating unit tests using dynamic code analysis tools, leveraging the \emph{executable nature} of code. Accordingly, we propose a \emph{selective code generator} that abstains from uncertain generations -- based on the functional correctness evaluated by generated unit tests -- to theoretically control the correctness among non-abstained answers, \ie the false discovery rate. Finally, we propose to use generated unit tests in evaluation as well as in learning for precise code evaluation, calling this paradigm \emph{FuzzEval}. We demonstrate the efficacy of our method along with the controllability of code hallucination and reasonable selection efficiency.

2025-05-19T06:29:16Z ICML 2026 Jaewoo Jeong Taesoo Kim Sangdon Park http://arxiv.org/abs/2606.16429v1 2026-06-15T09:04:13Z

Hybrid linear attention models offer an appealing path to faster long-context inference: they reduce the quadratic cost and KV-cache burden of full softmax attention while retaining much of the quality of Transformer models. A practical way to obtain such models is to convert a pretrained Transformer instead of pretraining a new architecture from scratch, but this conversion is still brittle. Simply copying the teacher attention projections into a Gated DeltaNet (GDN) student does not specify the new recurrent decay, write, and output-gating dynamics. As a result, the converted model often starts in a poor dynamical regime and must spend many distillation tokens repairing initialization rather than learning the remaining teacher behavior. We propose Taylor-Calibrate, a lightweight initialization method for hybrid GDN students. The method uses Taylor-guided teacher attention statistics to set the value projection, memory timescale, write gates, and output gate, then applies a short per-layer alignment step to match each converted layer to the teacher output. Across four teacher settings and three retained-layer policies, Taylor-Calibrate gives substantially stronger zero-shot students, with up to an 88x improvement in a representative ablation, and reaches matched recovery targets with 4.9x--9.2x fewer training tokens than naive conversion.

2026-06-15T09:04:13Z 24 pages, 9 figures Zhongzhu Zhou Qingyang Wu Junxiong Wang Mayank Mishra Shuaiwen Leon Song Ben Athiwaratkun Chenfeng Xu http://arxiv.org/abs/2606.17113v1 2026-06-15T09:01:49Z

Distinguishing causal adverse drug events (ADEs) from spurious correlations remains a central challenge in pharmacovigilance. The InferBERT framework integrates transformer models with Do-calculus, but its success hinges on the underlying classification model. This study evaluates the impact of model choice in InferBERT, assessing whether simpler models suffice, if domain-specific pre-training helps, whether scaling to LLMs improves causal detection, and the effect of post-hoc calibration. We performed a comparative study on two benchmarks: Analgesics-induced Acute Liver Failure (AILF) and Tramadol-related Mortalities (TRAM). Four models were evaluated-XGBoost (baseline), ALBERT (original InferBERT), BioBERT (biomedical transformer), and Med-LLaMA (medical LLM)-using 5-fold cross-validation repeated over 20 runs. We measured accuracy, Expected Calibration Error (ECE) pre- and post-isotonic regression, and Jaccard concordance of causal terms with PRR, ROR, and EBGM; significance was tested with paired t-tests. BioBERT achieved the highest accuracy on both datasets, while Med-LLaMA underperformed despite its size and parameter-efficient fine-tuning. Domain-specific pre-training was decisive. Calibration improved ECE but had mixed effects on accuracy and causal discovery. BioBERT's superiority also yielded the strongest concordance with traditional pharmacovigilance signals. These results show that domain-specific pre-training provides a clear advantage over simpler baselines and larger LLMs. Investing in manageable, domain-aware models is more effective for computational pharmacovigilance than simply scaling model size.

2026-06-15T09:01:49Z 10 pages, 5 figures Csaba Kiss Roland Molontay Gabriele Pergola http://arxiv.org/abs/2511.05963v4 2026-06-15T08:56:56Z

Transformers replace recurrence with a memory that grows with sequence length and self-attention that enables ad-hoc lookups over past tokens. Consequently, they lack an inherent incentive to compress history into compact latent states with consistent transition rules. This often leads to learning solutions that generalize poorly. We introduce Next-Latent Prediction (NextLat), which extends standard next-token training with self-supervised predictions in the latent space. Specifically, NextLat trains a transformer to learn latent representations that are predictive of its next latent state given the next token. Theoretically, we show that these latents provably converge towards belief states, compressed information about the history necessary to predict the future. This simple auxiliary objective injects a recurrent inductive bias into transformers while leaving their architecture, parallel training efficiency, and inference unchanged. NextLat effectively encourages transformers to form compact internal world models with coherent belief states and transition dynamics -- crucial properties not guaranteed by standard next-token prediction alone. Empirically, across benchmarks in world modeling, reasoning, planning, and language modeling, NextLat demonstrates significant gains over standard next-token prediction and other baselines in downstream accuracy, representation compression, and lookahead planning. Furthermore, NextLat enables variable-length self-speculative decoding, accelerating inference by up to 3.3x in language modeling. NextLat offers a simple yet effective paradigm for learning compact, predictive representations in transformers that generalize better. Our code is available at https://github.com/JaydenTeoh/NextLat.

2025-11-08T10:41:26Z Microsoft Research Preprint Jayden Teoh Manan Tomar Kwangjun Ahn Edward S. Hu Tim Pearce Pratyusha Sharma Akshay Krishnamurthy Riashat Islam Alex Lamb John Langford http://arxiv.org/abs/2605.23234v3 2026-06-15T08:49:35Z

Assessing the spatial fairness of predictive models involves establishing whether they are statistically penalizing (favoring) individuals associated with certain geographical locations. Literature on this topic makes the fundamental assumption that each individual is assigned to a single geographical location (e.g., place of residence). However, fairness with respect to the set of locations where one has been, i.e., their movement patterns over different regions, also matters when fairness is considered. Consequently, we argue that it is necessary to generalize the notion of spatial fairness to also include movement patterns, leading to the novel problem of assessing predictive models for fairness relative to the movements of individuals. To deal with this problem, we propose an approach that first associates the movements of individuals to certain geographic regions, considering multiple spatial partitions with different resolutions and alignments, and then employs a suitable spatial scan statistic to assess whether a predictive model is fair based on movement patterns. In the experimental evaluation, we study the performance of our approach over thousands of synthetic unfair datasets, showing that it is effective at detecting this new type of unfairness and at retrieving the set of objects treated unfairly, while localization performance exhibits a consistent multi-resolution trade-off.

2026-05-22T04:53:34Z 33 pages, 10 figures, 7 tables Francesco Lettich Mario A. Nascimento Chiara Pugliese Chiara Renso