AnomalyGFM is a graph foundation model designed for zero- and few-shot anomaly detection in graphs. It learns graph‑agnostic normal and abnormal prototypes by aligning data‑independent class representations with node representation residuals, enabling consistent abnormality measurement across diverse graph datasets. Experiments on 11 real‑world GAD datasets show that AnomalyGFM outperforms state‑of‑the‑art methods in both zero‑shot and few‑shot settings.
By Hezhe Qiao, Chaoxi Niu, Ling Chen, Guansong Pang
The paper introduces Dr. Post-Training, a framework that treats general training data as a data‑induced regularizer to prevent overfitting to scarce target objectives, rather than merely a pool for selection. It constructs a feasible set of model update directions from general data and projects target‑driven updates onto this set, unifying standard training and existing data‑selection methods as special cases along a bias‑variance spectrum. Experiments on SFT, RLHF, and RLVR demonstrate that these methods outperform both standard training and state‑of‑the‑art data‑selection baselines, especially when general and target data are highly misaligned, while system optimizations keep overhead minimal.
By Pingbang Hu, Xueshen Liu, Z. Morley Mao, Jiaqi W. Ma
The paper introduces a practical quantum encoding technique that allows Euclidean distance calculation between vectors with different L2 norms by appending a single component to each vector. This method is applied to the quantum k‑nearest neighbors (QKNN) problem, offering a simpler and more qubit‑efficient implementation than existing approaches. The authors provide an implementation and benchmark it against a recent Euclidean‑distance QKNN method, showing improved performance in simulations with realistic noise due to the reduced qubit count and more coherent circuits.
By Christopher Rector, Jiangwen Sun
The paper demonstrates that language models often accept and use unsupported premises when answering questions, even though the source text does not provide the necessary information. By manipulating whether an unsupported fact is usable, the authors show that models’ refusal rates drop significantly while their commitment to an answer increases. This behavior is consistent across multiple models and datasets, indicating that models’ verification checks are heavily influenced by the presence of usable material rather than strict source support.
By Naveen Madiraju
The paper investigates whether System‑1‑style decision models can replace dedicated rerankers in retrieval‑augmented generation (RAG) systems. After screening 21 open‑weight models on 11 benchmarks, the authors compare a promising decision model with a dedicated reranker on identical candidate lists. While limiting candidate text length improves reliability and makes some configurations competitive, a tuned reranker remains faster and uses less GPU memory, indicating that replacing a reranker requires balancing quality, latency, and memory.
By Qiaomu Li, Ying Xie
Auto-Research Arena is a multi-target benchmark designed to evaluate autonomous research systems and uncover new ideas in language modeling. It separates admissibility—determined by model quality and resource constraints—from improvement, using nine diverse targets covering model, data, training, and inference efficiency. The benchmark, applied to language model pretraining, distinguishes existing research harnesses and reveals significant room for improvement, while also enabling the reuse of discovered ideas to achieve lower training loss and better held‑out performance.
By Xi Chen, Chongjian Ge, Yiran Xu, Yiqun Mei, Henrik I. Christensen, Hao Tan, Ziwen Chen, Hanwen Jiang
RoboJev-27B is an open‑weight vision‑language model that predicts whether a robot has successfully completed a manipulation task by analyzing a task instruction, ten frames of the attempt, and optional auxiliary camera frames. Trained with a rank‑32 LoRA adapter on 52,598 episodes, it achieves 95.2% balanced accuracy and 0.993 AUROC on held‑out episodes of the same robot model, and maintains strong performance on unseen labs, robots, and community‑collected data. The model runs in 0.26 seconds per episode on an H100 GPU and is released with inference code and a large, transparent evaluation dataset for reproducible benchmarking.
By Josef Liyanjun Chen
Matcha is a new molecular docking method that refines ligand poses through three flow-matching stages—translation, rotation, and torsion—before applying GNINA for minimization, filtering, and ranking. It achieves high physical plausibility and is approximately 28 times faster than current large-scale co‑folding models, outperforming existing benchmarks. The authors provide code, benchmarking artifacts, and model checkpoints on GitHub and Zenodo.
By Daria Frolova, Talgat Daulbaev, Egor Sevriugov, Sergei A. Nikolenko, Dmitry N. Ivankov, Ivan Oseledets, Marina A. Pak
The paper introduces MUNKEY, a memory‑augmented transformer that enables machine unlearning by design. By decoupling instance‑specific memorization from model weights, unlearning is achieved simply by deleting an instance‑identifying key, allowing zero‑shot forgetting without retraining or accessing original data. Experiments on natural image, fine‑grained visual recognition, and medical datasets show that MUNKEY outperforms existing post‑hoc unlearning methods while maintaining predictive performance.
By Sonia Laguna, Jorge da Silva Goncalves, Moritz Vandenhirtz, Alain Ryser, Irene Cannistraci, Julia E. Vogt
Markovian Circuit Tracing (MCT) is a benchmark that tests whether transformer activations encode a compact state abstraction with recoverable transition structure. In experiments with small causal transformers trained on sequences from four‑state hidden Markov models, models achieve near‑optimal validation loss and recover transition matrices accurately under easy and medium observability, but performance collapses to control levels under hard observability. Various controls—including shuffled data, untrained activations, and sparse autoencoders—demonstrate that the recovered transition structure is genuinely causally useful rather than a trivial artifact.
By Abdullah X
The paper introduces SNTO, a simultaneous neural topology optimization framework that represents state, adjoint, and design variables with custom neural networks optimized in a single loop. It employs a hardware‑aware training scheme that partitions the design domain across multiple GPUs and slices subdomains to manage memory, enabling stress‑constrained topology optimization at millions of degrees of freedom. Experiments on 2D benchmarks show sharper boundaries and reduced gray areas compared to finite element methods, while 3D problems with up to eight million points are solved efficiently on eight GPUs.
By Xiangyu Sun, Shirin Hosseinmardi, Ramin Bostanabad
The study examined 305 quantized large language models (LLMs) from public registries, testing them on a 15‑task code screen and 164 HumanEval+ tasks. Only seven low‑bit, small‑parameter models failed to solve more than six tasks, and none of the files were corrupted—reproducibility did not guarantee functionality. The authors released audit data and a tool, recommending pre‑publication screening and calibration of low‑bit tags.
By Aditi Patodiya
ETC-Bench is a multilingual benchmark for evaluation‑side transcript canonicalization (ETC), covering 181 phenomenon types across 12 languages. Each instance is a minimal pair that shares a sentence context and differs only in a target span exhibiting one phenomenon. The benchmark introduces Semantic‑preserving Exact‑match Normalization Accuracy (SENA) to evaluate canonicalization success and semantic preservation, and a Taxonomy‑guided Agent Canonicalization Skill (TACS) to guide LLM agents in canonicalizing transcripts.
By Zhiheng Qian, Junhao Du, Jing Peng, Hanqi Li, Kunyang Peng, Feng Xue, Kai Yu
The paper investigates whether token‑confidence scores used to prune reasoning chains in language models actually reflect model certainty or merely chain length. By treating per‑token confidence as a signal and applying spectral analysis and change‑point detection, the authors find that the windowed‑minimum confidence score largely correlates with how long a chain is, offering no additional predictive power for correctness once length is accounted for. Although the confidence spectrum contains a weak signal of rapid token‑to‑token fluctuation in incorrect chains, it is too faint to improve majority voting, and early‑stopping methods based on confidence or change detection do not outperform simply sampling fewer chains.
By Pragna Mallappa
LinMix proposes a hyperparameter‑free vertex sampling strategy for data mixing in language model pretraining, treating domain attribution as a linear regression problem. By sampling mixes at the vertices of a capped simplex, it decorrelates domain effects and yields an interpretable linear model whose coefficients directly reflect each domain’s contribution. Experiments on 1B‑parameter OLMo2 and 4B Qwen3.5 models demonstrate that LinMix outperforms traditional Dirichlet‑based mixing, achieving significant gains on reasoning, code, and math benchmarks and improving overall accuracy by 11% on average over the best nonlinear baseline.
By Hubert Baniecki, Przemyslaw Biecek, Martin Pawelczyk
EffiBench-X is a new multi‑language benchmark that measures the efficiency of code generated by large language models (LLMs). It covers six programming languages—Python, C++, Java, JavaScript, Ruby, and Golang—using competitive programming tasks with human‑expert solutions as efficiency baselines. Experiments show that while LLMs produce functionally correct code, they consistently fall short of human efficiency, achieving only about 62 % of human performance on average, with notable language‑specific differences.
By Yuhao Qing, Boyu Zhu, Mingzhe Du, Zhijiang Guo, Terry Yue Zhuo, Qianru Zhang, Jie M. Zhang, Heming Cui, Siu-Ming Yiu, Dong Huang, See-Kiong Ng, Luu Anh Tuan
AxisQ introduces a depth‑adaptive bit‑allocation framework for camouflaged object detection (COD) that allocates precision in the channel domain for shallow features and in the local‑frequency domain for deep features, guided by COD‑aware sensitivity. It also incorporates a boundary‑aware Hessian approximation to better preserve subtle foreground regions and ambiguous boundaries during post‑training quantization. Experiments on multiple COD models and benchmarks show significant accuracy gains at low bit widths, with relative weighted F‑measure improvements of 12.7% for CFRN and 33.3% for CamoFormer‑P on the CAMO dataset under W2A2 quantization.
By Chi Ruan, Tianqi Li, Xin He, Xue Geng, Ming-Ming Cheng, Yun Liu
The paper introduces a Graph-informed Kriging framework that embeds aerodynamic interaction data into the surrogate mean function for wind farm layout optimization. By representing the farm as a directed aerodynamic graph and assembling pairwise turbine interactions into a farm-scale mean, the method captures dominant variation in annual energy production, allowing the Gaussian process to focus on residuals. Numerical experiments demonstrate that this approach improves AEP prediction accuracy and sample efficiency over conventional Kriging, achieving competitive solutions on the IEA Wind Task 37 benchmark with only about 280 evaluations and outperforming other methods in both 16- and 48-turbine cases.
By Zhen-Fan Wang, Kai Zhang, Xianping Du, Zhaolong Han, Dai Zhou, Onur Bilgen
The article introduces deep Bayesian filtering techniques for nonlinear structural system identification, emphasizing how learnable dynamics can be integrated with physical inductive biases to achieve uncertainty-aware inference. It reviews key models such as the Deep Markov Model, Physics-guided Deep Markov Model, and Neural Extended Kalman Filter, and evaluates them on a three‑dimensional shear‑frame benchmark with hysteretic Bouc‑Wen connections. The study compares measurement‑conditioned reconstruction versus open‑loop forecasting, explores the impact of physical structure on latent‑state interpretability, and discusses robustness in response and parameter estimation, offering practical guidelines for model construction and deployment.
By Wei Liu, Konstantinos Vlachas, Eleni Chatzi
The paper introduces a physics‑guided generative AI framework that uses flow matching to forecast global total electron content (TEC) one day ahead. By embedding ionospheric dynamics described by a partial differential equation into the transformation process, the model blends physical knowledge with statistical patterns, adjusting their influence locally. Trained on a year of TIEGCM‑generated TEC data, the approach outperforms statistical, deep learning, and other generative baselines, achieving a 39.2% lower RMSE than ConvLSTM and a 23.8% improvement over the same framework without physical guidance.
By Si-Tong Yang, Feng Ye, Xian Lu, Dong Lin, Benjamin C. Martinez