The paper introduces Skill Schema Transport (SST), a post‑training framework that mitigates catastrophic forgetting in continual domain adaptation by translating new domain knowledge into a model’s existing skill space. SST abstracts each example type into a reusable skill schema, preserving essential operations while removing domain‑specific surface forms, thereby turning knowledge updates into procedural reuse rather than trajectory memorization. Experiments on two 3B instruction‑tuned backbones show that SST consistently reduces forgetting and improves generalization, achieving up to a 5.75‑point performance gain and a dramatic drop in maximum forgetting from 42.41 to 7.10 points on the Qwen‑3B MATH stream.
By Zijie Liu, Zhen Tan, Charles Fleming, Tianlong Chen
The paper introduces GPSCS, a learning-based framework that uses genetic programming to evolve a symbolic scoring function for selecting cutting planes in mixed-integer linear programming. It also employs a statistics-based method to decide how many top-ranked cuts to add during each separation round. Experiments on diverse synthetic and real-world benchmarks show that GPSCS reduces solving time by 6% over SCIP’s default strategy and outperforms a neural-network approach and a no-cut baseline, indicating strong generalization and robustness.
By Gwen maudet, Mathis Da Cunha, Gr\'egoire Danoy
The paper introduces methods to improve training of stochastic flow maps for stochastic differential equations. It proposes representing Brownian paths with a shifted Legendre polynomial basis and using a Wong‑Zakai styled objective to enable derivative‑based losses. The authors also provide a fine‑tuning recipe that turns large pre‑trained flow models into strong stochastic flow maps, achieving a three‑fold improvement on ImageNet compared to SSFM.
By Zander W. Blasingame, Antonio Franca, Raul Minan, Christian Koke, Niklas Rindtorff, Alexander Tong
arXiv:2610. 28994v1 Announce Type: cross Abstract: We address the Diverse Traveling Salesman Problem (D-TSP) that seeks a set of $k$ distinct TSP tours.
By Ssu-Yuan Lo, Ching-Kai Wang, Kuan-Lun Chen, Hao-Tsung Yang
The paper introduces internal guidance as a new sampling technique for tilted diffusion models, which modifies the evaluation point of the base denoiser rather than altering its output. For discrete distributions in the hypercube, the denoiser of a linearly tilted density can be obtained by evaluating the base denoiser at a shifted and rescaled input. By augmenting training with random diagonal rescalings, a single denoiser network can sample binary reward tilts, class labels, and infilling without fine‑tuning or external guidance.
By Curtis McDonald, Peter Bartlett
arXiv:2610. 29052v1 Announce Type: cross Abstract: A typed decision maps evidence, a question and a fixed set of options to a probability for every option, computed in one forward pass, by reading a language model's logits or a small head over its hidden states, instead of generating and parsing text.
By Yuchuan Weng, Chi Liu, Xin Chen, Srinivasan Manoharan
The paper investigates how diffusion language models maintain factual fidelity when generating text from tables, comparing masked diffusion models (MDLM, LLaDA) to autoregressive models (GPT‑2, GPT‑Neo) on the ToTTo dataset. Using a retrieval‑based few‑shot protocol, the study finds that while diffusion models score higher on reference similarity (BLEU), they exhibit lower source grounding and different failure modes—autoregressive models drift from the table, whereas diffusion models stay on topic but swap entities. The authors conclude that reference‑based metrics alone are insufficient for assessing faithfulness in table‑to‑text generation.
By Zaynab Merimi, Noussayba El Marrakchi, Arthur Satouf
NodeSynth is a method that generates socially aligned synthetic data by combining a fine‑tuned taxonomy generator with credible sources and extracted sensitive attributes to produce queries that reflect real societal contexts and nuanced harms. In evaluations across ten open‑weight and proprietary large language models on Medical Advice and Self‑Harm tasks, NodeSynth outperformed generic synthetic baselines, structured synthetic benchmarks, and authentic human‑authored queries, inducing higher failure rates that highlight model vulnerabilities. Ablation and variance attribution studies show that the granular Level 3 taxonomic expansion and sociotechnical grounding of queries drive these failure rates, underscoring the importance of nuanced data for high‑stakes evaluation.
By Qazi Mamunur Rashid, Xuan Yang, Zhengzhe Yang, Yanzhou Pan, Erin van Liemt, Darlene Neal, Kshitij Pancholi, Jamila Smith-Loud
The paper introduces Bilateral Sharpness-Aware Minimization (BSAM), a new optimization method that combines the traditional Max-Sharpness (MaxS) metric with a newly defined Min-Sharpness (MinS) metric. By using the difference between the training loss and the minimum loss in a neighborhood around the current weights, BSAM creates a more comprehensive Flatness Indicator (FI) that guides the search toward flatter minima. Experiments show that BSAM improves generalization and robustness over vanilla SAM in tasks such as classification, transfer learning, human pose estimation, and network quantization.
By Jiaxin Deng, Junbiao Pang, Baochang Zhang, Qingming Huang
The paper investigates why AI-generated text has become harder to understand, focusing on psycholinguistic features linked to human processing difficulty such as object relative clauses, long syntactic dependencies, and low-frequency n-grams. Using a new corpus of AI-generated text from diverse domains, the study finds that frontier and near-frontier language models employ these difficult features more frequently than paraphrases from round-trip translation and human writing, including more inanimate subjects with animate verbs, longer dependencies, stacked nouns, and reduced object relatives. The authors hypothesize that AI models prioritize information density for tasks like math, code, and reasoning, diverging from human communication goals, and demonstrate this by fine‑tuning a model with a length penalty, which reduces human‑like linguistic behavior.
By Thomas Hikaru Clark, Matias B. Martin, Kyle Mahowald
The paper explores how to leverage parallel corpora for multilingual pretraining by concatenating each document with its translation in the same training sequence, a method called Parallel Pretraining (PPT). Compared to standard pretraining (SPT), PPT alone does not consistently improve language understanding, but it builds a stronger cross‑lingual foundation. A two‑stage approach—first PPT to establish cross‑lingual knowledge, then SPT to fine‑tune for standard language modeling—consistently outperforms standard pretraining in both high‑resource and low‑resource settings.
By Zhihong Huang, Yao Lu, Derek F. Wong, Hamid Alinejad-Rokny, Min Yang
The paper introduces CS‑SeLeRoSa, a synthetic dataset of over 600k code‑mixed Romanian sentences with English, German, French, and Italian, generated via four methods (random replacement, phrase translation, dependency parsing, and large language models). It evaluates multilingual large language models on satire detection in this dataset and a human‑annotated gold benchmark, CS‑SeLeRoSa‑Gold, showing that newer models like Gemma 4 31B outperform older open models and achieve competitive zero‑shot performance, with fine‑tuning further improving results.
By R\u{a}zvan-Alexandru Sm\u{a}du, Ioana Buhnila, Daniela Tanase, Dumitru-Clementin Cercel, Florin Pop, Mihaela-Claudia Cercel
AfrIFact is a new dataset designed to support automatic fact‑checking in ten African languages and English, covering information retrieval, evidence extraction, and fact verification. Experiments show that current embedding models struggle with cross‑lingual retrieval, and that cultural and news documents are easier to retrieve than healthcare documents. Large language models also lack robust multilingual fact‑verification, but few‑shot prompting and task‑specific fine‑tuning can significantly improve performance.
By Israel Abebe Azime, Jesujoba Oluwadara Alabi, Crystina Zhang, Iffat Maab, Atnafu Lambebo Tonja, Tadesse Destaw Belay, Folasade Peace Alabi, Salomey Osei, Saminu Mohammad Aliyu, Nkechinyere Faith Aguobi, Bontu Fufa Balcha, Blessing Kudzaishe Sibanda, Davis David, Mouhamadane Mboup, Daud Abolade, Neo Putini, Philipp Slusallek, David Ifeoluwa Adelani, Dietrich Klakow
The paper introduces DONIC, a generalist bimanual tracking controller trained on diverse operation categories and human-motion data, and BiHOI, a multi‑view dataset of 1,575 sequences with 229 objects that captures sustained bimanual interactions. DONIC employs proximal policy optimization combined with behavioral cloning, an object‑centered per‑hand reward, and Clip‑Centered Advantages to learn coordinated manipulation from a single policy. The method achieves 86.79% success on unseen trajectories of training objects and 57.32% on held‑out objects, outperforming existing bimanual tracking baselines by large margins.
By Mingze Sun, Ziang Cao, Yuan Fei, Qiankuan Tao, Zihan Wang, Tianqi Liu, Junhao Chen, Haiwen Diao, Fangzhou Hong, Zhaoxi Chen, Ziwei Liu
The paper introduces PLOVIS, a data‑efficient framework for semantic segmentation of 3D point clouds that tackles three simultaneous data shortages: few training scenes, sparse point‑level labels, and missing 2D image sequences. PLOVIS generates pseudo‑labels by applying an open‑vocabulary image segmentation model directly to 3D point clouds, then refines these labels through a two‑stage filtering process and a class‑balanced memory bank. Experiments on ScanNet, S3DIS, Toronto3D, and Semantic3D show that PLOVIS outperforms existing fine‑tuning and weakly supervised methods under realistic, data‑scarce conditions.
By Takahiko Furuya
Goal images used as control targets in robotic manipulation are not best chosen by general image‑editing quality. The authors show that large proprietary editors, though top on editing‑quality leaderboards, produce poorer control targets than smaller fine‑tuned open models, especially on long‑horizon tasks. They introduce SGE‑Goal, a spatially‑grounded editing framework that predicts edit‑region masks before synthesis, enabling open base models to become the strongest goal generators on CALVIN and nearly as strong on SIMPLER‑Bridge, with superior performance on chained tasks.
By Riddhi Chatterjee, Ritish Shrirao, Kushagra Mehrotra, Viswanath Gopalakrishnan
The paper introduces a model‑agnostic, nonparametric framework for classification when training data contains label noise, particularly useful when a large noisy dataset is supplemented by a small clean dataset. It addresses the distribution shift between noisy and clean data that undermines traditional parametric transfer learning methods, and it is applicable to a wide range of classifiers. The authors provide a rigorous statistical theory and validate the approach with simulations and a real‑world medical imaging study on pneumonia diagnosis.
By Zhu Guojun, Zhang Sanguo, Ren Mingyang
The paper introduces PARED, an approach that uses inverse reinforcement learning to derive an explicit, interpretable reward from expert demonstrations. By training a lightweight discriminator on response-level features, PARED can guide both inference-time reranking and on-policy reinforcement learning, improving alignment for models that have not undergone supervised fine-tuning and further refining those that have. The method also supports contextual alignment, enabling a single policy to adapt its responses to different audience preferences.
By Micha{\l} Wili\'nski, Liu Leqi, Chirag Nagpal
The paper studies how long‑context continual pre‑training (LCCP) scales with token budget and model size. An empirical additive power law is derived that predicts LCCP performance across five model scales and over a hundred budget configurations, and a new Constant‑Last4 protocol is introduced for stable monitoring. Experiments show that performance continues to improve without clear saturation, even after 128K context extension and fine‑tuning, indicating that scaling LCCP yields transferable gains for downstream tasks.
By Ziqing Qiao, Xiaoyue Xu, Sikui Zhang, Yinuo Xu, Bingxiang He, Chaojun Xiao, Xu Han
NeuralSRender is a sim‑to‑real video translation framework that improves the photorealism of synthetic driving videos while preserving geometric consistency. It builds on a pretrained video diffusion transformer and uses a lightweight projector to inject G‑buffer geometric and material priors, with LoRA for efficient adaptation. The two‑stage training first learns G‑buffer conditioning from synthetic data, then adapts to real‑world appearance, adding only 1.45 % parameter overhead and requiring 24 GB GPU memory. The authors also release GenReal, a large‑scale dataset of synthetic and real driving videos with G‑buffers, and show that training perception models on NeuralSRender‑translated videos boosts detection AP₅₀ on nuImages from 35.1 to 40.6 and segmentation mIoU on Cityscapes from 32.4 to 40.8.
By Yipeng Lu, Youqi Liao, Zhenyu Wang, Yuning Peng, Zhen Dong