EGCForest is a multi‑view hierarchical memory framework designed for enterprise group chats. It organizes the immutable message log into three complementary forests—Channel, Actor, and Artifact—each capturing temporal workflows, speaker‑conditioned context, and deliverable lifecycles, respectively. By using a Placement Map and sparse structural bridges, EGCForest jointly leverages message‑level relevance, forest summaries, and cross‑view links to retrieve and assemble distributed evidence, outperforming RAG and memory‑system baselines by up to 20.85 points on three benchmarks.
By Lingxi Cui, Jinhao Ruan, Wenqing Wei, Guo Yi, Li Rao, Ke Chen, Lidan Shou
The paper introduces Length‑Conditioned Majority Voting (LCMV), a test‑time training method that enhances the standard Majority Voting (MV) technique for Large Language Models (LLMs). LCMV incorporates an empirical bias that longer Chain‑of‑Thoughts (CoTs) are less likely to be correct by modeling a length‑conditioned prior with a logistic function. This prior is adapted on unlabeled test data without fine‑tuning the LLM, and the resulting weighted votes consistently outperform standard MV and other length‑ or confidence‑based aggregation strategies across multiple reasoning benchmarks.
By Fiorenzo Parascandolo, Enver Sangineto, Xiaoyi Yu, Wenhui Tan, Jipei He, Ruihua Song, Rita Cucchiara
The paper presents an industrial pipeline that converts U.S. state legislation into strict, schema‑constrained XML, emphasizing that the accuracy of the conversion stems more from the deterministic harness engineering than from the underlying language model. The authors conduct a controlled ablation study across three harnesses—zero‑shot, a single tool‑using agent, and a decomposed multi‑agent pipeline—paired with three Claude models on 153 gold bills from two jurisdictions. Results show that the full pipeline with the weakest model outperforms the strongest model alone (0.948 vs. 0.806), and that the harness narrows the model performance gap across four scored dimensions, with the remaining discrepancy largely due to citation normalization handled by an external lookup tool.
By Palvika Bansal, Rittika Jindal, Ruichong Wang
AFCS (Atomic‑Fact‑Conditioned Synthesis) is a new pipeline for generating synthetic data for language‑model pretraining that focuses on extracting and using fine‑grained atomic facts from source documents. Unlike previous methods such as WRAP and Nemotron‑CC, AFCS explicitly requires secondary entities, list/table entries, and precise temporal attributes to appear in the synthesized corpus, employing structure‑aware extraction, source‑grounded verification, local context expansion, and multi‑format generation. In experiments with OLMo‑3‑7B and a 32B model on the FineWiki corpus, AFCS outperforms the baselines on several QA benchmarks and improves general benchmark performance by +1.09 when integrated mid‑training with 100B tokens.
By Yunhao Wang, Feng Zhang, Tinghao Yu, Xing Wu, Jiang Zhou
TreeSpark is a new training and inference framework that extends the semi‑autoregressive DSpark architecture to improve speculative decoding. It builds parent‑conditioned candidate trees in a single backbone pass and then replays their expansion, pruning, and packing during training, using survival negative log‑likelihood and a capped coverage hinge to weight retained edges and promote target‑important alternatives. Evaluations on Qwen3‑4B and Qwen3‑8B across nine general benchmarks and six writing domains show that TreeSpark increases mean acceptance length by 23.4%–31.6% and mean single‑request throughput by 15.1%–25.0% compared to DSpark, with additional gains when using the proposed supervision.
whyItMatters":"TreeSpark demonstrates that a unified training and inference approach can substantially boost both acceptance length and throughput in speculative decoding, offering measurable performance improvements for large language models."
By Jihao Zhao, Qianyu Jiang, Shuaishuai Zu, Pengfei Li, Guanghuan Fang, Ben Feng, Biao Qin, Zhibin Wang, Baochuan Yang
The paper introduces HICO-Bench, a benchmark of 180 everyday coordination scenarios where two agents with private information converse and then record a shared plan. A program verifies that the plans are feasible and an LLM judge checks that the conversation supports the plan, defining a Verified Success. Using Paired-Credit Policy Optimization (PaCPO), the authors improve the Verified Success rate of Qwen3-8B from 20.4% to 28.3%.
By Caixin Kang, Ke Wang, Yuchuan Wu, Xiaoqiang Wang, Ling Li, Bingxue Qiu, Yongbin Li, Yoichi Sato
CSWBench is a benchmark designed to evaluate context-sensitive code-switching in creative writing, featuring two settings: Focused Narrative and Extended Narrative. The accompanying CSWEval framework assesses event-level aspects such as Licensing, Contextual Appropriateness, and Functional Gain, as well as whole-text quality metrics tailored to each narrative setting. Experiments with twelve large language models reveal that while models show some code-switching competence, performance varies significantly across narrative contexts, with stronger challenges in multilingual literary integration and clearer associations between narratively salient switches and contextual/functional realization.
By Zewen Jiang, Yahan Li, Chenlu Guo, Diana Inkpen, Yuan Wu, Yi Chang
The paper introduces PUlSE, a lightweight probing method that uses a language model’s internal representations to decide when to retrieve external biomedical evidence during medical reasoning. By analyzing inter‑layer volatility, PUlSE identifies erroneous reasoning steps and triggers adaptive retrieval, achieving comparable accuracy to exhaustive retrieval while cutting retrieval calls by 72%. The approach also transfers to an unseen legal domain, suggesting broader applicability.
By Danyang Zhao, Sendong Zhao, Yanrui Du, Haochun Wang, Ming Ma, Wenhao Li, Jingbo Wang, Jiatong Liu, Bing Qin
The paper introduces V-PaSS, a physics‑aware semantic system that predicts perovskite solar cell performance by interpolating crystal graphs of 23 pure $ABX_3$ compounds using Vegard’s law and learned bandgap bowing. It couples device stack information—including layer sequence, polarity, and interfacial band offsets—with a scientific language model to encode chemical descriptions. In an institution‑grouped evaluation, V‑PaSS outperforms fourteen baseline models, achieving the lowest mean absolute error (2.861%) and demonstrating significant gains on unseen absorber alloys, new laboratory samples, and devices with unobserved transport layers.
By Qiwei Han, Chi Zhou
The paper introduces mixed integer linear programming (MILP) formulations for training sparse disjunctive normal form (DNF) networks, which are small networks with a conjunctive hidden layer and a disjunctive output layer. By fixing the network structure and optimizing weights, the authors achieve globally optimized rule sets that balance misclassification costs against sparsity. Their enhanced formulation reduces integer decision variables and redundant constraints, and a combined initialization and dropout approach breaks weight symmetries to promote diverse, stable rule representations, leading to improved computational efficiency and competitive accuracy on synthetic and UCI datasets.
By Thomas Pammer, Johannes F\"urnkranz, Markus Sinnl
OrionMoE-v1.0 is a compact tabular foundation model that uses a dual‑axis Transformer and a sparse mixture‑of‑experts architecture to perform in‑context learning on labeled support tables. It selectively activates experts along the feature axis, guided by both individual feature representations and overall dataset context, enabling adaptive computation across diverse tabular problems. Pretrained on synthetic classification tasks, it achieves competitive performance on multiple benchmarks, especially on small‑context, high‑dimensional, and class‑imbalanced datasets.
By Mohamed Bouadi, Nassim Bouarour, Aditya Tanna, Vinay Kumar Sankarapu
The paper investigates general‑purpose coding agents that generate images by writing and executing programs, contrasting them with conventional generative models. On the AgentDrawBench benchmark, these agents excel at preserving structure and meeting precise visual constraints, achieving competitive performance in tasks where compositional accuracy outweighs photorealism, though they fall short in perceptual richness. The study further shows that the agents’ visual code can serve as an external language of thought, retaining substantial information for visual question answering and suggesting an internal, compositionally structured representation that guides visual operations across different drawing interfaces.
By Hengguang Zhou, Yuanhao Ban, Jiaqi Feng, Xiaohuan Pei, Justin Cui, Cho-Jui Hsieh
The paper introduces physics-informed weak regression (PIWR) for estimating parameters in compartmental epidemic models, avoiding the need for data derivatives and jointly estimating initial states. PIWR, combined with split conformal prediction, provides distribution‑free uncertainty intervals with finite‑sample coverage guarantees. Across eleven benchmark models and real influenza outbreak data, PIWR achieves orders of magnitude higher accuracy than existing physics‑informed neural networks while delivering calibrated confidence bands in under two minutes per fit.
By Allan Peter Engsig-Karup
arXiv:2610.20370v1 Announce Type: new
Abstract: Hate speech has become increasingly prevalent on social media, raising concerns about its detection, interpretation, and moderation. Persian, a relativ...
By Ghazal Kalhor, Sima Kamandloo, Mehrsa Shamsian, Behnam Bahrak
arXiv:2610.19108v1 Announce Type: cross
Abstract: Multimodal Large Language Models (MLLMs) have been widely applied in embodied agents and robotics for vision tasks. However, MLLMs suffer from limita...
By Mahdieh Mallahnezhad, Masoud Jafaripour, Mohamadali Tofigh, Sheila Schoepp, Osmar R. Zaiane
arXiv:2610.19475v1 Announce Type: cross
Abstract: Multimodal in-context learning (ICL) is an important capability of multimodal large language models (MLLMs). Its performance is closely tied to the d...
By Yanshu Li, Ruixiang Tang, Junyi Jessy Li, Elias Stengel-Eskin
arXiv:2510.07231v5 Announce Type: replace
Abstract: Socio-economic causal effects depend heavily on their institutional and environmental contexts. The same intervention can produce different, even o...
By Donggyu Lee, Hyeok Yun, Meeyoung Cha, Sungwon Park, Sangyoon Park, Jihee Kim
arXiv:2601.10085v3 Announce Type: replace
Abstract: Conversational dynamics describe how rapport, emotion, and goals change over a dialogue. These changes are important in therapeutic conversations,...
By Viet Cuong Nguyen, Nhi Yen Nguyen, Kristin A. Candan, Mary Conlon, Vanessa Rumie, Kristen Risola, Michael L. Birnbaum, Munmun De Choudhury
arXiv:2604.23069v2 Announce Type: replace
Abstract: Large language model (LLM) agents often struggle in long-context interactions. As the history grows, sliding windows and prompt compression may dro...
By Yating Wu, Yuhao Zhang, Sayan Ghosh, Sourya Basu, Anoop Deoras, Jun Huan, Gaurav Gupta
arXiv:2606.00981v2 Announce Type: replace
Abstract: LLMs can plan by generating schedules directly as a Planner, or by translating a task into a formal specification for a solver as a Formalizer. Mos...
By Jiayi Zhang, Jianing Yin, Ben Zhou, Li Zhang