The paper evaluates six structured decision models for hate‑speech moderation across four datasets, comparing them to specialized, zero‑shot, commercial, and supervised baselines. It investigates whether providing dataset definitions or decomposing them into multiple questions improves classification, and measures latency and cost. Results show that commercial LLMs outperform decision models on only one dataset, definitions alter up to 28% of predictions without consistent gains, and decomposition helps in only 20% of cases, yet the best hosted decision model approaches the best commercial LLM’s performance with roughly 97% lower inference cost.
By Demetris Paschalides, George Pallis, Marios D. Dikaiakos
The paper "Code2Math: Can Your Code Agent Evolve Math Problems Through Exploration?" explores how code agents can autonomously generate more complex variations of existing math problems. It introduces a multi‑agent framework that evolves problems while ensuring they remain solvable and increasingly difficult. Experiments show that with sufficient exploration, code agents can produce new, structurally distinct, and challenging problems, demonstrating a scalable method for creating high‑difficulty mathematical reasoning tasks.
By Dadi Guo, Yuejin Xie, Qingyu Liu, Weixian Huang, Jiayu Liu, Zhiyuan Fan, Qihan Ren, Shuai Shao, Tianyi Zhou, Jianjie Feng, Wenze Su, Yujiu Yang, Dongrui Liu, Yi R. Fung
The paper introduces Mawqif-XT, a new Arabic benchmark dataset comprising 996 manually annotated tweets from three public targets: Women Driving, E-Cars, and Trimester System. Each tweet is labeled for stance, sentiment, and sarcasm following the Mawqif annotation scheme, and the dataset is intended as a held‑out evaluation set to test cross‑target generalization. Baseline results are provided using Arabic and multilingual transformer models as well as zero‑shot large language models, enabling reproducible evaluation alongside the original Mawqif dataset.
By Rasha Albalawi, Nuha Albadi, Hamzah Luqman, Maram Kurdi, Saad Ezzini, Asma Yamani, Ahmed Ashraf
The paper introduces GAW-PO, a gradient‑aligned token reweighting technique for Direct Preference Optimization (DPO). It assigns weaker penalties to rejected tokens whose gradients align with preferred behavior, while maintaining stronger penalties for conflicting tokens. Across 11 benchmarks in mathematics, reasoning, coding, and question answering, GAW‑PO outperforms standard DPO and other baselines, and remains robust when the DPO regularization parameter is reduced.
By Andreea Dutulescu, Stefan Ruseti, Mihai Masala, Traian Rebedea, Mihai Dascalu
The paper introduces a simulation‑grounded vision‑language model (VLM) framework for wildfire monitoring that converts 2D wildfire simulations into labeled video episodes using a fixed Blender mapping to create low‑detail 3D proxies. These proxies, along with controllable video generation, provide a multimodal memory that a training‑free multi‑agent VLM system uses to retrieve reference episodes, reconcile visual and memory‑based predictions, and generate structured wildfire reports. The system achieves 77.3% accuracy on six simulator‑derived report fields, outperforming direct VLM querying and text‑only memory baselines.
By Duowen Chen, Yuchen Sun, Zhiqi Li, Yuxuan Liao, Sinan Wang, Bart van Bloemen Waanders, Bo Zhu
SpectralCache is a training‑free framework that accelerates diffusion‑based world models by exploiting the stability of singular subspaces and predictable evolution of singular values across denoising steps. It reuses these stable subspaces and estimates low‑dimensional singular values via linear extrapolation, while also skipping expensive backbone evaluations through singular value scaling. Experiments on models such as HunyuanWorld‑Voyager‑13B show a 5.22× speed‑up with no loss in generation quality, outperforming existing caching methods.
By Zhendong Mi, Pu Zhao, Ziyu Hu, Xiaodong Yu, Yanzhi Wang, Grace Li Zhang, Shaoyi Huang
The paper introduces “SpaceConflict”, a benchmark of 23,196 multimodal inputs designed to test whether large language models can not only report spatial facts but also use them in reasoning. Experiments show a gap between a model’s ability to recover an initial spatial state from visual evidence and its ability to apply that state to perform transformations, with the gap narrowing but not closing as model size increases. To address this, the authors propose Operational State Supervision (OSS), which supervises task‑relevant spatial states and their transformation trajectories, improving accuracy on judgments that require state organization and use.
By Jinchang Zhang, Guoyu Lu
The paper introduces Behavior Pack Optimization (BPO), a post‑training method for video multimodal large language models that replaces single‑response rewards with a set of outputs across counterfactual views. BPO enforces stability when interventions are irrelevant, sensitivity when key evidence is removed, and abstention when no evidence remains, using an anchor‑relative advantage to keep the objective stable with small pack sizes. Experiments on datasets such as TempCompass, MVBench, and NExT‑QA show that BPO improves macro accuracy and abstention metrics for models like Qwen2.5‑VL‑7B‑Instruct, with gains that transfer to other benchmarks and models.
By Zhaolu Kang, Shiyu Liu, Tailong Luo, Wei Zhang, Yingjie He, Lei Wei, Guansu Wang, Liang He, Siheng Wang, Guangyuan Dong, Jiaqi Su, Shuang Chen, Haoyu Ji, Qishi Zhan, Kaiyue Zhou
GeoScaffold introduces a geometric supervision framework that embeds 3D geometry into vision‑language navigation policies during training, eliminating the need for depth sensors or 3D encoders at inference. The method learns a compact depth tokenizer from training trajectories, then fine‑tunes the policy with geometry query tokens that reconstruct depth, connectivity, and traversability, turning these states into compact geometric latents for action decoding. After training, the tokenizer, generators, and reconstruction heads are discarded, leaving a lightweight backbone that outperforms vision‑only navigators on continuous VLN benchmarks.
By Yixuan Jiang, Wentong Li, An Liu, Zihao Xin, Fulin Tang, Cong Leng, Yang Gao, Jian Cheng
I2CD is a method that directly predicts a convex decomposition of 3D geometry from a single RGB image, bypassing the traditional slow reconstruct‑then‑decompose pipeline. It leverages a frozen pretrained diffusion transformer and shape decoder, adding only a lightweight cross‑attention head to output half‑plane parameters for K convex polytopes. The resulting geometry is compact, convex by construction, and can be loaded into physics engines in about 0.5 s per image, outperforming eight baseline pipelines in speed and volumetric IoU on benchmark datasets.
By Qian Wang, Liam Merz Hoffmeister, Brian Scassellati, Daniel Rakita
The paper investigates how affective fine‑tuning shapes the internal architecture of multimodal foundation models. By analyzing 13 model instances across nine designs, it finds that adapting the feed‑forward network (FFN) consistently outperforms attention‑only adaptation and nearly matches full‑model tuning, revealing the FFN as an efficient adaptation substrate. Moreover, joint optimization leads to emergent functional specialization, notably a prominent gate projection pathway, which the authors exploit in Gate‑Focused Efficient Tuning (GET) to achieve 96.2–98.0% of full‑model performance with only 19.3–24.5% of the trainable parameters.
By Zhen Zhang, Runhao Zeng, Sicheng Zhao, Xiping Hu
LiteTrajEval is a lightweight architecture designed for budget‑bounded trajectory evaluation of large‑language‑model agents. It creates compact domain‑specific rule profiles offline, preprocesses trajectories online to flag heuristic failure signals, and then uses a single rubric‑guided LLM judge to generate structured diagnostic reports. On public Magnetic‑One‑style and τ‑bench‑style datasets, LiteTrajEval improves failure‑localization alignment with human annotations by 20–35 percentage points and reduces cost and evaluation time by 6× and 8× respectively, and has been deployed in an enterprise agentic platform.
By Linh-An Phan, MingXue Wang, Guangyu Wu, Feng Pan, Zhaoyu Pang, Yanbin Zhang
The paper introduces a constraint‑aware training objective that externalizes program analysis constraints used during inference, avoiding redundant learning of these constraints during model training. It presents three theorems demonstrating that this approach can reduce model size and improve data efficiency. A controlled synthetic experiment confirms that constraint‑aware training achieves lower prediction loss than standard cross‑entropy training at the same parameter count and data volume.
By Jinwoo Kim
JOVE is an online framework that jointly assigns large language model (LLM) executors and selects intermediate outputs for paid verification in directed acyclic task graphs. It balances immediate execution costs with future learning by solving per-query mixed‑integer linear programs and updating task‑dependent LLM quality estimates through verification feedback. Experiments on four reasoning benchmarks show that JOVE matches standard inference accuracy while cutting average cost and latency by at least 3.17×.
By Haoran Zhang, Dongjun Kim, Seohyeon Cha, Kevin S Chan, Ananthram Swami, Gustavo De Veciana, Haris Vikalo
This thesis introduces methods for assessing and enhancing the robustness of large language models (LLMs) against adversarial input variations. It proposes a generative robustness metric, R_stab(f), based on Jensen‑Shannon divergence, and proves bounds for localized attacks. The work also presents adaptive attacks, Trojan detection techniques, defense strategies for multi‑layer systems, and lightweight attestation for Model Context Protocol (MCP) agentic systems, all implemented in the JudgeGuard, TrojanArmor, and MCPSec suites.
By Narek Maloyan
The study investigates how two computational dimensions—model depth and refinement steps—affect intelligibility and speaker identity in masked-diffusion text‑to‑speech systems. Experiments with 15 models (19–133 M parameters) and up to 16 refinement steps show that refinement improves intelligibility more than identity, with a 1.86× asymmetry that persists even after retraining. Best‑of‑K search can recover identity when refinement fails, and analysis indicates that depth and steps target distinct bottlenecks, requiring separate optimization.
By Nityanand Mathur, Hamees Sayed, Ayush Pratap Singh
The paper tackles the challenge of compressing diffusion language models (dLLMs) by addressing the mismatch between calibration on clean activations and inference on partially masked states. It introduces a trajectory‑aware low‑rank objective that considers corruption levels and masking realizations, and proposes Traj‑MC, a Monte Carlo method to efficiently estimate the trajectory second moment. Experiments show that trajectory‑aware calibration yields better reconstruction along the generation trajectory and preserves significantly more mathematical reasoning compared to clean calibration.
By Tian Liang, Zishan Shao, Yiran Chen
The paper introduces JIL, an attack that manipulates length predictions used by large language model (LLM) schedulers to gain higher priority and reduce completion time. By optimizing an adversarial suffix, JIL causes lightweight output-length probes to underestimate request lengths, leading to up to an 83.4 % reduction in predicted length and an average 1.53× faster completion in end‑to‑end serving experiments. The study also evaluates defenses, showing that grouping predictions into coarse intervals can mitigate JIL’s advantage and protect benign requests.
By Yuyang Dai, Rana Shahout, Mahmood Sharif
The paper introduces FLINT, a domain‑specialized Text‑to‑SQL system designed for complex financial databases where data is stored as opaque integer keys. FLINT employs a lookup agent to resolve natural‑language concepts, uses embedding‑based retrieval of query templates from a compact expert‑authored bank, and prunes large schemas by traversing foreign‑key chains. Evaluated on 359 questions across two production financial schemas, FLINT outperforms state‑of‑the‑art baselines and is already deployed in a financial data retrieval service.
By Arijit Sehanobish, Bruno Gomes Coelho, Guillaume Michel, Sophia Zhi, Valerie Faucon-Morin, Kristen Howell
The paper introduces LS-AR, a dual‑channel architecture that separates continuous goal steering from discrete token decoding using FiLM conditioning. It employs a Static Goal Encoder for long‑term macro‑objective retention and a Dynamic State Tracker for recurrent latent updates. Experiments show LS‑AR achieves perfect target recall on long‑horizon retrieval tasks, improves throughput by ~35%, reduces VRAM usage by 52.8%, and maintains higher completion rates in Blocksworld planning under perturbations.
By Anubha Gupta, Eduardo Pignatelli