The paper investigates why supervised AI‑text detectors, specifically a RoBERTa‑based model, can be fooled by subtle changes in language. By applying semantic, structural, and tokenizer‑level perturbations to a large dataset and controlled Mistral‑7B‑Instruct outputs, the authors show that increasing verb diversity makes machine text easier to detect and that detection scores correlate with statistical complexity, leading to a high false‑positive rate on formal human writing. They also evaluate an event‑based latent space detector, finding that paraphrasing and homoglyphs significantly alter extracted event sequences and verbs, yet the detector’s performance remains modest (AUC 0.577).
By Claudiu Creanga, Liviu Dinu
The paper introduces a controlled anchor–stress protocol to analyze compositional failures in text‑to‑image diffusion models. It presents a text‑only Compositional Stress Index (CSI) that distinguishes common from rare compositions across SD1.5, SDXL, and SD3, and conducts a six‑prompt localization study showing how different denoiser interventions affect outcomes such as residual minimization and color hit rate. The study reveals that diagnostic signals appear before denoising, but the intervention that exposes risk is not always the one that improves generation, highlighting modality‑ and architecture‑dependent responses.
By Fangzheng Wu, Brian Summa
GISTBench is a benchmark designed to assess how well Large Language Models understand users by extracting and verifying user interests from their interaction histories in recommendation systems. It introduces two new metric families—Interest Groundedness (IG) and Interest Specificity (IS)—to measure the accuracy and distinctiveness of LLM-generated user profiles. The benchmark includes a synthetic dataset built from real user interactions on a global short‑form video platform, validated through user surveys, and evaluates a range of open‑weight and proprietary LLMs, uncovering limitations in their ability to count and attribute engagement signals.
By Iordanis Fostiropoulos, Muhammad Rafay Azhar, Abdalaziz Sawwan, Boyu Fang, Yuchen Liu, Jiayi Liu, Hanchao Yu, Qi Guo, Jianyu Wang, Fei Liu, Xiangjun Fan
The paper argues that continual learning for world models must differentiate between knowledge that should never be revised—such as physics and object permanence—and knowledge that should be updated when the environment changes. It critiques existing forgetting metrics and benchmarks for failing to capture this distinction, and proposes a differential retention approach that tracks invariant regression testing and revision latency throughout the adaptation process.
By Nishit Anand, Ramani Duraiswami, Dinesh Manocha
MaskCoFT introduces a masked co‑adaptive fine‑tuning approach for mixture‑of‑experts language models, training both routers and experts jointly with a cross‑entropy loss. A learnable binary mask limits each layer’s Top‑K routing to a subset of experts during fine‑tuning, allowing experts to adapt to the tokens they receive. In simulated GPU cache scenarios, MaskCoFT reduces expert fetches per token by 23.7% for Mixtral‑8x7B and 10.1% for DeepSeek‑V2‑Lite, and lowers inference time per output token by up to 16.4% and 5.5% respectively, while maintaining or improving accuracy across nine benchmarks.
By Junfeng Wu, Zehao Fan, Hadjer Benmeziane, Kaoutar El Maghraoui, Liu Liu, Yinan Wang
RNADynBench is a new standardized benchmark for RNA molecular dynamics, comprising 2,585 high‑quality 100‑nanosecond all‑atom trajectories with carefully controlled data splits. RNADynNet is a unified model that simultaneously generates RNA trajectories and extracts dynamics fingerprints from a single conformer, leveraging coordinate denoising, alignment, and physical grounding. The model achieves strong agreement with MD data, achieving RMSF correlations of 0.875 and 0.766 on test sets, including a high‑flexibility challenge set.
By Yiming Huang, Lennart Bastian, Hanqun Cao, Luis Vollmers, Tolga Birdal
The paper introduces AdsCVR, a benchmark for e‑commerce cross‑video reasoning with 2,483 videos and 6,110 QA pairs across six reasoning dimensions. It proposes AdSeek, an agentic framework that actively selects visual and audio tools during multi‑turn exploration and uses an offline trajectory rectification mechanism to improve reinforcement learning. AdSeek achieves 74.30% accuracy on AdsCVR, outperforming its backbone by 27.90 percentage points and generalizes to the open‑domain CrossVid benchmark.
By Jinghan Zhao, Yiman Hu, Liang Wu, Jian Xu, Bo Zheng
The paper introduces S$^{2}$-PINN, a stochastic separable physics-informed neural network that models the solution of random partial differential equations using a learnable Gaussian spatial dictionary, Fourier temporal features, and a generalized polynomial chaos stochastic basis linked by a low-rank Canonical Polyadic tensor core. Training employs a hybrid strong-form and gPC-projected residual loss, and theoretical analysis shows the separable class is dense in $L^2$ and that the projected residual matches a stochastic Galerkin constraint. Experiments on four manufactured benchmarks and additional Poisson, Darcy, Navier–Stokes, and inverse problems demonstrate that S$^{2}$-PINN achieves higher mean and variance accuracy, better calibration, and fewer parameters than nine baselines, supporting stochastic separability as an effective design principle for physics-informed neural UQ.
By Zhendong Li, Akwum Onwunta
The paper introduces SymCE, a dataset of 4,707 false undergraduate‑algebra and real‑analysis conjectures each paired with a deterministic Python verifier that can be executed to check truth. It studies counterexample generation by training a 4B‑parameter language model (Qwen3‑4B) first with supervised fine‑tuning and then with reinforcement learning (GRPO) using the verifier as a reward. The results show that counterexample‑only fine‑tuning can destroy true‑theorem recognition, while reinforcement learning with a sparse outcome‑only reward restores and surpasses baseline performance, achieving higher success rates than larger open‑weight models and competitive performance on other math benchmarks.
"whyItMatters":"The study demonstrates that reinforcement learning guided by a per‑theorem verifier can effectively repair the shortcomings of supervised fine‑tuning in theorem proving, offering a practical approach to improve language models’ mathematical reasoning capabilities."
By Omar Farouk Zouak, Houssam Eddine Boukhalfa, Soumaya Lakehal, Shiv Katiyar, Samia Nefti-Meziani
TasteBench is a multimodal benchmark designed to accelerate sustainable protein discovery by providing computational proxies for sensory prediction. It includes a food-level ranking task based on over 21,000 human evaluations of 215 plant-based foods across 24 categories, and a molecular-level taste classification task covering 15,000 flavor molecules. The benchmark offers baseline models, characterizes inter-rater agreement and reliability limits, and demonstrates that the best model achieves pairwise accuracy comparable to individual human panelists.
By Anna T. Thomas, Sohum Patnaik, Caroline Cotto, Benjamin Sanchez-Lengeling
The paper presents a method for zero‑shot cross‑subject continuous affect regression using synchronized EEG‑fNIRS data. It decomposes affect trajectories into a shared component across subjects and an individual component derived from a label‑free alpha‑band cross‑channel synchrony marker, which rescales the shared trajectory for each test subject. Extensive validation—including leave‑one‑subject‑out correlation, functional‑form comparison, component ablation, and ceiling analysis—shows the approach achieves lower mean absolute errors than EEGNet and ASAC‑Net baselines on unseen subjects.
By Xuan Wang, Bing Wang, Shuai Chang, Hao Yuan, Xinbo Qi, Xinyue Zhang
S2S-JEPA is a new AI weather model that focuses on predicting only the slowly varying components of the atmosphere at subseasonal-to-seasonal timescales, from two weeks to two months ahead. It adapts the Joint-Embedding Predictive Architecture (JEPA) from computer vision to discard unpredictable fine-scale details, thereby addressing the long‑known predictability desert. The model matches the skill of the ECMWF physics‑based ensemble and even outperforms it on several metrics during weeks 5 to 6.
By Chenyu Dong, Gianmarco Mengaldo
HakushoBench is a Japanese chart and table visual question answering benchmark created from 33 governmental white papers, comprising 2,053 images across more than ten types. The dataset includes manually annotated and independently verified QA pairs that evaluate holistic understanding of charts and tables rather than just local visual cues. Experiments show that HakushoBench is significantly harder than existing Japanese benchmarks, with sub‑10B open‑weight models achieving at most 58.6% accuracy and even large models like Qwen3.5-397B-A17B lagging behind Gemini‑3‑Pro by 8.1 points.
By Issa Sugiura, Shuhei Kurita, Yusuke Oda, Naoaki Okazaki
The paper presents a method for improving cross‑lingual retrieval in multilingual encoders by adding a query‑only low‑rank adapter and deterministic routing based on language. The adapter, trained against frozen document embeddings, is used for cross‑lingual queries while the original encoder handles same‑language queries, preserving existing performance. Using the SampoTron LoRA adapter with the Nemotron‑3‑Embed‑1B model, the authors report a 20.9% relative gain in nDCG@10 across six English, Finnish, and Swedish directions on a financial benchmark, with all cross‑lingual directions improving and same‑language performance maintained.
By Akshay Jain, Edward Kim
The paper introduces Function‑Space Guided Multimodal Optimization (FGMO) to address modality imbalance in multimodal learning. FGMO uses Functional Progress Estimation (FPE) to measure each modality’s function‑space response and Functional Response Control (FRC) to redistribute learning‑rate budgets, aiming to balance optimization progress across modalities. The authors provide theoretical analysis of FRC’s target‑contraction property and demonstrate FGMO’s effectiveness on several multimodal benchmarks.
By Zhongjing Gu, Fengqiang Wan, Yiming Cui, Yufa Feng, Yang Yang
The paper introduces the Hop-Decayed Influence (HDI) attack, which targets auxiliary schema-level structures—semantic summaries, hierarchical edges, and pre-computed scores—used in GraphRAG pipelines for retrieval prioritisation. By propagating query-aware influence, HDI identifies high-impact targets and corrupts a minuscule fraction (as low as 0.016%) of these structures, achieving an 88–94% success rate across HotpotQA and 2WikiMultiHopQA benchmarks on Microsoft GraphRAG and HippoRAG2 architectures. The modifications affect up to six queries each, yielding a 1:N amplification that instance-level attacks cannot achieve, and they evade perplexity and paraphrase defenses with over 99% evasion, exposing a structural blind spot in current GraphRAG defenses.
By Jisung Park, John Le, Heath Cooper
The paper introduces Multimodal Hypergraph Flow Matching (MHG‑FM), a framework that jointly generates structural connectivity (SC) and functional connectivity (FC) by constructing modality‑specific hypergraphs and learning higher‑order representations with Hypergraph Neural Network encoders. It employs Dual Cross‑Attention for bidirectional cross‑modal fusion, a variational autoencoder to map fused representations into a latent space, and conditional flow matching to synthesize connectivity and perform multimodal translation. Experiments on the Human Connectome Project Young Adult dataset demonstrate that MHG‑FM outperforms state‑of‑the‑art baselines in reconstruction quality, topology preservation, distributional similarity, and SC‑FC coupling, while achieving roughly eight times faster sampling than a comparable diffusion backbone.
By Chyong Yi Poh, Hwa Hui Tew, Junn Yong Loo, Rapha\"{e}l C. -W. Phan, Fuad Noman, Pew-Thian Yap, Chee-Ming Ting
The paper introduces scEGFlow, an energy‑guided flow matching framework that predicts how single cells respond to perturbations by modeling continuous transitions from control to perturbed states. It uses conditional flow matching and condition‑specific energy gradients to dynamically adjust predictions without retraining, achieving superior performance on imaging and transcriptomic benchmarks, even for unseen perturbations or limited data. The method preserves cellular manifold geometry, heterogeneity, and accurately recapitulates gene‑expression regulatory patterns.
By Jianan Wei, Jiajun Hong, Guikun Chen, Ning Yang, Lifeng Fan, Wenguan Wang
The paper introduces Prefix‑Certificate Retention (PCR), an exact optimization framework for deciding which prefix states of a language model to cache in order to balance recomputation and storage time. PCR models usable‑prefix rewards as nodes with prerequisites tied to timeout thresholds and preceding hit certificates, reducing the problem to a single minimum‑cut graph whose size scales linearly with block lookups and timeout choices. The authors provide a breakpoint theorem that extends the construction to all nonnegative timeouts without discretization error, a linear‑time dynamic program for ordered timeouts, and empirical validation on 39,632 public Mooncake requests, showing that ordered timeouts achieve the unrestricted optimum in most trace‑grouping cases and that heterogeneous retention can improve memory‑time tradeoffs.
"whyItMatters":"The study offers a tractable, auditable optimization model for retention policies that directly measures usable prefix blocks and storage time, providing a concrete benchmark for evaluating memory‑time tradeoffs in language‑model serving."
By Shivam Gupta
The paper investigates where to insert Cross‑Attention Layer (CAL) blocks between a system prompt and text in a frozen causal‑decoder backbone. A sweep on a 1.5B model shows that later placements of CALs are generally more effective and parameter‑efficient, while an 8B scaling study confirms that the impact remains task‑dependent, affecting instruction‑following and security behavior without significantly harming general performance. These results highlight placement as a key design choice when injecting system‑prompt information via cross‑attention.
By Li Lixing