The paper introduces SSS, a semi‑supervised framework that builds on the Vision Foundation Model SAM‑2 to improve medical image segmentation. It combines a weak‑to‑strong consistency regularization with a Discriminative Feature Enhancement mechanism and a prompt generator that uses Physical Constraints with a Sliding Window to supply prompts for unlabeled data. Experiments on the ACDC and BHSD datasets show that SSS outperforms prior methods, achieving a 53.15 Dice score on BHSD, a +3.65 improvement over the state of the art.
By Hongjie Zhu, Xiwei Liu, Rundong Xue, Zeyu Zhang, Yong Xu, Daji Ergu, Ying Cai, Yang Zhao
MedGEN-Bench is a new benchmark for open‑ended multimodal medical generation that addresses limitations in current medical visual benchmarks, such as query‑image misalignment, closed‑ended answer spaces, and text‑centric outputs. The dataset contains 6,422 image‑text pairs across six imaging modalities, 15 clinical tasks, and 27 subtasks, including VQA, image editing, and contextual multimodal generation pairs. Evaluation combines reference‑based fidelity metrics with a structured, checklist‑guided assessment by a medical VLM judge, and preliminary results show that image‑output tasks remain unsaturated while contextual augmentation improves image‑instruction similarity.
By Junjie Yang, Yuhao Yan, Gang Wu, Rui Qian, Zhisheng Chen, Haijiang Li, Yuhe Wu, Qichao Zhao, Dawen Tian, Xiang Wan, Fenglei Fan, Wenjian Qin, Yongquan Zhang, Feiwei Qin, Changmiao Wang
SegKAN is a new model for high‑resolution medical image segmentation that tackles fragmentation and noise in hepatic vessel CT scans. It replaces the standard embedding module with a novel convolutional network to smooth noise and avoid gradient explosion, and reinterprets spatial relationships between Patch blocks as temporal relationships to better capture positional dependencies. Experiments on a hepatic vessel dataset show a 1.78% improvement in Dice score over the current state‑of‑the‑art model, indicating that the new structure enhances segmentation performance for extended objects.
By Shengbo Tan, Rundong Xue, Shipeng Luo, Zeyu Zhang, Xinran Wang, Lei Zhang, Daji Ergu, Zhang Yi, Yang Zhao, Ying Cai
The study compares two ways of obtaining predictions from language models fine‑tuned on customer behavior: scoring answer tokens directly versus generating a written rationale and then scoring the resulting answer. Across 13 model‑domain cells covering four retail tasks, scored readouts consistently rank outcomes more accurately than generated readouts, with an AUC improvement ranging from 1.5 to 14.5 points. The authors also find that a third readout—eliciting a probability before any verdict—improves calibration but only when outcome rates are represented in training, and they recommend using generated rationales for interpretability while relying on scored heads for ranking.
By Touchapon Kraisingkorn, Krittin Pachtrachai, Wachiravit Modecrua
The paper introduces the Belief-State Engine (BSE), an inference module that supplies a large language model (LLM) with a Bayesian posterior over hidden states in a partially observable Markov decision process (POMDP). By keeping the raw action‑observation log hidden from the LLM, the BSE ensures the agent behaves as a sound Markov policy on the belief MDP, thereby inheriting classical POMDP optimality guarantees. Experiments on the Tiger POMDP and a red‑team attack‑graph task show that BSE‑augmented agents outperform six baselines in task return, belief calibration, and decision consistency.
By Arnab Chattopadhayay, Debdipta Halder
uFlowCSP is a MeanFlow-based crystal structure prediction model that learns the average probability‑flow velocity, enabling it to generate complete crystal structures in one to five network evaluations. It achieves inference speeds 5×–58× faster than diffusion and flow‑matching methods while matching or surpassing their performance, with a chemistry‑ and symmetry‑aware Transformer that uses canonical atom ordering and per‑token chemistry embeddings. On the MP‑20 benchmark, uFlowCSP attains comparable or higher accuracy with far fewer evaluations and significantly lower wall‑clock time, demonstrating improved accuracy per network evaluation.
By Sourin Dey, Dipannoy Das Gupta, Lai Wei, Sadman Sadeed Omee, Jianjun Hu
The study examined how clinicians and GPT‑4 evaluate emergency department revisit pairs to determine if further assessment is needed. GPT‑4 over‑identified nearly all pairs as requiring follow‑up, while clinicians used clinical gravity and differential diagnosis factors. An algorithm using a knowledge graph populated by an LLM achieved high positive predictive value for identifying pairs that clinicians deemed warranting review.
By Jonathan A. Handler, Marlene I. Robles-Granda, Jacob E. Mefford, Jeremy S. McGarvey, Gregory S. Podolej, Colleen J. Klein, Matthew D. Dalstrom, William F. Bond
The paper introduces HorizonMath, a benchmark of 113 largely unsolved mathematical problems across eight domains, paired with an open-source framework for automated verification. It focuses on the generator‑verifier gap, targeting problems that are hard to discover but easy to verify computationally, thereby avoiding costly formal proof verification or manual review. Using this framework, the authors found six novel solutions—three each from GPT‑5.4 Pro and GPT‑5.6 Sol—demonstrating that current models can contribute to mathematical research, while most state‑of‑the‑art models score below 10%.
By Erik Y. Wang, Sumeet R. Motwani, James V. Roggeveen, Eliot Hodges, Dulhan Jayalath, Charles London, Kalyan Ramakrishnan, Jakob Foerster, Cheng Zhang, Flaviu Cipcigan, Philip Torr, Alessandro Abate
The paper introduces REVA, a method for compressing retrieval-augmented generation (RAG) prompts by aggregating historical query–document–model interactions into reusable evidence views. REVA mines attention traces from the target generator, maps token-level attention to readable words, aggregates importance across repeated document accesses, and produces budget‑specific plain‑text views that maintain document order and the standard RAG interface. Experiments on four benchmarks with modern LLMs show that REVA improves generation quality by 1.0–5.8 points over existing compressors while reducing compression overhead by 5.3 to 15.6 times and adding less than 40 ms of latency.
By Tuan Nguyen, Qiran Hu, Banruo Liu, Khoa D. Doan, Kok-Seng Wong, Fan Lai
NCP-ArchPreview is a latent‑space language model that extends standard next‑token prediction (NTP) with a Next Concept Prediction (NCP) objective, allowing the model to predict discrete concepts spanning multiple tokens. The architecture builds a product‑quantized concept vocabulary from hidden states, uses a dedicated Concept Module to forecast future concepts, and feeds these predictions back to guide token‑level generation, all trained jointly end‑to‑end. Trained on 5.73 T tokens with 8.9 B parameters, it achieves the final pretraining loss of OLMo‑3‑7B using only 51.3 % of the tokens, outperforms OLMo‑3‑7B on downstream tasks (including a 5.99‑point GSM8K gain), and demonstrates that the learned latent space enables lightweight domain adaptation and improved drafting performance.
By NCP Team, Jiaqi Cao, Chiyu Chen, Shuang Cheng, Xu Cheng, Beiya Dai, Yufan Feng, Kewen Ge, Ruijun Ge, Jiayi Huang, Yang Jiao, Dahua Lin, Zhouhan Lin, Yifan Liu, Yuliang Liu, Biqing Qi, Mowen Ruan, Junzhe Shen, Yunchong Song, Hao Sun, Zhongbo Tian, Yixuan Wang, Rubin Wei, Jiaxin Xiong, Kangyu Yang, Qian Yao, Qi Zhang, Bowen Zhou
The paper introduces MUtE, a dual framework that simultaneously erases concept-specific information from representations and generates counterfactual mappings. By deriving new erasure functions based on optimal bounds, MUtE imposes a translational bias on counterfactual trajectories, aligning with geometric properties of concepts in language models. The authors demonstrate that this approach improves downstream algorithmic fairness and enables the generation of counterfactual texts.
By Antoine Saillenfest
LLMAR is a tuning‑free recommendation framework designed for sparse, text‑rich industrial B2B domains. It transforms user behavioral history into structured semantic motives using LLM inference, employs a reflection loop to self‑correct hallucinations, and operates cost‑effectively with asynchronous batch processing. Experiments on MovieLens‑1M, Amazon Prime Pantry, and a construction risk dataset show LLMAR surpasses state‑of‑the‑art learning models, achieving up to a 54.6% nDCG@10 improvement while keeping inference costs around $1 per 1,000 users.
By Ryogo Hishikawa, Ichiro Kataoka, Shinya Yuda
The paper investigates how to combine synthetic and real historical images to improve OCR for the endangered Manchu language. Using 60,000 synthetic and 20,306 real word images, the authors evaluate three vision‑language models and a compact CRNN across synthetic‑only, real‑only, joint, and sequential training regimes. Adding real data boosts word accuracy to 95–96%, and ensembling the best recognizers raises it to 98.27% without further training.
By Yan Hon Michael Chung, Hanlin Wang
DeFiFlowBench is a benchmark comprising 207 natural‑language prompts for synthesizing DeFi workflows, evaluating graph coverage, configuration completeness, and declared safety predicates, and testing trade configurations on a local EVM. The study shows that direct, constrained, and few‑shot prompting still yield unsafe executions, and that a slippage bound derived from a quote does not prevent price impact. The proposed Koan‑Safe system—combining a prompt‑only intent parser, a replaceable generator, and structural repair—achieves a higher static safety proxy score and records no unsafe executions on the benchmark, while ablation studies reveal the limits of default safety thresholds and the need for explicit trade protections.
whyItMatters":"The results demonstrate that current prompting methods can still authorize costly trades and that explicit safety mechanisms like Koan‑Safe are necessary to prevent unsafe DeFi workflow executions."
By Abhinav Rajeev Kumar, Harshit Arora, Varun Singh, Manikandan Nanjappan
Post‑training quantization compresses large language models by storing weights at reduced precision, introducing errors into hidden states that could accumulate with depth. However, pretrained models accumulate far less hidden‑state error than randomly initialized ones, largely preserving downstream performance. The study identifies two key mechanisms: (1) each layer’s new error tends to oppose inherited error, partially canceling it, and (2) the LM‑head geometry preserves high‑rank token scores, mitigating output changes.
By Yuxiang Chen, Michael Beyer, Jun Zhu, Jianfei Chen
The paper introduces a method that learns dynamic reaction representations directly from textual descriptions using a fine‑tuned language model coupled with Gaussian process surrogates. This approach enables multi‑objective Bayesian optimisation for chemical reactions, achieving faster convergence than traditional descriptor libraries or one‑hot encodings across nickel‑, palladium‑, and iridium‑catalysed systems. Prospective experiments on a palladium‑catalysed cyanation and an asymmetric hydrogenation produced high‑yield, high‑enantiomeric‑excess conditions after only two rounds of high‑throughput testing, translating directly to gram‑scale synthesis.
By Joshua W. Sin, David Ming Segura, Bojana Rankovi\'c, Siu Lun Chau, Marius D. R. Lutz, Andrea Anelli, Ryan P. Burwood, Kurt P\"untener, Maximilian J. Notheis, Raphael Bigler, Philippe Schwaller
The paper argues that AI deployment performance depends on interactions among compression, compiler transformations, and serving policies rather than just model architecture. It introduces a three‑layer taxonomy—model‑level techniques, compiler transformations, and system policies—and frames deployment as a constrained multi‑objective optimization problem over accuracy, latency, throughput, memory footprint, and energy. The authors propose an evidence protocol for comparable benchmarking and synthesize data from edge and data‑center platforms to show that cross‑layer interactions drive deployment outcomes, concluding with a constraint‑aware selection procedure and open research problems.
By Tejinder Singh, John Pflueger, Jeebak Mitra, Robert Lincourt, Mitchell Markow, Bhavesh A. Patel
OpenDiscoveryTrace is a public dataset of 558 complete AI scientific agent trajectories that records the reasoning process—thoughts, tool calls, observations, errors, revision triggers, and confidence—across 124 scientific tasks in drug discovery, materials science, genomics, and literature analysis. The dataset includes seven models (three frontier models and four open‑weight models) and 60 live‑retrieval variants, providing a balanced view of performance and error patterns. Pilot analysis shows that process traces reveal behavioral differences invisible to output‑only evaluation, such as differing error rates and types among frontier models.
By Aayam Bansal, Keertan Balaji
The paper introduces Multi-Agent Agentic Graph Learning (MAAGL), a framework that partitions a graph into communities and assigns a dedicated agent to each community for specialized reasoning. MAAGL addresses two key challenges in existing agentic graph learning: it preserves permutation invariance by summarizing structural evidence with a dynamic structural signature, and it controls context size by filtering semantic evidence to the top‑k relevant nodes. Experiments on four benchmark datasets demonstrate that MAAGL outperforms state‑of‑the‑art agentic graph learning methods.
By Liang Qu, Jianxin Li, Hua Wang
The paper introduces a black-box framework for evaluating agentic AI systems, focusing on multi-step vulnerabilities that standard single-turn tests miss. It presents a seven-domain taxonomy linking observable behaviors to risk categories, an automated SAGE-RT red-teaming process generating 120 adversarial scenarios per domain, and a human-validated evaluation using LLM judges. Empirical tests on CrewAI and AutoGen agents show significant governance, privacy, and behavior risks, demonstrating the framework’s ability to uncover critical architectural weaknesses without privileged access.
By Divyanshu Kumar, Nitin Aravind Birur, Tanay Baswa, Sahil Agarwal, Prashanth Harshangi