The paper investigates Active Feature Acquisition (AFA), where features are selected sequentially to maximize predictive accuracy within a budget. It introduces a theoretical certificate that decomposes the gap between a greedy Expected Information Gain minus cost (EIG-Cost) policy and the optimal hard‑budget policy into a budget‑slack term and a greedy‑suboptimality term, δ_λ. Experiments on synthetic data, standard benchmarks, and a large clinical dataset (MIMIC‑IV) show that δ_λ is close to zero in many cases, and that EIG‑Cost achieves competitive performance while using only a small fraction of the full feature cost.
By Joseph Bingham
P2TPO introduces a structure‑aligned credit‑assignment framework for full‑page document parsing, aligning block‑level quality contrasts with generated spans and redistributing credit within matched blocks. The method employs Unified Anchor Pricing to handle matched and missing elements and Span Credit Redistribution to adjust token‑level credit while preserving overall block advantage. Experiments show P2TPO surpasses existing policy optimization baselines across six parsing benchmarks and scales well with model size and initialization.
By Yong Zu, Ninghua Yang, Zezhong Tan, Xinhong Ma, Ziqiang Dong, Yang Yang
LeanGraphSearch is a graph‑based retrieval method for the Lean 4 theorem prover that builds a dependency graph from Mathlib’s forward dependencies and traverses it to find supporting lemmas via multi‑hop chains. By linking each declaration to its premises, propagating relevance through a graph‑ranking algorithm, and re‑ranking candidates against the original query, it improves recall by up to 12% over LeanSearch v2 on MathlibQR and increases coverage of complete annotated proof routes on MathlibMPR by up to 3.7%. When integrated into an iterative proving loop, LeanGraphSearch raises pass rates at 8, 16, and 32 reflection turns by as much as 40% compared to a vanilla prover and up to 10% compared to LeanSearch v2, demonstrating higher accuracy‑cost efficiency.
By Yihan Xu, Jia-Chen Gu, Junyi Zhang, Di Wu, Haikang Deng, Po-Nien Kung, Kai-Wei Chang, Amit Sahai, Wei Wang, Nanyun Peng
The paper introduces DomSet-MIL, a multiple instance learning framework designed to recover evidence of coexisting lesions in whole-slide images when only the most severe diagnosis is weakly supervised. It models the dominant-label mechanism through a censored likelihood that treats the reported diagnosis as the maximum severity without assuming other lesions are absent, and employs severity-ordered residual routing with top‑q pooling to preserve secondary findings. Experiments on natural and controlled datasets show that DomSet-MIL outperforms standard MIL approaches in dominant-label prediction while better recovering latent coexisting-lesion evidence.
By Lubin Gan, Jing Zhang, Xinming Zhang, Chengpeng Xu, Siying Wu, Xiaoyan Sun
FutureLift is a refiner that transforms multimodal motion proposals into heterogeneous future graphs, linking temporal nodes to agent histories, map elements, alternative modes, and other agents’ future trajectories. Its bidirectional temporal edges enable later evidence to revise earlier segments in a single pass, producing residual coordinate and confidence corrections. Evaluated on Argoverse2 and the Waymo Open Motion Dataset, FutureLift consistently improves frozen proposals and outperforms other refiners, achieving top leaderboard rankings on the WOMD motion‑prediction benchmark without LiDAR or ensembling.
By Daniel Grimm, J. Marius Z\"ollner
The paper presents FANTF, a Fuzzy Attention Network-Based Transformer that integrates fuzzy logic into transformer architectures via a novel fuzzy attention mechanism inspired by Type‑2 fuzzy set theory. This mechanism employs a learnable Gaussian uncertainty perturbation to model uncertainty in attention relevance scores, producing softer, more diverse attention distributions and uncertainty‑aware interpretations. Experiments across five transformer backbones on real‑life datasets show statistically significant improvements in forecasting, classification, and anomaly detection, especially under noisy and non‑stationary conditions.
By Sanjay Chakraborty, Fredrik Heintz
HERO is a heterogeneous continual graph learning framework that preserves historical knowledge at both structural and semantic levels while adapting to new tasks. It uses DiSCo to select representative target nodes and reconstruct compact heterogeneous contexts, and applies multi-level knowledge distillation to align task-conditioned predictions with semantic responses. Experiments on four datasets with three HGNN backbones show HERO achieves best or tied-best average performance while maintaining competitive forgetting.
By Guiquan Sun, Xikun Zhang, Jingchao Ni, Dongjin Song
The paper introduces FPCD, a framework that stabilizes the modality gap in vision‑language models during continual learning. It combines Fast Null‑space Projection, which projects gradients onto approximate null spaces to reduce interference, with Contrastive Distillation that aligns image‑to‑text and text‑to‑image similarity distributions using a small reference set. On the 11‑task MTIL benchmark, FPCD achieves strong performance, maintaining stable modality gaps and superior zero‑shot retrieval on COCO Captions and Flickr30k.
By Tiantian Peng, Yuyang Liu, Shuo Yang, Qiuhe Hong, YongHong Tian
TabClustPFN is a Prior-data Fitted Network designed for clustering tabular data. It jointly addresses the lack of supervision, unknown cluster cardinality, and label switching by integrating a hybrid pretraining prior, a decoupled partition‑inference and cardinality‑inference architecture, and a permutation‑invariant SoftARI training objective. On a curated benchmark of 44 real‑world tabular datasets, it outperforms classical, deep, and amortized baselines while maintaining efficient runtime.
By Tianqi Zhao, Guanyang Wang, Yan Shuo Tan, Qiong Zhang
SEVerA is a three‑stage framework that synthesizes self‑evolving agents with formal safety guarantees. It introduces Formally Guarded Generative Models (FGGM) that wrap parametric models in verified rejection samplers, ensuring every output satisfies a first‑order logic contract. The framework searches for candidate programs, verifies them against hard constraints, and then fine‑tunes them to improve task utility while preserving correctness, achieving zero constraint violations on tasks such as Dafny verification, symbolic math synthesis, and policy‑compliant agentic tool use.
By Debangshu Banerjee, Changming Xu, Eugene Ie, Ming Zhang, Daiyi Peng, Chu-Cheng Lin, Tarun Suresh, Gagandeep Singh
The paper presents an end‑to‑end hybrid quantum‑classical workflow for sampling from discrete Markov random fields (MRFs). By precomputing the $2^n$ target probabilities classically, the authors compare amplitude‑encoded quantum sampling against several modern classical MCMC methods across 60 instances from five graph families, finding that classical samplers achieve mean effective sample size (ESS) ratios close to the quantum approach. Wall‑clock benchmarks show that exact inverse‑CDF sampling outperforms the quantum sampler by a factor of 36 on average, and that matrix product state (MPS) techniques yield far higher fidelities than variational quantum circuits (VQC) at comparable compression budgets.
By Arul Rhik Mazumder, Elias Towe
The paper investigates how few-shot on-device personalization of virtual assistants, specifically using LoRA fine‑tuning, affects the system’s ability to reject out‑of‑scope (OOS) inputs. It finds that naive personalization consistently lowers OOS‑rejection AUROC across five intent‑classification benchmarks while preserving in‑scope accuracy, partly due to miscalibration of the confidence scorer and partly due to a genuine shrinkage of embedding margins. The authors propose a low‑cost fix—scoring rejection with Mahalanobis distance instead of softmax confidence—that restores most of the lost OOS performance on most datasets without harming accuracy, and they release all code for reproducibility.
By Vaibhav Prajapati, Pushp Garg
The paper introduces SnapJudge, an image‑based AI detection system for math‑heavy scientific text, trained on screenshots of LaTeX‑generated PDFs. It shows that existing text‑only detectors like Pangram 4 lose accuracy as mathematical content increases, while fine‑tuned LaTeX detectors recover performance. SnapJudge achieves 92–93% detection rates at a 0.1% false‑positive rate, outperforming both Pangram 4 and text‑based models.
By Yixuan Huang, Danish Pruthi, Bhuwan Dhingra
VALCON introduces a role-driven framework to benchmark and improve value consistency in large language models (LLMs) by evaluating how well an LLM’s generative outputs align with its evaluative judgments when the same model alternates between generator and evaluator roles. The framework creates three benchmark suites—VAB-Consistency, VAB-Boundary, and VAB-Challenge—using dual-role variation to expose hidden value discrepancies. Experiments across 28 LLMs reveal widespread cross-role inconsistency, and mechanistic probing identifies a role-dependent representational decoupling; a representation-directed intervention is then proposed to mitigate these failures, establishing value consistency as a distinct alignment dimension.
By Junru Wu, Jingting Zheng, Yuqi Ren, Deyi Xiong
RenoLayout is a new framework that unifies floorplan renovation and furniture layout for residential design. It takes an existing residential layout and a set of design points, then generates a Geometry-as-Code program that preserves immutable structural elements while optimizing furniture placement under design objectives and physical constraints. The authors also introduce a Design Map with 123 fine‑grained design points, a Difference‑Grounded Design‑Point Annotation method, and a three‑stage training pipeline (CPT–SFT–RL) to improve spatial reasoning, instruction following, and constraint compliance. Additionally, they present the RenoLayout Benchmark, a collection of 200 real‑world before‑after renovation cases with complete furniture layouts, marking the first benchmark for unified residential design.
By Jingguo Liu, Yuanqing Liu, Chong Zhang, Yue Yang
The paper introduces PACE, a method that enhances multimodal reinforcement learning with verifiable rewards (RLVR) by using pairwise token contrast and adaptive trajectory confidence. Pairwise token contrast evaluates each token under four teacher views—combining correct/incorrect hints with image presence or absence—to isolate image-dependent support, while adaptive trajectory confidence normalizes teacher predictions to guide credit redistribution. Experiments demonstrate that PACE outperforms existing approaches on multiple multimodal reasoning benchmarks, achieving top accuracy across three backbones and the Qwen3‑VL‑8B‑Instruct model.
By Yiran Zhang, Hanbo Huang, Jielan Li, Zhen Zhao, Shiyu Liang, Hongxia Hao
The paper argues that machine unlearning, which seeks to erase specific information from large language models, is often evaluated only on a single benchmark distribution, leading to an incomplete assessment of privacy. The authors demonstrate that forgetting can differ significantly when the same information is queried in different languages and scripts across multilingual models, revealing a gap between behavioral forgetting and true inaccessibility. They call for broader evaluation that includes access‑path generalization and adversarial recoverability to ensure genuine privacy protection.
By Tyler Lizzo, Larry Heck
The paper introduces LURCH, a benchmark of 6,550 question‑answering instances that quantifies long‑context reasoning difficulty using three properties: evidence scope, evidence dispersion, and reasoning depth. Experiments with both proprietary and open‑weight models show that these properties are strongly negatively correlated with performance, that performance drops sharply for dispersed evidence as context length grows, and that difficulty increases when multiple properties are high simultaneously. The study demonstrates that context length alone is insufficient to explain model failures, and that the new metrics provide a more nuanced understanding of long‑context challenges.
By Shiwen Qin, Cameron Barker, Henry Gouk, Elliot J. Crowley, Shay B. Cohen, Mirella Lapata, Miao Li
The paper introduces a method for training large language models to generate diverse, high‑quality scientific ideas grounded in input documents. By using a set‑level reinforcement learning objective that rewards ideas for being distinct from earlier ones, the authors train an 8B model that outperforms Claude Opus 4.8 and GPT‑6 Astra on a contamination‑free benchmark. Human experts rate the model’s idea sets as equally promising while containing slightly more unique ideas, and automated experiments show comparable executability.
By Manya Wadhwa, Pranav Narayanan Venkit, Hiroaki Hayashi, Yu Li, Tuhin Chakrabarty, Chien-Sheng Wu
MagiC is a benchmark for grounded multimodal cognition that evaluates vision‑language models on answer accuracy, step‑by‑step reasoning quality, grounding fidelity, and self‑correction. It contains about 5,500 weakly supervised QA examples generated from strong model outputs and 900 human‑curated examples with fine‑grained annotations, including rationales and bounding box groundings. The benchmark tests 15 models ranging from 7B to 70B parameters, introduces new metrics such as MagiScore and StepSense, and includes diagnostic settings to probe robustness against adversarial visual cues and introspective error correction.
By Chengfei Wu, Ronald Seoh, Bingxuan Li, Liqiang Zhang, Fengrong Han, Dan Goldwasser