The paper introduces PrecepTron, a 32‑billion‑parameter language model fine‑tuned with low‑rank adaptation to evaluate clinical reasoning in large language models (LLMs) at a physician level. It also releases GRAND‑ROUNDS, a benchmark of 9,217 scored responses from 160 clinicians across seven studies. Using PrecepTron, the authors replicate key findings from major medical AI studies and explore new questions about LLM diagnostic accuracy, demonstrating that fine‑tuned models can provide consistent, scalable physician‑level scoring.
By Thomas A. Buckley, Zahir Kanjee, Peter G. Brodeur, Byron Crowe, Anthony M. Pettinato, Aashna P. Shah, Adrian D. Haimovich, Liam G. McCoy, Daniel Restrepo, Jason A. Freed, Ethan Goh, Jonathan H. Chen, Laura Zwaan, Katherine E. Goodman, Daniel J. Morgan, Raja-Elie E. Abdulnour, Adam Rodman, Arjun K. Manrai
The study investigates how vision models represent objects compared to humans during physical reasoning tasks. By comparing segmentation outputs from several architectures (DINOv2, SegFormer, DeepLabV3+, UPerNet) to human behavior in time‑to‑collision and change‑detection tasks, the authors find that briefly trained models are too coarse, fully trained models are too fine, and an intermediate training stage best matches the coarse, volumetric bodies humans use. Larger models reach this intermediate alignment earlier, suggesting that resource constraints drive the emergence of human‑like representations in general‑purpose vision systems.
By Andrey Gizdov, Andrea Procopio, Lorenzo Caputi, Georgi I. Ivanov, Yichen Li, Daniel Harari, Tomer Ullman
ReSolve is a training‑free inference method that reuses candidate reasoning by selectively moderating generative outputs. It examines existing derivations when candidates disagree or lack a parseable answer, then incorporates new solutions into a bounded loop. On 130 competition‑mathematics problems, ReSolve achieves 100 and 99 correct answers with significantly fewer tokens than eight‑sample self‑consistency, while a controlled ablation shows that visible derivations improve accuracy.
By Bangji Yang, Jiajun Fan, Hongba Ma, Xi Zhu, Weizhi Zhang, Minghao Guo, Ye Li, Hamid Palangi, Jiaxuan You
The study investigates whether a second review by a different large language model (LLM) improves error detection in code, documentation, and analyses. Using 30 artifacts with 150 planted errors and 900 review sessions across three reviewer models, the authors find that a top-tier cross-model reviewer performs similarly to same-model reviews in fresh sessions, identifies partly different errors, and that a combination of a fresh same-model review and a cross-model review detects more errors than two fresh same-model reviews. However, the results do not conclusively separate the effects of model difference from reviewer capability, and a lightweight cross-model reviewer does not outperform same-model review.
whyItMatters":"The findings suggest that while cross-model reviews can enhance error detection, the benefit depends on reviewer skill and the specific review strategy, informing how teams might structure LLM-based code review processes."
By Tae-Eun Song
DynaFlow is a framework that enables transparent and flexible intra‑device parallelism by separating logical model definitions from physical execution schedules. It offers a flexible frontend with annotations for graph partitioning and a programmable interface for custom parallelism strategies, while its efficient backend handles asynchronous control/data‑flow, custom memory management, and remains compatible with CUDA Graphs and TorchInductor. The authors demonstrate that DynaFlow can integrate representative parallelism strategies into six state‑of‑the‑art ML systems with minimal code changes, achieving up to a 1.29× throughput improvement.
By Yi Pan, Yile Gu, Jinbin Luo, Yibo Wu, Ziren Wang, Hongtao Zhang, Ziyi Xu, Shengkai Lin, Baris Kasikci, Stephanie Wang
PhaseAT introduces a Fourier phase‑aware adversarial training framework to improve medical image domain generalization. By perturbing only the phase spectrum while keeping amplitude unchanged, the method stresses spatial organization under consistent appearance statistics, applying updates to the luminance channel and using a phase‑saliency mask to focus on influential frequencies. Experiments on two challenging medical datasets show PhaseAT achieves over 20% improvement in single‑source domain generalization compared to several state‑of‑the‑art methods.
By Ahmed Sharshar, Asif Hanif, Naveen Kumar Kummari, Mohammad Yaqub, Mohsen Guizan
RASteer is a training‑free method for concept erasure in text‑to‑image diffusion models that retains other concepts. It constructs a retain subspace from concepts to preserve, then removes components aligned with this subspace from the erasure direction using Retain‑Orthogonal Steering (ROS). Overlap‑Adaptive Calibration (OAC) further adjusts the removal of shared components at each layer and denoising step, balancing target erasure with concept preservation. Experiments show RASteer matches or outperforms existing activation steering and weight editing baselines on unsafe‑content, instance, and artistic‑style erasure tasks across multiple backbones and benchmarks.
By Yongliang Wu, Haori Lu, Yulun Wu, Jinqi Luo, Xingyu Zhu, Yaoyao Liu
MosaiChunk introduces a spatio-temporal memory mechanism that composes a mosaic of selected historical key-value entries across space and time to improve long-horizon autoregressive video generation. By keeping the video generator frozen and learning a lightweight router to select historical sections within a fixed memory budget, the method enables the generator to recover fine-grained visual details when objects or scenes reappear. Experiments on the RememBench benchmark demonstrate that MosaiChunk consistently outperforms sliding-window inference and whole-chunk retrieval in both text-to-video and image-to-video settings.
By Yiwen Zhang, Haocheng Xi, Michael Tian-Yue Liu, Alexei A. Efros, Hadar Averbuch-Elor, Qianqian Wang, Haiwen Feng
The paper introduces a deployment‑focused evaluation protocol for pedestrian tracking that isolates tracker performance by using shared detections, assessing initialization, gap continuation, identity recovery, close‑neighbor association, and runtime under load, while still reporting HOTA as an aggregate metric. Applied to the JackRabbot Dataset, the protocol reveals that trackers vary widely in capability, with none maintaining spatially correct, same‑identity output after 1 s of missing detector support, and that runtime on an NVIDIA Jetson Orin often falls below the 10 Hz target in crowded scenes. The authors provide open‑source code and scripts to enable reproducible assessment of tracker suitability for real‑world pedestrian‑centric robotics.
whyItMatters":"The protocol offers a practical, tracker‑only framework that directly informs the design and deployment of mobile robots operating among pedestrians by highlighting specific failure modes and computational constraints that aggregate scores overlook."
By Dominik Wojcikiewicz, Diego Paez-Granados
LangMap introduces a human‑verified benchmark for language‑conditioned goal navigation (LGN) that spans four hierarchical semantic levels—scene, room, region, and instance—within real‑world indoor 3D scans. The dataset, built on HM3D, contains 18K tasks with concise and detailed descriptions for 414 object categories, and its contrastive annotation protocol ensures high‑quality, discriminative region and instance labels. Evaluation shows that LangMap’s descriptions improve text‑to‑view matching accuracy by 23 points over GOAT‑Bench and achieve a 92.5% unique‑and‑correct match rate in an independent human audit, while a proposed RGB‑only baseline, PlaNaVid, attains top‑tier success rates without depth or 3D scene representations.
By Bo Miao, Weijia Liu, Jun Luo, Lachlan Shinnick, Jian Liu, Thomas Hamilton-Smith, Yuhe Yang, Zijie Wu, Vanja Videnovic, Feras Dayoub, Anton van den Hengel
The paper introduces a 3D keypoint detector that replaces traditional heuristic post‑processing with a learned suppression module. This module, implemented as a graph neural network over candidate keypoints, learns to keep, suppress, or relocate points, and is built on a Point Transformer backbone extended with a directional graph neural network. Experiments show the approach outperforms DBSCAN and greedy non‑maximum suppression, surpasses KeypointDETR on most KeypointNet categories, and achieves top performance on the Building3D benchmark while remaining competitive on the Tallinn split.
By Batuhan Arda Bekar, Can Sar{\i}, H\"useyin Can G\"ulkan, Bar{\i}\c{s} \"Ozcan
The paper introduces AEGIS, a buffer‑free, layer‑wise orthogonal gradient projection framework that enables continuous fine‑tuning of Vision‑Language Models for robotic manipulation while protecting pre‑trained visual reasoning. AEGIS uses pre‑training activation statistics as static anchors and applies a Wasserstein‑2 transport penalty to generate an anchor‑restoration gradient, followed by a dual‑backward pass that orthogonalizes task gradients against this restoration vector. Experiments on the LIBERO manipulation benchmark with PaliGemma2‑3B show that AEGIS preserves Visual Question Answering performance and baseline loss while achieving continuous action convergence, all without replay buffers, teacher models, or co‑training data.
By Guransh Singh
M$^2$Weather is a new benchmark for joint multi‑station and multi‑variable weather forecasting that gathers 2,809 high‑quality stations with five physically coupled weather variables across France, Europe, and the globe. The benchmark includes unified training and evaluation protocols and a lightweight adapter that can introduce missing station or variable relationships without retraining. Systematic evaluation of 16 models shows that jointly modeling station and variable relationships reduces mean‑squared error across all datasets.
By Rongwen Li, Xiao Wang, Mingyang Wang, Hongwu Liu, Changjian Chen, Zhuo Tang, Kenli Li
The paper "Where's Waldo? Query-language Preference under Cross-lingual Knowledge Disparities" introduces the Waldo benchmark, a multilingual QA dataset built from Wikipedia that focuses on knowledge gaps and conflicts across languages. It evaluates eight models in five languages and finds that when a fact is missing in one language, models tend to use evidence from the other language, but when conflicting accounts exist, responses align strongly with the query language, leading to different answers for semantically identical questions. The study also explores mitigation strategies, including ablating attention heads and LoRA-based training, which can reduce the preference gap by up to 61.5%.
By Dayeon Ki, Ruochen Zhang, Silviu Cucerzan, Ryen W. White, Ning Gao
The paper presents a matched‑budget audit framework for evaluating recaptioned image‑text supervision distributions, which are created by applying a captioning policy, captioner, and source corpus to generate dense descriptions for text‑to‑image training. The framework operates under a fixed text budget of 64 tokens and produces a five‑axis profile covering prompt coverage, image‑conditioned faithfulness, and caption surface quality, using controllable basic units as a common claim metric. The authors apply the framework to seven paired comparisons across five public corpora, demonstrating improvements in supported claim units and revealing consistent long‑vs‑dense trade‑offs across different judges and budgets, and they release the audited multi‑source recap corpus and audit artifacts.
By Giyeong Oh, Junghun Park, Yuhan Bae, Youngjae Yu
The paper investigates how the choice of reconstruction loss, specifically mean squared error (MSE), affects optimization in sequential post‑training quantization (PTQ) of large language models (LLMs). It identifies an Optimization Imbalance where MSE causes reconstruction loss magnitudes—and thus gradient magnitudes—to vary dramatically across quantization stages, leading to uneven parameter updates. The authors propose that root mean squared error (RMSE) variants, which implicitly normalize gradients, can decouple optimization strength from loss scale and outperform MSE as a drop‑in replacement.
By Chao Li, Shigeng Wang, Anbang Yao
The paper presents an automated framework for building ontology networks by discovering typed semantic links between concepts across domains. It uses domain‑adapted DistilBERT embeddings, clustering to prune candidates, and GPT‑4o for generating interpretable relationships, reducing 800k raw pairs to 95k high‑quality candidates. Human validation on 429 links shows an 80.19% precision and 0.890 F1, outperforming five similarity‑based baselines and demonstrating the need for LLM reasoning in ontology link construction.
By Nouha Hayouni, Sheeba Samuel, Alsayed Algergawy
HumanoidToolBench is an 18‑task benchmark designed to evaluate humanoid robots on tool selection, manipulation, and locomotion across three scenarios, execution levels, and tool‑set modes. It is accompanied by ToolBook, a dataset of 3.1k demonstrations collected in simulation and on a real Unitree G1 robot. Experiments with seven policies in simulation and three on the real robot show significant gaps between tool selection and task completion, especially on unseen tools and under unrelated instructions.
By Kyochul Jang, Seohyeon Park, Ohchul Kwon, Sangjun Park, Junhyeok Choi, Seungyeop Yi, Chaeyun Kim, Sangkyu Lee, Idan Szpektor, Avi Caciularu, Jongmin Park, Youngjae Yu
Argo-Bench is an evaluation framework that tests data agents on 210 enterprise-scale data science and analytics tasks, simulating a 2024 New York City food delivery platform with 81 million orders and 7.5 billion rows across 235 tables. Unlike traditional text‑to‑SQL benchmarks, the tasks require agents to navigate a realistic ERP warehouse, reconstruct facts, and perform actions such as banning accounts or allocating budgets, with outcomes scored by a simulator. Each task includes an executable reference solution that proves solvability using only the warehouse data.
By Gabriel Tomitsuka, Arman Raayatsanati, Emma Xing, Duke Gand, Joseph J Ma
The paper introduces Watch, Infer, Coordinate, a benchmark and inference method for robots to deduce a partner’s physical constraints by observing joint behavior in physically coupled manipulation tasks. It demonstrates that understanding these constraints enables zero‑shot coordination on new tasks, achieving performance close to an oracle with true constraint knowledge. The study focuses on scenarios where a robot’s limitations may be unknown to its partner, such as hardware degradation or actuator faults.
By Suyu Ye, Zheyuan Zhang, Vaishnav Tadiparthi, Hossein Nourkhiz Mahjoub, Ehsan Moradi Pari, Tianmin Shu, Homanga Bharadhwaj, Nakul Agarwal