Evolve Quick Reviews
Public Paper Reviews
Structured preliminary reviews of recent open-access papers, generated and continuously updated by the Evolve analysis engine. Each review condenses what the paper claims, how solid the evidence is, and the questions a rigorous peer reviewer would ask — English analysis with abstracts in your language. These are independent AI analyses, not peer review; the canonical record always lives at the publisher.
Local Edits, Global Ripples: Replay-Informed Policy Adaptation for Workflow Synthesis
Manqing Mao, Hong Wang, Samson Koelle, Jie Yuan et al.
RIPPLE gates prompt edits via advancing-policy replay, lifting Haiku validation +23.1pp on a 39-task synthetic benchmark; promising but small-scale.
Molecular representation shapes the balance between target fidelity and exploration in flow based polymer generation
Tianren Zhang
Claims representation choice governs fidelity vs exploration in latent-flow polymer generation; controlled, but validation is surrogate-only.
Competence-Gated Pooling of Language Models and Priors for Event Forecasting
Aditi Tiwari, Aashrith Bandaru, Heng Ji
Credible study: outcome-estimated domain gates give modest, significant Brier gains over global pooling, defer on strong markets; stacker parity caps novelty.
Beyond Generation and Accuracy: Diagnosing and Enhancing Visual Chain-of-Thought for Geometry Problem Solving
Zhitong Dong, Jicai Pan, Yingguo Gao, Jingting Ding et al.
GeoVAD-Bench diagnostic benchmark + GeoWeave-8B (+25.3pp accuracy): strong diagnostic design, but LLM-judge circularity and thin stats warrant caution.
BlueLM-GUI Technical Report: A Real-Device-Centric Flywheel for Self-Improving Mobile GUI Agents
Tong Ye, Kunyang Han, Guozhi Wang, Longqiang Luo et al.
Real-phone GUI agent: SOTA on own benchmark (87.4), 84.9 AndroidWorld; strong flywheel, but self-referential evaluation and a confounded ablation temper it.
Can LLMs in Draft-Verify-Revise Pipelines Resolve Deictic Ambiguity?
Obinna I. Ekekezie
Rigorous e-value experiment shows revise-stage LLMs often misresolve deictic shifts in pipelines; convincing, but rests on just 10 synthetic stimuli.
Agentic TCAD Calibration Workflow for Oxide Semiconductor Transistors
Gyujun Jeong, Junmo Lee, Sungwon Cho, Woohyun Hwang et al.
Claims first agentic TCAD calibration on a fabricated IWO transistor; rigorous transfer tests but no manual/optimizer baseline to prove agent value.
GAUGE: When Not to Trust LLM-as-a-Judge in User-Simulated Evaluation of Task-Oriented Agents
Umesh Bodhwani, Thanh Tran, Kai Wei
Satisfaction misleads LLM-judge release gates while rankings hold; credible audit with strong controls, but small panel and proprietary models limit reach.
Simulating Disengaged Students to Evaluate LLM-based Tutors
Xianghui Meng, Jionghao Lin
DAS² credibly operationalizes disengaged student states for tutor testing; honest about limits, but model confounds and judge dependence cap tutor-eval claims.
SoK: Rethinking Jailbreaking in the Era of Agentic AI: Attacks, Defenses, and Practical Consideration
Md Jueal Mia, Yanzhao Wu, Selcuk Uluagac, M. Hadi Amini
SoK taxonomizing 116 jailbreak attacks and 63 defenses for agentic AI; controlled study shows defense value is model- and attack-dependent; stats thin.
Affective Agent: On-Device Personalized Intervention Reasoning for Wearable Systems
Reina Mun, Zishen Wan, Vijay Janapa Reddi
On-device sub-1B LLM architecture for personalized wearable intervention is sound, but results are all simulator-derived; real-world value stays unproven.
UniPart: Towards Zero-shot Language-Grounded 3D Part Segmentation for Embodied Interaction
Xinqiang Yu, Zekun qi, Jiawei He, Wenyao Zhang et al.
UniPart+LangPart-1M 在 FIND3D 上带来了显著的零样本三维部件分割性能提升；对照研究令人信服，尽管评估中混杂了自建基准和单次运行结果。
LettuceVisSim: A Simulator That Generates Lettuce Image Time-series for Vision-Based Reinforcement Learning
Ziye Zhu, Bert van 't Ooster, Congcong Sun, Eldert van Henten et al.
Modular lettuce simulator generating labelled crop images; honest multi-level validation, but single cultivar and full-pipeline GCR R²=0.40 cap the claims.
ORQA: An Occupation-Realistic Question and Answer Framework for LLM Professional Knowledge
Shreyas Krishnan, Serina Chang, Abhishek Nagaraj
ORQA: automated pipeline yields a 480-item source-traceable occupational benchmark; scalable and honest, but 67% filter agreement and thin contamination power.
Agent as Policy for Robotic Manipulation
Mengzhao Jia, Yang Lin, Xixin Zhang, Zhihan Zhang et al.
Frontier coding agents as full robot policies show strong zero-shot real-robot success (80-100% in 7/8 configs) but lack baselines and statistical rigor.
A deep dictionary network-based foundation model for ultra-low-dose CT denoising
Baoshun Shi, Shuangyi Yang, Ke Jiang, Bin Zhu et al.
DDN foundation model for multi-organ ULDCT denoising claims SOTA with interpretable unfolding; credible, but no significance tests, split details, or code.
Bridging Thought and Action: Taming Long-Horizon Instability in Open-Source LLM Agents with a MetaTool-Enhanced ROS Framework
Kazi Abrar Mahmud, Nilotpaul Kundu Dhurubo, Tamal Kirttonia, Sabbir Hossain Ujjal et al.
MetaTool adds persistent pseudo-code planning to ROS-Agent, stabilizing open-source LLM robots; results plausible but small-n, no stats, one table error.
Causal multi-modal AI for personalized chemosensitivity prediction
Dhruva Biswas, Jeroen Berrevoets, Alec McClean, Linus Bao et al.
Causal AI predicts chemo benefit from H&E + clinical data (n=11,135); strong multinational held-out results, but proprietary code, retrospective only.
Direct Preference Density Alignment for Conversational Audio Equalization
Ioannis Stylianou, Sven Ewan Shepstone, Jon Francombe, Pablo Martinez Nuevo et al.
无代理的强化学习（Proxy-free RL），通过 KDE 偏好密度图加上用于 EQ 的 GRPO→DPO 混合模型；统计数据扎实，但主要性能提升仅在波动网格的最佳配置下成立。
I Am AdMan: A Pipeline for Automatic Generation of Personalized Advertising Imagery
Victor Kolominsky-Rabas, Leopold Müller, Claudius Budcke, Niklas Kühl
Proposes multi-agent pipeline turning customer data into personalized ad images; credible feasibility demo, yet no baselines, single-run and judge-limited.
Beyond Point Forecasts: A Survey on Probabilistic Forecasting for Time Series and Spatiotemporal Data
Donia Besher, Rajdeep Pathak, Madhurima Panja, Tanujit Chakraborty
Survey unifying probabilistic forecasting from stats to foundation models; strong taxonomy and guidance, but headline empirical claims unverifiable here.
Scaling Clinical Judgment to Evaluate Medical AI
Thomas A. Buckley, Zahir Kanjee, Peter G. Brodeur, Byron Crowe et al.
LoRA-tuned 32B judge reaches physician-level agreement on 3/5 tasks and reproduces 5 landmark studies; valuable benchmark, but 'across tasks' claim overshoots.
Offline Reinforcement Learning for Wind Farm Control: A Wind Tunnel Study under Dynamic Wind Directions
Yuhan Su, Hongyang Dong, Simone Tamaro, Filippo Campagnolo et al.
First wind-tunnel-validated offline RL yaw controller; ~10% gain over greedy, matches LUT benchmark, but single-run comparisons lack statistical support.
TRACE: Two-Stage Detector-Response Estimation With Angular Cosine Expansion for Ring Artifact Correction in Photon-Counting CT
Jigang Duan, Heran Wang, Ligen Shi, Zheng Sun et al.
Unsupervised fixed + DCT sinogram decomposition for PCD-CT ring artifacts; convincing measured-data results, yet thin statistics and implementation details.