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.
OmniFysics-Nano-V2 Technical Report: Understanding the Physical World Across Modalities
Yizhou Liu, Jinghang Han, Kaixiang Qiu, Qi He et al.
4B omni model with physics-aware data and 2-stage GRPO claims 17/21 benchmark wins; strong but top omni rivals untested and 2 benchmarks self-published.
Towards Omni-dimensional GUI Agent Navigation with Masked Trajectory Prediction
Yan Zhang, Pei Fu, Daiqing Wu, Huawen Shen et al.
MaP unifies GUI navigation tasks via masked trajectory prediction; solid 2-3pt gains over mixture training on 5 benchmarks, but lacks significance tests.
Beyond Class Marginals: Bounding Rehearsal Gaps without Freezing Class Co-occurrence
Congren Dai, Nat Roongjirarat, Fei Ye
RPR bounds replay rehearsal gaps at 2⌈C/b⌉−1 with balanced marginals; theory tight, statistics careful, but accuracy gains are modest and regime-dependent.
Minimal Recurrent Behavioral Memory for Imitation under Partial Observability
Xianyao Li, Fang Xu, Rui Min, Ruitong Tian et al.
Rigorous info-theoretic characterization of minimal recurrent imitation memory with exact certificates; learning evidence mostly custom small-scale tasks.
CogenPVG: Cognitive-Enhanced Reflective Multi-Agent Framework for Persuasive Video Generation
Yuntian Xiao, Shoulong Zhang, Wenfeng Song, Yan Wang et al.
First ELM-guided multi-agent framework for persuasive video generation; novel task framing, but human-study evidence is weak: κ=0.33, no significance tests.
Neurosymbolic Action Model Learning under Partial Observability
Adem Kikaj, Lennert De Smet, Giuseppe Marra, Luc De Raedt
NeSyAM learns action models from partially observable traces; strong relevant-role recovery under occlusion, mixed full-model accuracy, no planning tests.
The Limits of Simulated Societies: How Post-Training and Survey Fine-Tuning Erase Cross-Cultural Variance
Rojin Ziaei
Claims LLM post-training causes 'consensus collapse' erasing survey diversity; credible diagnostic study, but 6 held-out questions limit generalization.
Reply to comments arXiv:2512.07881 and arXiv:2601.06104 on quantum structure in human and AI-generated language
Massimiliano Sassoli de Bianchi, Roberto Leporini
Conciliatory reply: concedes LLM protocol limits and BE-fit limits, defends 'entanglement as meaning' conceptually, but offers no quantitative re-analysis.
Video-HopChain: Multi-Hop Questions and Confidence-Gated Exploration for Video Reasoning Models
Trung Nguyen Quang, Yuhao Dong, Shuo Sun, Shuai Liu et al.
Multi-hop video RL data plus compute-neutral CGE lift Qwen3-VL-8B from 55.4 to 59.3 on 8 benchmarks; solid engineering, but single-run gains lack error bars.
A Leakage-Aware Multimodal Evaluation Framework for Early Intraoperative Acute Kidney Injury Prediction
Quang Minh Nguyen, Duc Minh Le, Ho Nhat Minh Nguyen, Thuy Quynh Nguyen et al.
Claims leakage-aware eval shows waveform-only models insufficient and stacked multimodal AKI prediction wins; the stats are thin (5 folds, 180 events).
Latest Exact Match Attention
Moritz Brösamle
LEMA binarized exact-match attention: word-RAM equivalence proofs, training to 834M, honest limits; theory novel but quality gap to softmax stays large.
Auditing Proxy-Based Validation Across Text Spans
Daein Weon, Dong Ho Kang
Careful span-boundary audit shows same-span proxy agreement can be surface containment; solid method, evidence limited to diagnostic scores on one model pair.
The Tasteful Agent: Measuring and Improving Taste in Long-Horizon Tasks
Wenbo Pan, Zhichao Liu, Shujie Liu, Jingying Zeng et al.
Proposes Taste-Bench: 502 auto-mined decision forks labeled by hindsight outcomes; careful protocol and distillation gains, but label validity gaps remain.
When Are Aggregate Agent Traces Diagnosable? Traffic-Governed Interpretation and Calibrated Abstention
Peiying Zhu, Sidi Chang
Claims a pre-registered two-gate diagnosability test for aggregate agent traces; rigorous simulation evidence, but external validity rests on one toy simulator.
Multi-View Fair Clustering Guided by Cross-View Sensitive Information Discrepancy
Mudi Jiang, Jiahui Zhou, Xinying Liu, Zengyou He et al.
Fair multi-view clustering via bias-ranked asymmetric alignment; ranks first in mean NMI, but fairness edge is thin and variance unreported.
You Only Need 2/3 of the Chosen Experts: An Empirical Study of Dynamic Expert Pruning in Fine-Grained MoE LLMs
Yuanteng Chen, Qiwei Lai, Chen Tianqi, Peisong Wang et al.
Empirical study: uniform 2/3 expert retention keeps 98.8% of MoE performance; dynamic pruning pays only under aggressive budgets. Convincing, solid work.
MorphoSHAP: Rethinking the Unit of Attribution in Explanation for Deep Visual Models
Anirudh Prabhakaran, Alexandre Rocchi, Gianni Franchi
MorphoSHAP makes morphological shapes the Shapley players for image attribution; strong insertion/deletion results, but faithfulness metrics need scrutiny.
CacheDyG: Decoupling Temporal Propagation for Efficient Dynamic Graph Learning
PinHeng Zong, Ye Yuan
Cache-decoupled dynamic GNN wins AP on 5 benchmarks with 19K params; solid vs snapshot baselines, but lacks continuous-time rivals and significance tests.
Sequential Confidence Sets for Coverage-Constrained Conformal Model Selection
Jing Li, Haibin Zhu
Proposes CC-SMCS: finite-sample time-uniform confidence sets for coverage-constrained conformal model selection. Sound theory; modest experiments, no code.
Protocol before progress: leakage-aware evaluation of AIS trajectory prediction
Zobeir Raisi, Vali Mohammad Nazarzehi Had
Claims protocol, not architecture, drives AIS prediction gains: oracle 2.1–3.2×, vessel leakage 23–25%, region collapse; rigorous two-corpus audit, convincing.
Beyond Reconstruction Error: Analytical and Data-Driven Action Tokenization for Autoregressive Vision-Language-Action Models
Yuxin Yang, Gaohan He, Changxue Guan, Hangming Liu
Claims reconstruction fidelity mispredicts VLA closed-loop control; careful controlled LIBERO study, honest stats, but 3-seed evidence stays suggestive.
ARAFA: An LLM-Generated Arabic Fact-Checking Dataset
Christophe Khalil, Shady Elbassuoni, Rida Assaf
Largest Arabic fact-check dataset (182k LLM-built pairs); valuable resource but test-set contamination and validator circularity weaken benchmark claims.
In-Context Guidance: Learning Inter-Task Synergies via Numerical Foundational Models for Few-Shot Multitask Optimization
Tingyang Wei, Haofeng Wu, Jiao Liu, Zhao Wei et al.
Frozen TabPFN guides inter-task coupling in few-shot multitask optimization via in-context priors; strong results, but no ablation isolates informed guidance.
MemoryAthena: Adaptive Routing over Latent and Generated Memories
Mingyuan Li, Guangsheng Yu, Juyuan Zhang, Xu Wang et al.
E-anchored routing blends generated memory into direct Engram retrieval; consistent, modest QA/NLP gains, but no significance tests and task-specific tuning.