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.
Support Thresholds, Not Algorithms, Limit Rare-Association Recovery in Co-Purchase Networks
Xiao Han, Zhen Zhang, Xin Zhao, Jiechun Lei et al.
Argues support thresholds, not mining algorithms, block rare co-purchase associations; NC filtering recovers 80–100% with persistent significance. Plausible and well-designed, but untested inference and no code require revision.
Small Enough to Know Everything: The Fully-Enumerable Transformer as an Instrument for the Science of Delayed Generalization
Yoshiyuki Ootani
Tiny-transformer 'grokking instrument' claim put to a preregistered 4000x scale test: 2 of 3 laws hold to 50M; rigorous, honest, limited to enumerable tasks.
A Noise Optimum in Rehearsal-Free Continual Learning: Isolation, Mechanism, and Scope
Gunner Levi Howe
Simulations show an inverted-U noise optimum in rehearsal-free consolidation driven by a σ²-coupled anchor gain; well-isolated but MNIST-scale, code unreleased.
Local Sparsity Enables Unsupervised LLM Safety Detection
Xin Chen, Gil Kur, Alexander Shevchenko, Andreas Krause
Proposes locally-masked SAE anomaly detection for LLM safety with theory and six-model validation; strong ablations, honest limits; gap to supervised remains.
QoS-Aware Federated Learning for Multimodal In-Cabin Interaction in Smart Vehicles
Baran Can G\"ul, Mert Nak{\i}p, Nasser Jazdi, Michael Weyrich
Proposes FedQoS, an event-triggered async FL framework cutting vehicular communication 76.7% for ~5% accuracy loss, but only vs FedAvg and without code/data.
SETTer: Sparse-Encoder Transformer for Long-term Multivariate Time Series Forecasting
Abraham Ezema, Chijioke Eze, Ferdinanda Ponci, Antonello Monti
Single-layer sparse transformer with hybrid masking claims SOTA on 8 LMTF benchmarks; results broad but single-seed, tie-inflated, efficiency unmeasured.
M$^2$PFN: End-to-End Disentangled Alignment for Generalizable Multimodal In-Context Learning in Alzheimer's Disease
Lujia Zhong, Shuo Huang, Jianwei Zhang, Xinyu Nie et al.
M²PFN repurposes frozen TabPFN for multimodal AD diagnosis via differentiable prompt tuning; strong external validation, though gains over MMPFN are modest.
MATCH: Model-Aware Tool Learning with Curriculum Scheduling and Hierarchically Gated Rewards
Shihao Liu, Hao Yin, Lijun Liu, Zhengzong Chen et al.
MATCH couples gated hierarchical rewards with a policy-tracking curriculum for RL tool learning; strong ablations, but single-run results lack variance tests.
Uncertainty-Gated Exploration Noise Suppresses Task Collapse in Online RL Fine-Tuning of a Flow-Matching Vision-Language-Action Policy
Mehmet Turan Yardımcı, Yunus Emre Çoğurcu
Claims uncertainty-gated noise suppresses per-task collapse in VLA RL fine-tuning; rigorous, narrow, 3 seeds/arm, no gain over BC baseline.
Evaluating Explanation Methods by the Predictors They Induce
Jacob Selb{\ae}k, Hugo L. Hammer
Turns explanations into parameter-free predictors to score PDP/ALE/SHAP/LIME; theory sound, controls strong; SHAP wins on dependent data, PDP if independent.
Blockchain-Enabled Artificial Intelligence and AI Agents for Secure Data Sharing and Cybersecurity Applications
Harsh Verma
Conceptual meta-synthesis linking the author's four AI-security studies to blockchain trust literature; coherent framing, no new data, heavy self-citation.
Correct Now, Insufficient Later: Auditing Update Sufficiency in Context Compression
Guangzhe Zhang
A candid pilot audit of update-sufficiency in compressed agent memory; unusually honest scoping and failure diagnosis, but synthetic data only and no external validation.
Past, Future, All at Once: Mitigating Stability-Plasticity Dilemma via Post-hoc JANUS Rectification
Zhilong Zheng, Letian Tao, Yang Guan, Yujie Yang et al.
事后雅可比零空间投影（Post-hoc Jacobian null-space projection）声称能消除 LLM 微调中的遗忘；一阶理论合理，但测试范围狭窄，且重放集与评估集存在重叠。
LastOPD: Taming Collapse in Latent On-Policy Distillation
Jie Yang, Zhengyu Fang, Zelin Xu, Jiarui Sun et al.
LastOPD claims latent OPD collapses after early gains and fixes it with a 10-step last-layer crossfade; strong diagnostics, but single-run, MATH-500-tuned.
Multimodal Routing and Region Refinement for Language-Guided Medical Image Segmentation
Md Maklachur Rahman, Md Hasan Al Banna, Saraf Anjum, Assame Arnob et al.
MRSeg adds routed LoRA + region refinement to text-guided segmentation; strong results, but the 0.30 Dice gain over prior SOTA lacks statistical testing.
Persuaded, Not Informed: Incentive-Misaligned Witnesses Defeat In-Context Grounding
Rahul Balakavi
Claims LLM agents are persuaded by self-interested CRM testimony, not missing info; controls are careful and honestly caveated, but causal proof is thin.
When Fancy Eviction Fails: Rethinking Cache Replacement For LLM Prefix Reuse
Yiyu Liu, Minlan Yu, Juncheng Yang
Finds sophisticated eviction rarely beats LRU for LLM prefix caching; convincing trace study with real vLLM validation, but narrow traces.
One Intervention per Component is Enough: Towards Identifiability in Linear Stochastic Dynamics from Steady State
Saber Salehkaleybar
Claims one intervention per SCC suffices to identify OU dynamics from steady state up to scaling; solid theory, but key genericity assumptions unproven.
Graph-Based Stochastic Power-UCT: Monte-Carlo Graph Search with Power Mean Estimation
Tung Tran, Viet Bao Mai, Hoang Ta, Tuan Dam
Adds power-mean backups to graph MCTS with depth-keyed DAG merging, proving O(n^-1/2) root convergence; proofs sound, conditional; experiments support gains.
Forecast-Dojo: Replayable Environments for Benchmarking and Training LLM Forecasting Agents
Liqin Ye, Haorui Wang, Fardin Ahmed, Rongzhi Zhang et al.
工程构建良好的可回溯预测基准；研究结果（研究工具对全部 12 个模型都有帮助，但全部逊于市场）是可信的，部分源于架构设计。
Amortizing Physics-Informed Neural Solvers via Graph Hypernetworks
Cheng Jing, Abhishek Verma, Kallol Bera, Yixuan He et al.
Operator-graph conditioning for amortized PINNs: 1D benchmarks support gains when coupling varies; honest CCP negative result; no code or significance tests.
Learning New Words from Unlabeled Test Data in Automatic Speech Recognition
Mengqi Wang, Mark A. Hasegawa-Johnson, Haolong Zheng, Chang D. Yoo
Proposes test-time unsupervised OOV word learning for ASR; results are real but modest, limited to recurring OOVs, and lack prior-method baselines.
Broadening Uncertainty Estimation for Audio Question Answering Across Methods, Formats, and Inputs
Aaron Isidore Grace, Weiran Wang
Broad audio-LM uncertainty study: cheap first-token confidence beats sampling and tracks audio evidence; solid breadth, but no CIs or significance tests.
Automatic Harness Evolution for Hardware Design Verification: Can LLMs Consolidate Gains Across Discovered Harnesses?
Kidus Seyoum, Ajay Mittur
Candid NVIDIA study: LLM-evolved harnesses lift completion and pass@5 on 12 proprietary verification tasks but gains don't consolidate; no significance tests.