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.
Learning Defensive Policies against Diverse Inference Attacks for Smart Meter Privacy
Ruichang Zhang, Mustafa A. Mustafa
Proposes HRL battery defense raising unseen NILM attackers' RMSE 107-166%; evaluation broad and significant, but theory is heuristic and code not yet released.
RAG-NAROK: Retrieval-Aware Knowledge Corpus Poisoning in RAG with Source-specific Refutation
Abdullahil Kafi, Alvi Ataur Khalil
Context-adaptive RAG poisoning showing strong retrieval gains over a naive baseline, but evaluation omits generator/judge models, detector calibration, and strong baselines.
ShowTellArena: Evaluating Business Workflow Understanding from Demonstrations
David Garg, Ritobrata Sarkar, Ehsan Azarnasab, Siddhartha Borah
Benchmark of agent comprehension after narrated business demos; dataset unusually transparent, but pilot favors the authors' own product—exploratory only.
Terminal Shrinkage Averaging Reveals a Schedule-Estimator Interaction in LLM Pretraining
Adam Ousherovitch, Yixin Wang
TSA: shrinkage blend of final iterate and checkpoint average. NanoChat experiments support schedule-estimator interaction; effects small, scale preliminary.
FAST-ML: A Hybrid Physics-Machine Learning Framework for Tropical Cyclone Intensity Forecasting
Shijie Xiao, Jonathan Lin, Thomas Ehrmann, Ali Sarhadi
FAST-ML embeds a neural ventilation closure in a differentiable hurricane model; gains over baseline are solid but FNV3 claims rest on selected cases.
Hill Sampling for Test-Time Scaling: A Simple and Better Alternative to Repeated Sampling, Evolution, and Training
Jacob Beck, Philip V. Ogren, Ari Kobren
Claims simple best-so-far sampling beats evolution/training for LLM discovery; convincing qualitatively, but SOTA margins sit at float-precision limits.
SAM-V: Geometry-Aware Segment Anything for Multi-View Instance Segmentation
Jiangshan Gong, Yuqun Wu, Qiqian Fu, Yao Xiao et al.
Strong SAM+VGGT fusion for multi-view segmentation (+5.7 T-mIoU vs PanSt3R; clear ablation), but single-run, small-N benchmarks without variance temper claims.
Queer inclusion in speech datasets: An audit and taxonomy of practical tensions
Brooklyn Sheppard, Anaelia Ovalle, Adina Williams, Levent Sagun
Audits 8 speech datasets, finding 0–1.4% queer speakers, and proposes a 4-way tension taxonomy; descriptive but a solid, timely audit with thin statistics.
RGSQ: Riemannian Geometry-Sensitive Quantization for Large Vision-Language Models
Zhiping Wu, Dongdong Ren, Yangchengyu Zhou, Zhengjie Zhang et al.
Proposes modality-aware Fisher-metric PTQ for VLMs with strong low-bit gains, but table label inconsistencies and missing rotation baselines weaken evidence.
Towards participatory speech dataset curation: A queer case study and conceptual framework
Brooklyn Sheppard, Anaelia Ovalle, Adina Williams, Levent Sagun
Position paper: a four-phase participatory framework for queer speech data curation — timely and well-motivated, but unvalidated: no pilot, no evidence.
NeuroRule: Making Black-Box Neural Networks Explainable through Rule-set Evolution
Tapaswini Kodavanti, Hormoz Shahrzad, Risto Miikkulainen
NeuroRule evolves concise rule-sets distilling NN behavior; promising OOD gains on 3 UCI sets, but baselines are the authors' own system and stats are weak.
Universal Fractal Natural Language Decision Map: Real-Time Edge Triage Across Heterogeneous Domains
Volkan Dağlı, Zerrin Dağlı, Dağhan Dağlı
Proposes weight-free Mandelbrot-boundary triage engine; headline 93.1% risks in-sample seed tuning, and 0% injection bypass rests on 10 attack classes.
SMTB: Fast Structure-Mapping with Tight Bounds
Daniel Weitekamp, Christopher MacLellan
SMTB claims 5-15x speedups and ~50% better mappings than SME in large domains; dual-objective eval on 5845 pairs is persuasive, but timing mixes C++ vs Lisp.
Continuous Optimization for p-adic Models
Julian Salazar, Dimitri Kanevsky, Matt Harvey, Pascal Getreuer et al.
Claims first continuous gradient descent for p-adic models via Berkovich hull Γp; careful math and controlled small experiments, but no convergence guarantees.
Matryoshka attribution: Learning to attribute language model outputs to representations and weights
Aryaman Arora, Kirill Acharya, Nathan Hu, Yanzhe Zhang et al.
MAttr learns one nested attribution ordering via sigmoid top-k; tops MIB (5.6 vs 1.95) and ablates refusal with 1% of weights; rigorous, partly self-metric'd.
West-WRF AI 2-km: High-Resolution Prediction of Integrated Vapor Transport and Precipitation
Nazak Rouzegari, Vesta Afzali Gorooh, Agniv Sengupta, Phu Nguyen et al.
Claims first 2-km stretched-grid AI forecast skill for western-US AR precipitation/IVT; evidence is broad and honest but confounded and short-period.
IndustrialVLA-Bench: A Traceable Multi-Axis Evaluation of Open Robot Policy Models
Yiqi Wang, Zhifeng Rao, Jiaqi Zhang, Xiaoyang Li et al.
Six-model VLA/WAM benchmark shows clean LIBERO near-saturated (1.6 pp) while robustness/paraphrase separate 15–31 pp; rigorous evidence tiers, but n=3 runs.
Compressing Long Context into Answer-Aligned Memory Embeddings for LLM Inference
Md Mostafizer Rahman, Md Faizul Ibne Amin, Md Shahajada Mia, Yutaka Watanobe et al.
CMC compresses context into answer-aligned embeddings (+7.3 EM on SQuAD), but gold-centered baselines and inconsistent ablation numbers weaken the claims.
LWCal: Loss-Weighted Calibration for Tabular Classifiers with Noisy Calibration Labels
Zeming Liu, Hang Lyu, Jingtao Zhang, Yuan Xie
LWCal re-weights noisy calibration labels for isotonic calibration; simple, honestly evaluated, but only synthetic noise on small tabular benchmarks.
A JEPA Recipe for Tabular Foundation Models
Mingyu Jeon, Suwan Cho, Jae Young Suh
Claims a JEPA term can train beside a tabular value loss without collapse yet trails value-only at convergence; credible single-seed negative result.
DefaultGNN: A Dual-Perspective GNN Framework for Predicting Corporate Default from Buyer-Seller Transaction Networks
Junghoon Kim, Hyunsung Kim, Seungyoon Choi, KyoungYong Park et al.
Claims dual buyer/seller-view GNN lifts corporate default prediction; large Korean tax-invoice study is solid but proprietary-data-bound with modest gains.
Weakly Supervised Quantum Error Mitigation
Seyed Mohamad Ali Tousi, G. N. DeSouza
Claims label-free QEM via 16 weak-supervision signals beats analytical baselines on 2 IBM 5-qubit devices; solid stats but scale and attribution unproven.
Synthesis and editing of multi-instrument audio mixtures using scalar-quantised latents with MIDI Span conditioning
Sungkyun Chang, Keshav Bhandari, Simon Dixon, Emmanouil Benetos
MIDI-guided mixture synthesis/editing via SQ latents + flow matching: strong metrics and honest analysis, but no listening tests and weak deletion control.
HABILIS Brain 0: Geometry-Change Supervision for Vision-Language-Action and Residual Flow Recovery
Jinu Pahk, Jesoon Kang, Taegeon Park, Jisu An et al.
GC-VLA claims geometry-change supervision lifts LIBERO to 95.20% and 99.55% with GCRF residual flow; ablations support the target, but single-seed eval and a null reader ablation temper the headline.