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.
Reading the News: Adapting Large Language Models to Swedish Journalism Through Continued Pre-Training
Lukas Borggren, Jenny Kunz, Marco Kuhlmann
Credible CPT study for Swedish journalism: 8.5B-token corpus, new 6-task benchmark; replay+LoRA key, but proprietary data/models limit reproducibility.
MolLedger: An Additive Graph Neural Network with Chemically Grounded ADME Attributions
Christina X. Ji
Additive GNN with anchor loss yields exact, chemically grounded ADME attributions at matched accuracy; promising, but faithfulness evaluation is circular.
TaxCE : A Framework for Automated Taxonomy Construction and Evaluation at Scale
Sandeep Sricharan Mukku, Albert Aristotle Nanda, Rohit Pyati
TaxCE automates multi-level taxonomy construction with EEG metrics, beating six baselines by 11.8–20.5 pp; circularity concerns and no code temper the claims.
REER-PT: Reverse-Engineered Reasoning for Perplexity-Guided Pre-training Data Augmentation
Haoran Que, Jiajun Shi, Ting Huang, Renming Pang et al.
Perplexity-guided reasoning annotations augment pre-training data; PPL drops and 680M-model gains look strong, but single-run results lack significance tests.
Staged Linguistic Seeding: Grounded Query Expansion for Verified-Unit QA in AI Contact Centers
Hyeonseop Yoon, Jeong-Eun Park
Honest application study: human-seeded LLM expansion lifts FAQ retrieval R@1 +0.27/+0.34 on private contact-center data; strong controls, in-distribution eval.
Three Steps at a Time: Learning Representations from Action Sequences in Contrastive RL
Michal Korniak, Kamil Dybek, Benjamin Eysenbach, Marco Bagatella et al.
将对比强化学习（Contrastive RL）扩展至动作块（action chunks），在离线（+31.7%）和在线（+93.1%）任务中均带来性能提升，并提供了可信的信息机制；但性能提升主要归功于与 CRL 自身的对比。
Where Identity Lives: Localized, Retain-Free Identity Unlearning in Multimodal Large Language Models
Kangwook Ko, Jaehyuk Jang, Wonjun Lee, Hee-Seon Kim et al.
Proposes localized retain-free MLLM identity unlearning (PAVA); strong ablations, but single-seed results and test-set hyperparameter tuning limit certainty.
Three Ways Classical Test Theory Misleads for LLM Judges
Louis Yiven Zhu
Shows KR-20, Φ(λ) and Livingston–Lewis lose attribution when applied to LLM judges; rigorous and honest, but shown on one small bank with one judge.
SwarmBench: Can Large Language Models Act as Agent Swarm Orchestrators?
Jinshan Gao, Zhuoran Jin, Tianyi Men, Kang Liu et al.
SwarmBench: a new benchmark for LLM swarm orchestration, plus SwarmExp; sound multidimensional eval, but in-distribution experience reuse weakens gains claim.
Contrastive Knowledge Distillation for Anomaly Detection in Multi-Illumination/Focus Display Images
Jihyun Lee, Hangil Park, Yongmin Seo, Taewon Min et al.
KD-based AD with class-flipped contrastive loss and channel blending hits 0.99 AUROC on a small private display dataset; promising but unverifiable.
PyKEEN-NSX: A Modular Framework for Static, Dynamic and Schema-Aware Negative Sampling in PyKEEN
Ivan Diliso, Nicola Fanizzi, Claudia d'Amato
PyKEEN-NSX factors negative sampling into pool+selector, unifying six samplers; useful availability diagnostic, though evaluation spans one model, one dataset.
What It Costs to Compose, Rebuild, and Correct Precomputed Memory
Asa Shepard
Maps KV-cache/cartridge memory failure modes (composition, rebuild, correction) with rare pre-registration rigor; single-model, ≤24k-token scope limits reach.
OCR-MetaReasoning Benchmark: Evaluating the Meta-Reasoning Ability of MLLMs in Text-Rich Image Understanding
Gengxu Li, Yuan Wu, Yi Chang
Rigorous 1,500-sample OCR meta-reasoning benchmark finds deduction and layout hardest for MLLMs; contamination and abductive-only edits limit type comparisons.
Fine-Grained Multi Image Object Hallucination Benchmark
Joonki Min, Chaeyun Kim, Hyungwook Choi, Yejin Kim et al.
MIOH probes multi-image object hallucination (4 tasks, 3 patterns, 29 models); sound design and findings, but no significance tests or release statement.
Geometry of Divergence: Tracking Hidden-State Trajectories for Adaptive Multi-Turn Reasoning
Jie Liang, Zhengxin Yu, Hamid Nasiri, Peter Garraghan
Hidden-state curvature/slope separate agent success pre-completion and lift τ-Bench 24.1%→39.6%; promising, but threshold tuning is under-reported.
MedAgent-R1: Faithfulness-Aware Reinforcement Learning for Evidence-Grounded Medical Reasoning
Jiangwang Chen, Chenghao Zhang, Hengxing Cai
Proposes faithfulness-gated RL that cuts citation fabrication 31.8%→4.7% at unchanged accuracy; well-controlled and mostly convincing, but LLM-judge-dependent.
Season-Aware Hybrid Convolutional-Transformer for Antarctic Sea Ice Concentration Forecasting
Danyang Li, John Taylor, Thang Bui, Quanling Deng
Claims MonthPE+SeasonBias priors improve Conv-Transformer Antarctic SIC forecasting; gains plausible vs DL baselines, untested against climatology.
MURANO: Design, Run, and Reproduce Mechanistic Interpretability Experiments as Composable Pipelines
Alireza Bayat Makou, Emirhan Böge, Phu Gia Hoang, Federico Tiblias et al.
Open-source pipeline framework for mech-interp experiments; reproductions largely match published findings, but usability is judged only via code-line counts.
HiRS-Agent: A Hierarchical Multi-Agent System for Reliable Long-Horizon Remote Sensing Task Solving
Boyang Mu, Zhiwei Wei, Mugen Peng, Wenjia Xu
Two-level RS agent with expert tuning + verification-guided HRL lifts Qwen3-4B Earth-Bench accuracy 15.7→43.9; strong ablations, yet no significance testing.
Learning Materials Properties from Scarce Labels and Unlabeled Crystals
Wentao Li, Yizhe Chen, Jiangjie Qiu, Yijun Li et al.
SemiMat benchmark + MatRank SSL give best aggregate NMAE (0.896) over 24 blocks; rigorous protocol, but lacks significance tests and pretraining baselines.
Commit-first LLM judging inherits the judge's own errors
Idil Gozel
Careful empirical audit and experiment showing commit-first judging moves the gaming anchor to the judge's own errors; convincing, though small in scale.
CoMPASS: Collaborative Molecular Property Prediction via Adaptive Small-Large Model Synergy
Wentao Li, Jiangjie Qiu, Yijun Li, Leyi Zhao et al.
Improves its own GAT anchor on 8 benchmarks via retrieval calibration plus a bounded LLM gate; credible ablations, but unpaired baselines temper the headline.
WildSEEK: Evaluating Language Models for Information-Seeking
Tanise Ceron, Joachim Baumann, Elisa Bassignana, Berat Cabuk et al.
Claims analytical queries elicit more LLM failures; WildSEEK dataset/framework is solid, but weak judge agreement on US bias (κ=.40) tempers findings.
Foundation models for electricity price forecasting and battery arbitrage: Can they replace market-specific forecasting models?
Arkadiusz Lipiecki, Rafał Weron
Rigorous 5-year, 3-market zero-shot FM test: only TabPFN beats EPF benchmarks statistically, but BESS arbitrage value hinges on risk appetite.