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.
Building Socio-Affective Artificial Intelligence for Interactive Multi-Agent Simulations
David Berga
AGIMUD: open MUD simulator adding Schwartz/Ostrom/Shapley reasoning and Ekman emotion to LLM agents; architecture solid but core validation fails (26.8% match).
A2M: Trace-Optimized Agent Hijacking in the MCP Ecosystem
Laizhen Li, Xuan Wang, Peicheng Zhao, Juanjuan Zhao et al.
Proposes A2M, a black-box attack that hijacks MCP agents (93.6% invocation, 74.4% ASR on GLM-4.6); convincing, but judge validation and variance lacking.
SWE-Serve: Benchmarking Agentic Engineering For Production Inference Serving
Jennifer Williams, Dave Farris, Jeff Farris, Jiantao Jiao
SWE-Serve benchmarks agents on SGLang production-inference tasks; rigorous pipeline and E2E gap, but matched-control CI includes zero, tempering causality.
PR-Smoother: Simulator-Preserving Non-Gaussian Smoothing for Data Assimilation
Yuta Tarumi
PR-Smoother keeps the prescribed simulator inside an amortized non-Gaussian smoother for DA; sound limiting-case theory, strong but synthetic-only experiments.
Flash-dLLM: IO-Aware KV Caching and Parallel Decoding for Fast, Memory-Efficient Diffusion LLMs
Quan Nguyen-Tri, Mukul Ranjan, Zhiqiang Shen
Training-free fused KV-cache kernel + self draft-and-verify for diffusion LLMs; 210.6 tok/s on GSM8K-512, but results shown only for LLaDA-1.5.
TwinCheck: Evidence-Grounded Negative-Twin Verification for Stateful Tool Agents
Jiaxuan Dai, Tianyi Huang
TwinCheck swaps an agent's tool call only when trace evidence and a twin comparison justify it; BFCL V4 +11.6pp, zero harms — but one benchmark, one model.
Ajar: Measuring Open Privilege in Agent Defenses
Reshabh K Sharma, Linxi Jiang, Shuo Chen, Zhiqiang Lin
Ajar adds a third axis—harm-weighted open privilege left by agent defenses—to AgentDojo; convincing and honest, but 73% of oracle labels are self-asserted.
Cross-Modal Contrastive Learning from Histopathology and CT for Automated Renal Cell Carcinoma Grading
Amit Das, Tanmay Shukla, Naofumi Tomita, Faraz Farhadi et al.
RCC-Align uses paired histopathology–CT contrastive training for CT-only ccRCC grading; the AUC 0.601 gain over baseline is significant yet near chance, so clinical value stays unproven.
CRISP: Scalable Importance-Stratified Coresets for Imbalanced Tabular Learning
Hardhik Mohanty, Indrayana Rustandi, Mohamadreza Sheibani
Score-stratified coreset for imbalanced GBDT training; Criteo evidence solid, but the proprietary headline, shifted settings and no tests temper it
On Preference Coverage Collapse from Hindsight Relabeling in Multi-Objective Reinforcement Learning
Baptiste Bonin, Caro Strickland, Audrey Durand
Careful, well-powered study: HER-style preference relabeling harms MORL via coverage collapse; her_mix fix convincing but only within MO-MuJoCo scope.
COMED: The Missing Middle Between Routing and Collaboration in Multi-LLM Inference
Norah Alballa, Wenxuan Zhang, Salma Kharrat, Fares Fourati et al.
COMED 是多 LLM 推理中的一种接受/验证/协作的锚点后控制器；其广泛的收益是可信的，但在稀疏设置下的获胜处于噪声范围内。
Classifying Interpretive Canons at the Sentence Level: A Benchmark from the German Federal Constitutional Court
Felix Ringe
Sound sentence-level benchmark for German legal canons with LLM baselines; strong reproducibility, but single-annotator labels and small tests limit findings.
ContraVis: Evidence-Grounded Visual Analytics for Contradiction Review in Legal Contracts
Luis Sante, Paula Lima, Mariana Rocha, Jorge Poco
ContraVis: typed paragraph graph conditions LLM and anchors visual inspection; suggestive gains over standalone LLM but no stats on 4 contracts and 3 lawyers.
Same evidence, different judgments: Evidence noncommutative in vision/speech-text conflicts
Zhuoyun Li, Boxuan Wang, Xiaowei Huang, Yi Dong
Same-evidence order swaps shift multimodal judgments toward image/speech; a well-controlled empirical claim, though "noncommutativity" slightly outruns two-element scope.
Experts Rise Where LLMs Disagree: Using Cross-Model Disagreement to Target Expert Effort in LLM Codebook Revision for Large-Scale Annotation
Zeyu He, Zhuqian Zhou, Kirk Vanacore, Rene F. Kizilcec et al.
Claims a few hours of expert rationale-labeling beats a 6-month codebook (64.9% vs 57.8% accuracy); strong single-project evidence, generality unproven.
A 3D Pose-Based Ensemble Framework for Cricket Shot Classification and Automated Biomechanical Analysis
Sourav Shome, M. D. Ashiquzzaman Rahad, Rameswar Debnath
Claims 97.68% 5-fold accuracy for a YOLO+MeTRAbs 3D-pose LRCN–TCN ensemble on four cricket shots; promising but undermined by inconsistent metric tables and no significance testing.
When Post-Processing Fairness Constraints Help and When They Harm: Evidence from Eight Cross-Domain Evaluations
Nithin Raghava Ramachandra Narla
Cross-domain study finds ThresholdOptimizer helps when baseline disparity is high, harms when near-fair; transparent but lacks uncertainty quantification.
Sharp Convergence of Wasserstein Gradient Flows for Spectrally Nonnegative Interaction Energies
Zhengjiang Lin, Philippe Rigollet
The paper proves global convergence at a universal o(t^-1) energy rate for nondiffusive Wasserstein gradient flows of spectrally nonnegative kernels and shows the rate is sharp; a clean, self-contained contribution.
Recognized but Not Produced: A Generation Benchmark for Culturally Specific Kinship Terms
Sahil Pardasani, Madhusudan Singh
Kinship-term generation benchmark (Hindi/Tamil/Korean, 5 LLMs) vs matched MC baseline: format gap is real; honest analysis, but scoring lacks human validation.
GeoRVQ: Decoder-aware geometry for residual-token prediction in physiological signals
Bo Cui, Yaowen Zhang
GeoRVQ: decoder-aware targets + coarse-to-fine heads cut decoded error 35% at near-flat token accuracy; consistent 3-run evidence, but purely descriptive.
Escaping Python Dependency Hell: A Hybrid Replay-and-Repair Pipeline for Python Dependency Resolution
Veronica Poweska, Ariana Oyanguren, Jessica Pourleyli, Sourena Khanzadeh et al.
PLLM+ lifts HG2.9K dependency repair to 51.9% from 40.4%, but 99.7% of fixes replay prior baseline solutions; honest yet thin on novelty and generalization.
LEGO: Synergizing Expert GraphRAG and Expert Chain-of-Thought for Legal Reasoning
Qingjing Chen, Junkai Zhang, Shaochun Wang, Jiahao Ding et al.
Expert-graph RAG + syllogistic CoT for Chinese civil law: strong same-backbone RAG gains, but claimed parity with larger models lacks statistical support.
CVaR anchor regression protects against rare shifts
Malte Londschien
Proposes CVaR anchor regression with exact shift-robustness guarantees for rare shifts; theory is clean and honest, though empirical validation is thin.
Untangling the Geometry and Speed for RF Sensing Spectrograms
Mert Torun, Darius Cuenca, Yasamin Mostofi
Physics-informed autoencoder disentangles speed and geometry from single-link WiFi Doppler; strong synthetic results, thin real evidence (4 trials).