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.
PIDS-Bench: Evaluating Prompt-Injection Detectors Under Over-Defense, Obfuscation, and Distribution Shift
Yusuf Khalid Shire, Sang-Chul Kim
Proposes a frozen multi-axis PI-detector benchmark showing F1 hides benign over-defense; claims carefully scoped, statistics sound, reproducibility strong.
Shared KV Caching for Replicated 27B Inference: Correctness Failures and Performance Boundaries
Frank Li
Claims 31.7→0.6s TTFT via shared KV caching; stream-ordering fault diagnosis is rigorous, but evidence spans one host, one model, and two block pairs.
TwinICL: Diagnosing Multimodal In-Context Learning through Paired Counterfactuals
Zihan Xue, Po-Yi Lu, Serhii Honcharenko, Zih-Ching Chen et al.
Paired text–image benchmark shows persistent multimodal ICL gaps across six models; strong controls, but toy shape tasks and partial run coverage limit scope.
SALUTE: Benchmarking and Adapting LLMs for the Defense Domain
Hyeongcheol Park, Sumin In, Suyeon Myeong, Hogun Park et al.
Solid multi-stage defense-domain LLM adaptation with careful data curation, but headline gains are measured chiefly on the authors' own new benchmark.
SpliTEE: Improving LLM Inference on Trusted Hardware with Differentially Private GPU Outsourcing
Shashie Dilhara Batan Arachchige, Robin Carpentier, Hassan Jameel Asghar, Dali Kaafar
Extends Slalom TEE-GPU split LLM inference with DP noise, not encryption; strong theory and attack eval, but privacy accounting skips offload composition.
Pick Your Poison: Learning to Select Poison Sets for Stronger LLM Backdoor Attacks
Aashiq Muhamed, Mona T. Diab, Virginia Smith, Andrew Ilyas et al.
Shows poison-set choice swings LLM backdoor ASR 3–80%; SAILS's learned set scorer is convincing, but its +30pp headline uses unequal budgets vs +21pp matched.
Validating Hybrid-State Cache Recovery for GLM-5.3-Flash with vLLM and LMCache
Frank Li
Case study validating a strict-prefix cache-recovery fix for vLLM+LMCache hybrid serving; bounded evidence convincing but one model only, no code release.
The average-farmer illusion in language-model simulations of agricultural decisions
Zhanliang Zhu, Ziwei Li, Yuchen Liu, Liujun Zhu et al.
LLM farmer agents match group means yet fail person-level fidelity and tails; a distribution-only generator beats all 24 configs — a rigorous cautionary study.
Horizon-specific Expert Fusion for Photovoltaic Power Forecasting
Xu Yuqing, Zhou Liguo, Sun Ze, Yu Lei et al.
PV ensemble of 8 experts edges LightGBM 4.11% on one station yet misses its own 0.2pp bar; rigorous and honest, but gains are small and site-specific.
MoARa: Module-Aware Rank Allocation and Structure-Preserving Decomposition for Low-Rank LLM Pre-training
Keunyoung Kim, Nojun Kwak
Claims 37% faster GaLore pretraining via module-aware ranks plus gradient decomposition; evidence broad and credible, though decomposition is host-dependent.
Data Attribution via Sketched Metadifferentiation
Yuxi Chen, Hamza Golubovic, Han Tong, Arian Maleki et al.
Casts budgeted attribution as influence-matrix sketching; MAGE/SPELL beat baselines on two modest benchmarks, but core assumptions and scale remain unverified.
Zero-SNR Analyticity of the Scalar MMSE Is Equivalent to Gaussianity
Yixing Zhang
Proves MMSE analytic at zero SNR iff input Gaussian (sub-Gaussian class) via MGF zeros and Borel analysis; consistent at spot-check, needs full refereeing.
Mirror, Mirror on the Wall: Prompt Echoing in Small Instruct Language Models
Inez Okulska, Bartosz Naskręcki, Jan Piotrowski, Tomasz Steifer
Claims small-model prompt echoing reflects partial training-data overlap but is mainly driven by induction-head copying; solid, though no statistical tests.
Personalizing Personal Health Interfaces: Co-Design with Generative AI
Karthik S. Bhat, Vidhi Shah, Vedika Agnihotri, Dong Whi Yoo et al.
Co-design study (n=14) shows generative AI helps users prototype health interfaces but injects model defaults; rich qualitative insight, low generalizability.
Ensemble Complexity in Photovoltaic Forecasting
Sun Ze, Zhou Liguo, Xu Yuqing, Yu Lei et al.
Transparent PV ensemble audit: fusion beats matched boosting only on OPSD after correction; weather gate adds nothing; oracle weather limits broader claims.
Structured Features Overfit Where Random Features Grok
Chon-Fai Kam, Miloud Bessafi, Frederic Cadet
Structured Fourier features remove the grokking delay XVS prove for random ones; masking shows active-mode count drives collapse; solid but one λ, one target.
Is INT8 Portable? A Cross-Platform Measurement Study of Quantized Inference on Embedded and Automotive Accelerators
Yuyeong Shin
Controlled 7-hardware-class study shows INT8 speedup, numerics, and deployability are not portable; convincing, though n=1 per class and one headline model.
Not All Prompts Are Equal: Exploration-Guided Prompt Scaffolding for Multimodal Reinforcement Post-Training
Yuanhao Yue, Qianli Ma, Chengyu Wang, Haoting Wang et al.
EPS prompt scoring plus teacher hint rewriting consistently beats GRPO in MLLM RL, but answer-aware scaffolds and a missing filter-only ablation temper claims.
Bridging the Gap in ECG-Based Emotion Recognition: A Unified Evaluation of Deep Learning Models
Timothy C Sweeney-Fanelli, Ajan Ahmed, Masudul Imtiaz
Unified benchmark of five deep models for ECG emotion recognition; accuracies near chance (51.8–57.0%) and model rankings lack statistical testing.
BusMA: A Bus Communication Substrate for Multi-Agent Systems
Yanwen Peng, Delvin Ce Zhang, Xi Wang, Nikolaos Aletras
BusMA lets agents address peers on a shared bus with four intents; wins on 13 benchmarks look broad, but with no significance tests small margins stay unproven.
Sensory Precision Inference for Multimodal Arbitration under Uncertainty
Tin Mišić, Takato Horii
Joint inference of latents and modality precisions via free-energy minimization shows synthetic gains, but lacks external baselines and statistical testing.
AdaVSkip: Adaptive Visual Token Skipping Across Layers For Efficient MLLMs Inference
Yuyao Sun, Tao Deng, Shuang Li, Deqing Wang
Claims adaptive per-module visual-token skipping halves MLLM FLOPs at near-zero loss; well-ablated, strong tables, but no significance tests or code.
DA-DLM: Explicitly Modeling Token Dependencies in Diffusion Language Models
Pengyu Ji, Zichen Zhang, Xiang Hu, Kewei Tu
Adapts DAT-style DAG dependency modeling to block diffusion LMs; consistent gains over BD3LM at small scale, but no head-to-head with competing methods.
Salesforce Koa: An Enterprise Language Model for Agentic Tool Use
Zixiang Chen, Sufeng Niu, Yingchi Liu, Wenting Zhao et al.
Spec-driven GRPO lifts an RL-post-trained base on agentic tool use (BFCL multi-turn 54.1→59.5), but gains over base are modest and lack significance testing.