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.
Listening and Mirroring: The Effects of Verbal Attunement and Behavioral Mimicry on Social and Empathic Perceptions of Embodied AI Agents in VR
Nathalia Gomez, Haig Shamlian, Omar Khan, Tiffany D. Do
Solid 2×2 VR study (n=20): verbal attunement drives perceived empathy, mimicry only marginal. Honest but underpowered; gender patterns exploratory.
On the Sample Complexity of Active Learning with Membership Queries
Ganghua Wang, Shaddin Dughmi
Argues membership queries make 2D halfspaces exponentially learnable where pools fail; rigorous but main theorem known and halving conjecture unresolved.
Combining LLMs and Genetic Search for ARC-AGI-2
Val Dyachenko
Claims GA evolves LLM-seeded programs to lift ARC-AGI-2 solves from 2/60 to 6/60; promising but tiny scale, non-random subset, incomplete control.
DRSR: Learning Set-Level Deletion Risk for Efficient Long-Horizon Agents
Mingxuan Wang, Bo Wang, Fei Luo, Guorun Yao et al.
DRSR learns set-level deletion risk via counterfactual supervision, reporting +0.103 reward at -20.82% tokens; results are suggestive but lack error bars.
Hunyuan-A13B Technical Report
Tencent Hunyuan Team, Ao Liu, Botong Zhou, Can Xu et al.
Tencent's 80B/13B-active MoE claims near-frontier scores; broad tables but no variance, hardware specs or ablations—statistically unverified but promising.
Full-Covariance Smoothing of Bayesian Neural Networks for Online Adaptation
Oren Wright, Haoming Jing, Qiaoan Shen, Koichiro Niinuma et al.
Claims full-covariance smoothing for BNNs; sound theory, small-scale experiments, distributional-target path untested. Moderate confidence.
Can One Adapted Model Do It All? Fine-Tuning Strategy Selection for Customer Support LLMs
Md Tahmid Rahman Laskar, Xue-Yong Fu, Shashi Bhushan TN
Controlled comparison of fine-tuning strategies across 13 support LLMs finds MT-full best; useful, but lacks significance tests and half the data is private.
UniDataAgent: An Ontology-Grounded Agent for Enterprise Question-to-Report Automation
Yutai Duan, Yahui Zhao, Zhangti Li, Yu Ma et al.
A deployed ontology-grounded enterprise QA system with strong reported gains (95.0% vs 72.5% RAG), but evaluation is small, single-site, and baseline-thin.
Teach-to-Crash: A Closed-Loop Student-Teacher LLM Framework for Collision-Inducing Test Scenario Generation
Zaid Ghazal, Khouloud Gaaloul, Bruce Maxim
Claims dual-LLM closed-loop generation tops CHR/TTC/diversity in CARLA; well-hedged, statistically careful, but the dual-LLM design itself is never isolated.
The Risk-Sensitive Schrödinger Bridge: Is Not a KL Projection
Hamidreza Behjoo
Shows the risk-sensitive Schrödinger bridge has no fixed-reference KL dual, via an exact θ/2 heat-equation defect; formal yet careful, existence left open.
CAVEAT: Towards Robust Computer-Use Agents in Incentive-Misaligned Environments
Yuxuan Li, Will Epperson, Wesley Deng, Zezhou Huang
Benchmark: marketplace steering drops CUA optimal purchases 78.6%→17.3%; sound matched-control design, but no stated code release and tiny Hard split.
TimeEvo: Failure-Driven Self-Evolution of a Time Series Agent
Jie Yang, Yan Zheng, Jiarui Sun, Xiran Fan et al.
Claims failure-driven tool synthesis with a paired admission gate wins all 30 task-backbone cells; protocol is rigorous but no self-evolving baselines are run.
EnSIMem: Entity-Structured Indexing for Long-Term Agent Memory
Xuanyu Meng, Xing Fan, Xinyi Fan, Chenlei Guo et al.
Claims SOTA agent memory via entity-property indexing; results plausible, well-ablated, but LLM-judge scoring and transcribed baselines weaken the SOTA claim.
Graph Learning with Spectral Connectivity Priors for Scarce Data
Mingxiao Liu, Bahar Oveisgharan, Bingyan Zou, Gene Cheung et al.
SCoGL adds Laplacian-spectral connectivity priors to GLASSO for scarce-data graph learning; clean convex theory but synthetic-only evidence, no error bars.
StateComp: Learning When to Compress History in Long Horizon Agents
Mingxuan Wang, Hongyue Chen, Yinglong Guo, Fei Luo et al.
StateComp cuts agent tokens 52.27% at near-flat reward via state-based compression timing; promising, but single-run evals lack statistical rigor.
Memory Control Signals Emerge Before Action in Long Horizon Agents
Mingxuan Wang, Guorun Yao, Fei Luo, Yinglong Guo et al.
Claims hidden states encode memory needs pre-action (AUROC .831/.765); PaMER cuts tokens ~73% at near-par scores; LLM-made labels and thin stats temper it.
Multitask Regression with Pairwise Fusion
Xiaodong Li, Zhentao Li
Proves minimax-optimal rates and exact pooling of shared tasks for pairwise fusion; solid theory, but oracle tuning and missing code/closest baselines temper impact
SR-Fraud: An Outcome-Supervised Reflective LLM Agent Framework for Non-Stationary Payment Fraud Detection
Xuwei Tan, Yao Ma, Xueru Zhang
Claims verified reflection stabilizes LLM fraud agents; gains look credible on proprietary data, but no code/data release blocks independent checks.
PotARCin: Multi-Dimensional Evaluation of Skill Acquisition in Abstract Reasoning Tasks
Claas Beger, Ryan Yi, Melanie Mitchell
PotARCin extends ARC to five rule-use dimensions, exposing 25-52pp drops and self-contradiction; a rigorous benchmark study with acknowledged confounds.
Ruby-ASR: Evidence-Preserving Supervision for Joint Orthographic and Lexical-Reading Recognition
Hao Shi, Yun Liu, Xuehao Yang, Jun Liu et al.
Proposes ruby span-bound targets for Japanese ASR; five-benchmark gains and released code, but reading-label provenance and speech causality lack verification.
NGN: Learning Neural Network Size as a Differentiable Count
Lixing Li
NGN learns model size as a differentiable boundary; broad, honest experiments, but size choice shifts to tuning λ and PEFT results are single-seed.
Sparse-Observation Atmospheric Thermal Forecasting with Physics-Informed Neural Networks for Climate-Aware Digital Twins
Tannaz Goodarzvand Chegini, Elyas Shivanian, Behzad Karimi, Faraz Dadgostari
PINN thermodynamic constraint improves 1-3h ERA5 thermal hindcasts vs matched NN baselines (14.6-24.4% at +3h); honest terrain-failure report, but few origins.
KITE: KV-Invariant Transformer Expansion for Efficient Agentic LLM Scaling
Zhiheng Hu, Yixun Wei, Jian Zhou, Yizhuang Zhou et al.
KITE/SST 67B upcycling reportedly beats from-scratch 47B/63B at equal FLOPs with cheaper prefill; solid controls, single-run, analytic-cost, no ablations.
Live Assistant: Learning Whether, When, and Whom to Assist in Real-World Live Social Streams
Shujian Gao, Jiamei Yan, Yuchen Yang, Penghao Zhou et al.
Frames livestream assistance as OBS/MEM/ANS decisions with a 320-hour corpus and two-stage training; solid gains, but LLM-generated labels and no CIs temper claims.