Dr. Claw is an open‑source AI scientist workspace that integrates existing command‑line coding agents into a single, auditable, human‑in‑the‑loop workflow. It uses persistent state objects, a reusable skill library, and multi‑executor coordination to link human decisions with AI execution, creating a traceable and recoverable loop for planning, execution, and writing. The authors demonstrate the system with an interactive scenario and a failure‑recovery walkthrough, and show that, when the underlying executor is held constant, Dr. Claw achieves higher research completeness while preserving an auditable process trail.
By Dingjie Song, Hanrong Zhang, Dawei Liu, Yixin Liu, Zongxia Li, Zhengqing Yuan, Siqi Zhang, Henry Peng Zou, Zhiling Yan, Yuxuan Zhang, Yanfang Ye, Philip S. Yu, Lichao Sun
SpecMind is a Multi-Agent Retrieval-Augmented Generation system designed to support spectrum intelligence by reasoning over heterogeneous data sources such as policy proceedings, legal regulations, and license databases. The system coordinates specialized sub-agents that retrieve and synthesize knowledge, enabling autonomous agents to handle diverse query types. SpecBench, a Q&A dataset based on real-world license records and policy proceedings, was created to evaluate RAG systems in the spectrum domain, and experimental results show SpecMind outperforms traditional RAG systems with over an 80% win rate against strong baselines.
By Songwei Dong, Bingyan Lu, Makayla Kienlen, J. Nicholas Laneman, Cong Shen
mimeo is an open‑source tool that compiles a public expert’s work into a file an agent can load, verifying each quotation against the source text. In experiments, mimeo enabled agents to answer all 20 obscure, quotation‑heavy questions and avoided misstatements that occurred with personas generated from model memory. While it improved knowledge access and reduced misidentification, it did not demonstrate clear transfer of expert judgment or outperform simple keyword search in all metrics.
By Timothy Kassis
The paper presents a camera‑pose‑free stereo endoscopic method for recovering tissue deformation by modeling geometry as a 3D point‑derivative map and deformation as a 3D displacement‑local deformation map. It optimizes inter‑frame deformation in a camera‑centric setting, eliminating the need for camera pose estimation, and introduces a canonical map for online geometry and deformation optimization. Experiments on in‑vivo and ex‑vivo laparoscopic data show accurate 3D reconstruction (≈0.37–0.39 mm surface distance) even under occlusion, and the method can estimate surface strain distributions during manipulation.
By Jiahe Chen, Naoki Tomii, Ichiro Sakuma, Etsuko Kobayashi
The paper introduces “Revise”, a runtime system that performs validity-guided, fine-grained recovery for online revisions in structured agent workflows. When a revision arrives, Revise intersects the change with recorded data and control dependencies, propagates the impact through the partially executed DAG, stops invalid work, preserves unaffected progress, and recomputes only the affected region. Experiments on real coding‑agent traces and LangGraph/LLMCompiler applications show that Revise matches a latest‑version oracle, reduces model calls by up to 56%, and improves service‑level objective goodput under load.
By Ruoling Qi, Xuaner Wu, Penghang Liu, Jian Chen, Yirui Liu
DramaChain Bench is the first benchmark that evaluates every stage of short‑drama production—from script to finished video—using a unified dimension system called DramaChain Dimensions. It incorporates three in‑house systems: a calibrated agent that mirrors commercial workflows, a labeling system with 5,785 items scored by professional annotators, and an agentic judge that automatically scores leaf dimensions with high correlation to human judgments. The benchmark demonstrates that defects in early stages cascade through the pipeline, showing that final episode quality depends on more than just video generation.
By Haoyuan Shi (Hunyuan, Tencent), Mingtao Chen (Hunyuan, Tencent), Shuo Jiang (Hunyuan, Tencent), Ziyan Chen (Hunyuan, Tencent, Beijing Film Academy), Xuyi Sheng (Peking University), Yiming Liu (Hunyuan, Tencent), Ying Zhang (Hunyuan, Tencent), Miao Wang (Hunyuan, Tencent, Shenzhen University), Jianxiang Lu (Hunyuan, Tencent), Fanyang Lu (Hunyuan, Tencent), Songyuanyi Lu (Hunyuan, Tencent), Xiele Wu (Hunyuan, Tencent), Zhichao Hu (Hunyuan, Tencent), Yuhong Liu (Hunyuan, Tencent), Richeng Xuan (Hunyuan, Tencent)
The paper introduces Agentic Empirical Asset Pricing (AEAP), a framework where autonomous LLM agents conduct the entire scientific discovery process for asset pricing. It outlines AEAP’s core components, critiques current evaluation methods that only test outputs, and proposes a new reference architecture with rigorous standards for factor discovery and out‑of‑sample backtesting. Using this framework, the authors evaluate SEADS against five baselines on US equity panels, finding no single metric consistently ranks the systems and highlighting the need for multi‑axis evaluation and rolling re‑execution to assess reliability of the discovery process.
By Yingjian Pan, Xiaowei Ding, Kay Giesecke
ContextPipe is a database-inspired framework for assembling context in long-horizon large language model agents. It treats context assembly like relational query execution, using a five-phase pipeline—Plan, Bind, Optimize, Execute, Feedback—backed by a structured catalog, deterministic cache-aware optimizer, and EXPLAIN ANALYZE tracing. In a preliminary evaluation on the SWE-bench Pro Qutebrowser subset, ContextPipe reduced token volume by 31%, LLM calls by 23%, and response time by 9% compared to an append-only policy, though it lowered KV cache-hit ratio.
By Peng Xu, Zuyu Zhang, Yuze Sun, Feng Tian, Long Wang, Chen Zhang
The paper introduces VGEBench, a new benchmark for evaluating Vision‑Language Models (VLMs) on generalizable, visually grounded exploration of household devices. Unlike existing datasets that rely on static images or annotated trajectories, VGEBench employs a logic‑driven state machine to simulate multi‑turn interaction loops, requiring agents to actively perceive, act, and refine their actions to achieve goals. Experiments show that current VLMs struggle to translate semantic knowledge into physical execution and to maintain long‑horizon state tracking.
By Linhao Zheng, Zeming Liu, Wangke Chen, Li Zeng, Wanxiang Che, Heyan Huang, Yuhang Guo
The paper introduces a heterogeneous federated learning approach using the FractalNet architecture tailored for satellite mega‑constellations. It formalizes contact‑window‑constrained, depth‑heterogeneous optimization and proposes a distributed path scheduler that assigns model depth based on satellite SWAP‑C constraints, predicted contacts, and training statistics. The framework includes periodic update pooling and a three‑tier agentic control plane, and is validated through a wildfire detection case study across LEO, MEO, and GEO/HEO shells, demonstrating improvements in convergence, communication efficiency, energy adaptation, and robustness.
By Sai Puppala, Koushik Sinha
The paper introduces Reinforcement Learning Enhanced LLM Agents (RLEA), a multi‑agent framework that automates the modeling of complex Vehicle Routing Problems (VRPs). RLEA employs a lightweight neural Planner trained with Soft Q‑learning to coordinate LLM‑based agents, and incorporates an evolutionary memory module and retrieval‑augmented generation to leverage experience and external solver knowledge. Experiments on 48 VRP variants show that RLEA outperforms the prior state‑of‑the‑art method, achieving a 16.67% higher success rate and significantly reducing runtime errors.
By Yi Chen, Zikang Yu, Jiahai Wang, Jinbiao Chen, Jianpeng Zhou, Zizhen Zhang
The paper introduces a covariance‑corrected Mahalanobis distance method for detecting out‑of‑domain (OOD) intents in conversational agents. It builds on prior work showing Mahalanobis distance’s effectiveness but addresses its shortcomings in few‑shot scenarios, where limited examples of new intents are available. The authors analyze why the original approach underperforms and propose a refined distance metric to improve OOD detection in practical settings.
By Jayasimha Talur, Oleg Smirnov, Paul Missault
The paper introduces “FigTree”, a multi-agent system that automatically converts a scientific paper into a structured vector figure by recursively constructing SVG programs. It decomposes figures into hierarchical regions, generates each region as a short SVG program, and assembles them, using a render‑critic refinement loop to trace and repair visual defects. Evaluations show that “FigTree” produces high‑quality figures and allows more effective editing than raster‑based methods.
By Yepeng Liu, Dasen Dai, Chengzhi Liu, Yiren Song, Hai Ci, Yu Zhang, Qi Zhang, Mike Zheng Shou, Xin Eric Wang, Yuheng Bu
WorldBench is a new multilingual benchmark that tests large language model agents on culturally grounded everyday workflows, offering 1,600 tasks in seven languages and eight cultures. The benchmark evaluates agents through structured sandbox actions and introduces Constrained Task Success (CTS), a metric that assesses task completion, minimal modification, and other complementary aspects via deterministic and LLM-as-a-Judge evaluations. Experiments show that even leading models achieve only 49.2% CTS, revealing significant gaps in correctness and state preservation across languages and cultures.
By Leonardo Ranaldi, Sherrie Shen, Jushi Kai, Alexandra Birch
Analog-DB is an open‑source, versioned database that stores analog integrated circuit designs in a process‑neutral topology, reusable testbenches, and machine‑readable datasheets using a domain‑specific language. The database’s parameterization scheme and schema‑governed contracts enable AI design agents to discover, compose, and retarget circuits across multiple process kits. In a case study, a coding agent used the released artifacts to size op‑amp cores for a chopper instrumentation amplifier, uncovering defects that a sizing‑only baseline missed.
By Danial Noori Zadeh, Mohamed B. Elamien
The paper introduces a dual‑process architecture for nonlinear motion planning that blends fast, learning‑based intuition (System‑1) with slow, robust symbolic reasoning (System‑2). A metacognitive controller decides when to use each component, aiming to balance speed, precision, and adaptability. Experiments on diverse benchmark environments show consistent improvements in planning efficiency, accuracy, and generalization, highlighting the benefits of integrating learning with structured reasoning.
By Jiayi Yan, Francesco Fabiano, Alessandro Abate
LEAP (Likelihood Elicitation and Aggregation for Probabilistic forecasting) is a new approach that reorganizes how evidence is used in LLM-based forecasting systems. Instead of a monolithic prediction that aggregates all evidence at once, LEAP examines each evidence item separately, elicits likelihood parameters, and combines them with an explicit prior to produce a posterior distribution. The method supports continuous, single-choice, and multi-choice forecasts and has been shown to improve prediction and calibration metrics across models on a benchmark covering forecasting, information-seeking, and browsing tasks.
By Yufei Chen, Yiran Zhao, Xiaogang Xu, Qipeng Xie, Jiafei Wu, Zhe Liu
The paper investigates how marketplace guardrails affect welfare in language‑model agent simulations of hotel transactions. It finds that initial reports of large welfare gains disappear when controlling for offer schemas and buyer choice, and that guardrails mainly redistribute rather than increase welfare unless sellers are explicitly forced to produce inefficient bundles. The authors propose a construct‑validity framework to flag invalid or inconclusive policy claims before they are reported.
By Peiying Zhu, Sidi Chang
The paper introduces a live trace model that incrementally folds an append‑only event ledger into typed run state, producing per‑consumer views for both human observers and the agent itself. Evaluations show that for observers, the compiled view reduces input tokens by 14–15× and cost by 5–7× while improving accuracy from 0.48 to 0.85–0.87. For agents, maintaining running statistics in per‑step state enables success on 120‑link sequential tasks where full‑context prompting fails, and a prompt‑level scratchpad matches the fold’s accuracy at lower cost.
By Egor Pakhomov, Erik Nijkamp
The paper introduces SAGE, a framework that selectively queries a Vision‑Language Model (VLM) teacher only when the learner is uncertain, using the teacher’s suggestions to guide training and distill them into a lightweight reinforcement learning policy. SAGE weights teacher actions by environment‑derived advantages, allowing the policy to improve beyond the imperfect VLM. Experiments on sparse‑reward visual reasoning and navigation tasks show that the learned policies can act without VLM guidance at evaluation, reduce VLM usage during training, and sometimes outperform the teacher itself.
By Matteo Merler, Giovanni Bonetta, Davide Zago, Rossella Cancelliere, Bernardo Magnini