Isolater - Feed

Ax Shubh Laddha, Lucas Changbencharoen, Win Kuptivej, Surya Shringla, Archana Vaidheeswaran, Yash Bhaskar 3/2/2026

HumanMCP: A Human-Like Query Dataset for Evaluating MCP Tool Retrieval Performance

HumanMCP dataset providing human-like queries for evaluating tool retrieval performance in Model Context Protocol (MCP) servers.

Ax Pavel Chernakov, Sasan Jafarnejad, Rapha\"el Frank 3/2/2026

An Agentic LLM Framework for Adverse Media Screening in AML Compliance

Agentic LLM system with RAG for automating adverse media screening in AML compliance, reducing false positives from keyword-based approaches.

Ax Lun Zhan, Feng Xiong, Huanyong Liu, Feng Zhang, Yuhui Yin 3/2/2026

MMKG-RDS: Reasoning Data Synthesis via Deep Mining of Multimodal Knowledge Graphs

MMKG-RDS is a framework for synthesizing training data using multimodal knowledge graphs to improve domain model reasoning capabilities.

Ax Yihan (Logon), Wen, Xin Chen 3/2/2026

PseudoAct: Leveraging Pseudocode Synthesis for Flexible Planning and Action Control in Large Language Model Agents

PseudoAct introduces pseudocode-based planning for LLM agents to reduce token consumption and improve stability on long-horizon multi-tool tasks.

Ax Siyuan Ma, Bo Gao, Xiaojun Jia, Simeng Qin, Tianlin Li, Ke Ma, Xiaoshuang Jia, Wenqi Ren, Yang Liu 3/2/2026

ODAR: Principled Adaptive Routing for LLM Reasoning via Active Inference

ODAR-Expert presents an adaptive routing framework for optimizing accuracy-efficiency tradeoffs in LLM reasoning via active inference instead of uniform sampling.

Ax Yawen Wang, Wenjie Wu, Junjie Wang, Qing Wang 3/2/2026

From Flat Logs to Causal Graphs: Hierarchical Failure Attribution for LLM-based Multi-Agent Systems

Method for hierarchical failure attribution in multi-agent LLM systems using causal graphs to improve observability and debugging of complex agent interactions.

Ax Jiangyuan Wang, Kejun Xiao, Huaipeng Zhao, Tao Luo, Xiaoyi Zeng 3/2/2026

ProductResearch: Training E-Commerce Deep Research Agents via Multi-Agent Synthetic Trajectory Distillation

ProductResearch proposes a multi-agent framework using trajectory distillation to train LLM-based agents for complex e-commerce product research tasks.

Ax Sheng Cao, Zhao Chang, Chang Li, Hannan Li, Liyao Fu, Ji Tang 3/2/2026

The Auton Agentic AI Framework

Auton framework addresses architectural mismatch between stochastic LLM outputs and deterministic backend systems for agentic AI deployment.

Ax Longyin Zhang, Shuo Sun, Yingxu He, Won Cheng Yi Lewis, Muhammad Huzaifah Bin Md Shahrin, Hardik Bhupendra Sailor, Heng Meng Jeremy Wong, Tarun Kumar Vangani, Yi Ma, Qiongqiong Wang, Minh Duc Pham, Ridong Jiang, Jingtao Li, Jingyi Liao, Zhuohan Liu, Yanfeng Lu, Manas Gupta, Ai Ti Aw 3/2/2026

Unlocking Cognitive Capabilities and Analyzing the Perception-Logic Trade-off

MERaLiON2-Omni: 10B multilingual MLLM for Southeast Asia addressing perception-logic tradeoffs in omni-perception tasks.

Ax Zhipeng Xu, Zilong Wang, Xinyang Jiang, Dongsheng Li, De Cheng, Nannan Wang 3/2/2026

Reasoning-Driven Multimodal LLM for Domain Generalization

Domain generalization method leveraging reasoning chains in MLLMs to improve robustness under domain shift.

Ax Yiyang Fang, Wenke Huang, Pei Fu, Yihao Yang, Kehua Su, Zhenbo Luo, Jian Luan, Mang Ye 3/2/2026

EMO-R3: Reflective Reinforcement Learning for Emotional Reasoning in Multimodal Large Language Models

EMO-R3 applies reflective reinforcement learning to improve emotional reasoning capabilities in multimodal LLMs.

Ax Chao Wang, Han Lin, Huaze Tang, Huijing Lin, Wenbo Ding 3/2/2026

RUMAD: Reinforcement-Unifying Multi-Agent Debate

RUMAD: reinforcement learning approach to multi-agent debate that adapts topology to task complexity while maintaining debate neutrality.

Ax Ning Gao, Xiuhui Zhang, Xingyu Jiang, Mukang You, Mohan Zhang, Yue Deng 3/2/2026

RF-Agent: Automated Reward Function Design via Language Agent Tree Search

RF-Agent uses LLM-based tree search to automatically design reward functions for control tasks with improved historical feedback utilization.

Ax Fan Zhang, Baoru Huang, Xin Zhang 3/2/2026

Pessimistic Auxiliary Policy for Offline Reinforcement Learning

Pessimistic auxiliary policy approach for offline reinforcement learning to mitigate overestimation from out-of-distribution actions.

Ax Vanya Priscillia Bendatu, Yao Lu 3/2/2026

Portfolio Reinforcement Learning with Scenario-Context Rollout

Scenario-context rollout reinforcement learning for portfolio rebalancing under market regime shifts and distribution changes.

Ax Reva Schwartz, Carina Westling, Morgan Briggs, Marzieh Fadaee, Isar Nejadgholi, Matthew Holmes, Fariza Rashid, Maya Carlyle, Afaf Ta\"ik, Kyra Wilson, Peter Douglas, Theodora Skeadas, Gabriella Waters, Rumman Chowdhury, Thiago Lacerda 3/2/2026

CIRCLE: A Framework for Evaluating AI from a Real-World Lens

CIRCLE: six-stage framework for evaluating AI systems under real-world conditions and user variability beyond model-centric metrics.

Ax Xiang Li, Jiabao Gao, Sipei Lin, Xuan Zhou, Chi Zhang, Bo Cheng, Jiale Han, Benyou Wang 3/2/2026

Human or Machine? A Preliminary Turing Test for Speech-to-Speech Interaction

Turing test evaluation of 9 state-of-the-art speech-to-speech systems with human judgments on conversational naturalness.

Ax Yue Xie, Zizhen Xu, William Beazley, Fumiya Iida 3/2/2026

Bi-level RL-Heuristic Optimization for Real-world Winter Road Maintenance

Bi-level RL-heuristic optimization for winter road maintenance routing on UK strategic and local road networks.

Ax Richard Csaky 3/2/2026

Artificial Agency Program: Curiosity, compression, and communication in agents

Position paper on Artificial Agency Program proposing resource-bounded, curiosity-driven agents as embedded systems within human-tool extended systems.

Ax Yanwei Ren, Haotian Zhang, Likang Xiao, Xikai Zhang, Jiaxing Huang, Jiayan Qiu, Baosheng Yu, Quan Chen, Liu Liu 3/2/2026

Recycling Failures: Salvaging Exploration in RLVR via Fine-Grained Off-Policy Guidance

Fine-grained off-policy guidance improves exploration in reinforcement learning from verifiable rewards for complex reasoning in large language models.

Ax Antoine Peyronnet, Fabian Gloeckle, Amaury Hayat 3/2/2026

LemmaBench: A Live, Research-Level Benchmark to Evaluate LLM Capabilities in Mathematics

LemmaBench: live, updatable benchmark evaluating LLMs on research-level mathematics by extracting lemmas from arXiv papers.

Ax Shishun Zhang, Juzhan Xu, Yidan Fan, Chenyang Zhu, Ruizhen Hu, Yongjun Wang, Kai Xu 3/2/2026

Learning Flexible Job Shop Scheduling under Limited Buffers and Material Kitting Constraints

Deep learning approach to flexible job shop scheduling with buffer and material constraints for production optimization.

Ax Gregory Kang Ruey Lau, Hieu Dao, Nicole Kan Hui Lin, Bryan Kian Hsiang Low 3/2/2026

Uncertainty Quantification for Multimodal Large Language Models with Incoherence-adjusted Semantic Volume

Method for uncertainty quantification in multimodal LLMs using semantic volume metrics to identify unreliable outputs.

Ax Borja Requena Pozo, Austin Letson, Krystian Nowakowski, Izan Beltran Ferreiro, Leopoldo Sarra 3/2/2026

A Minimal Agent for Automated Theorem Proving

Minimal agentic baseline for automated theorem proving that enables systematic comparison across AI-based prover architectures with iterative refinement and library search.

Ax Fan Shu, Yite Wang, Ruofan Wu, Boyi Liu, Zhewei Yao, Yuxiong He, Feng Yan 3/2/2026

DARE-bench: Evaluating Modeling and Instruction Fidelity of LLMs in Data Science

DARE-bench introduces a benchmark for evaluating LLMs on multi-step data science tasks with focus on instruction adherence and process fidelity.

Ax Cheng Qian, Yulun Zhang, Varun Bhatt, Matthew Christopher Fontaine, Stefanos Nikolaidis, Jiaoyang Li 3/2/2026

QD-MAPPER: A Quality Diversity Framework to Automatically Evaluate Multi-Agent Path Finding Algorithms in Diverse Maps

QD-MAPPER uses Quality Diversity and Neural Cellular Automata to automatically generate diverse maps for evaluating multi-agent path finding algorithms.

Ax H. C. W. Price, H. AlMuhanna, P. M. Bassani, M. Ho, T. S. Evans 3/2/2026

Let There Be Claws: An Early Social Network Analysis of AI Agents on Moltbook

Social network analysis of Moltbook, an AI-native platform, reveals rapid stratification and hierarchical structures emerge within 12 days across 15K+ agent accounts.

Ax Shreyas Subramanian, Adewale Akinfaderin, Yanyan Zhang, Ishan Singh, Mani Khanuja, Sandeep Singh, Maira Ladeira Tanke 3/2/2026

Keyword search is all you need: Achieving RAG-Level Performance without vector databases using agentic tool use

Demonstrates agentic tool-augmented LLMs achieve RAG-level performance using keyword search without vector databases.

Ax Xuanming Cui, Hong-You Chen, Hao Yu, Hao Yuan, Zihao Wang, Shlok Kumar Mishra, Hanchao Yu, Yonghuan Yang, Jun Xiao, Ser-Nam Lim, Jianpeng Cheng, Qi Guo, Xiangjun Fan 3/2/2026

Reason to Contrast: A Cascaded Multimodal Retrieval Framework

TTE-v2 hybrid multimodal retrieval framework extending reasoning-driven bi-encoder architectures with improved performance.

Ax Kaifeng Wu, Junyan Wu, Qiang Liu, Jiarui Zhang, Wen Xu 3/2/2026

Toward General Semantic Chunking: A Discriminative Framework for Ultra-Long Documents

Discriminative framework for semantic chunking of ultra-long documents improving topic segmentation and retrieval.

Ax Rakshita Goel, S Pranav Kumar, Anmol Agrawal, Divyan Poddar, Pratik Narang, Dhruv Kumar 3/2/2026

Domain-Partitioned Hybrid RAG for Legal Reasoning: Toward Modular and Explainable Legal AI for India

Domain-partitioned hybrid RAG for legal document reasoning across Indian statutes, codes and precedents.

Ax Qizhi Wang 3/2/2026

Democratizing GraphRAG: Linear, CPU-Only Graph Retrieval for Multi-Hop QA

SPRIG democratizes GraphRAG with CPU-only linear-time pipeline using NER co-occurrence graphs and PPR for multi-hop QA.

Ax Weixi Lin 3/2/2026

Higress-RAG: A Holistic Optimization Framework for Enterprise Retrieval-Augmented Generation via Dual Hybrid Retrieval, Adaptive Routing, and CRAG

Higress-RAG optimizes enterprise RAG with dual hybrid retrieval, adaptive routing and CRAG to reduce hallucination.

Ax Yueran Hou, Peilei Jia, Zihan Sun, Qihang Lu, Wenbing Yang, Yingming Gao, Ya Li, Jun Gao 3/2/2026

Hello-Chat: Towards Realistic Social Audio Interactions

Hello-Chat end-to-end audio language model for realistic social interactions with emotional resonance.

Ax Jiazheng Quan, Xiaodong Li, Bin Wang, Guo An, Like Liu, Degen Huang, Lin Liu, Chengbin Hou 3/2/2026

Learning to Generate Secure Code via Token-Level Rewards

Vul2Safe framework uses token-level RL rewards and LLM self-reflection for secure code generation from LLMs.

Ax Varun Gopal, Rishabh Jain, Aradhya Mathur, Nikitha SR, Sohan Patnaik, Sudhir Yarram, Mayur Hemani, Balaji Krishnamurthy, Mausoom Sarkar 3/2/2026

DesignSense: A Human Preference Dataset and Reward Modeling Framework for Graphic Layout Generation

DesignSense dataset and reward model framework for graphic layout generation using human preference learning.

Ax Alejandro Rodriguez Dominguez 3/2/2026

Human Supervision as an Information Bottleneck: A Unified Theory of Error Floors in Human-Guided Learning

Develops unified theory showing human supervision as information bottleneck explaining error floors in LLM training from human feedback.

Ax Gaurav Kamath, Sreenath Madathil, Sebastian Schuster, Marie-Catherine de Marneffe, Siva Reddy 3/2/2026

Humans and LLMs Diverge on Probabilistic Inferences

Study showing divergence between human and LLM behavior on probabilistic inference tasks requiring non-deterministic reasoning.

Ax Aishwarya Sarkar, Sayan Ghosh, Nathan Tallent, Aman Chadha, Tanya Roosta, Ali Jannesari 3/2/2026

Rudder: Steering Prefetching in Distributed GNN Training using LLM Agents

Rudder: LLM agent-based prefetching steering for distributed GNN training to optimize irregular communication patterns.

Ax Asiri Wijesinghe, Sevvandi Kandanaarachchi, Daniel M. Steinberg, Cheng Soon Ong 3/2/2026

Flowette: Flow Matching with Graphette Priors for Graph Generation

Flowette: flow matching generative model for graphs with recurring motifs using graph neural network transformers.

Ax Yun Wang, Xuansheng Wu, Jingyuan Huang, Lei Liu, Xiaoming Zhai, Ninghao Liu 3/2/2026

BRIDGE the Gap: Mitigating Bias Amplification in Automated Scoring of English Language Learners via Inter-group Data Augmentation

BRIDGE: data augmentation method mitigating bias amplification in automated scoring systems for English Language Learners.

Ax Xiang Ao 3/2/2026