HN epaga 5/18/2026

DeerFlow 2.0 – Deep Exploration and Efficient Research Flow

Open-source agent orchestration framework v2.0 combining sub-agents, memory, sandboxes, and extensible skills for complex tasks.

HN UtkarshPatel13 5/18/2026

RAG vs. Fine-Tuning – The Question Every AI Builder Gets Wrong

Comparison of RAG, fine-tuning, and emerging approaches for integrating proprietary data into LLM applications in 2026.

HN meysamazad 5/18/2026

LLM Tracing with MLflow AI Gateway

MLflow AI Gateway tutorial on tracing agentic app interactions, token consumption, and tool usage for LLM debugging.

HN praneetbrar 5/18/2026

Ask HN: What AI coding workflows have stuck for you?

Discussion thread asking which AI-assisted coding workflows remain useful beyond initial novelty.

HN vickychijwani 5/18/2026

Apple's treatment of AI coding apps could be shifting with Replit update

Replit released iPhone app update with AI agent support after resolving Apple App Store disputes on AI code generation.

HN logan_ng 5/18/2026

Show HN: Thuki – local Al overlay for macOS (double-tap Control, no API key)

Thuki is a local macOS overlay powered by Ollama enabling AI queries without API keys or cloud calls via double-tap Control.

HN edwinhao 5/18/2026

MouseClaw – a pixel-art desktop AI pet for macOS (hold to speak)

MouseClaw is a macOS desktop AI pet that takes screenshots and processes voice commands with local CLI tools.

HN cachebag 5/18/2026

Neovim sweetspot between trad coding and AI coding

Neovim plugin (jumpy) for controlled AI-assisted code editing, maintaining developer oversight over LLM modifications.

HN eigenBasis 5/18/2026

How to Work and Compound with AI

Workflow framework for iterative work with AI, focusing on context compounding and error reduction across sessions.

HN inishchith 5/18/2026

AnyFrame – Sandboxes for Your AI Agents

AnyFrame Python SDK for sandboxing AI agents with Claude Code, featuring MCP server connectors and per-agent skill management.

Ax Ming Yang, Zhiwei Zhang, Jiahang Li, Haoseng Liu, Yuzheng Cai, Weiguo Zheng 5/18/2026

DeepSlide: From Artifacts to Presentation Delivery

DeepSlide multi-agent system generates presentations with focus on delivery process including narrative planning and presentation preparation, not just artifacts.

Ax Zhantao Wang 5/18/2026

SDOF: Taming the Alignment Tax in Multi-Agent Orchestration with State-Constrained Dispatch

SDOF framework treats multi-agent orchestration as constrained state machines, enforcing business process constraints in frameworks like LangGraph.

Ax Nanxu Gong, Zixin Chen, Haotian Li, Zishu Zhao, Jianxun Lian, Huamin Qu, Yanjie Fu, Xing Xie 5/18/2026

Does Theory of Mind Improvement Really Benefit Human-AI Interactions? Empirical Findings from Interactive Evaluations

Empirical evaluation showing Theory of Mind improvements in LLMs do not directly translate to better outcomes in interactive human-AI conversations.

Ax Duling Xu, Zheng Chen, Zaifeng Pan, Jiawei Guan, Dong Dong, Jialin Li, Bangzheng Pu 5/18/2026

SkillSmith: Compiling Agent Skills into Boundary-Guided Runtime Interfaces

SkillSmith compiles agent skills into boundary-guided runtime interfaces, reducing redundancy from irrelevant context and repeated reasoning in LLM-based systems.

Ax Jagdish Tripathy, Marcus Buckmann 5/18/2026

Fair outputs, Biased Internals: Causal Potency and Asymmetry of Latent Bias in LLMs for High-Stakes Decisions

Study examining latent biases in instruction-tuned LLMs for high-stakes decisions like mortgage underwriting, showing behavioral fairness masks internal bias.

Ax Chenying Lin, Yichen Hai, Yi He, Ran Wang, Haiyan Qiang, Liang Yu 5/18/2026

CAX-Agent: A Lightweight Agent Harness for Reliable APDL Automation

Agent harness architecture for MAPDL finite-element simulation providing structured execution control, tool encapsulation, and fault recovery for LLM reliability.

Ax Salman Avestimehr, Ken Duffy, Muriel M\'edard 5/18/2026

NOVA: Fundamental Limits of Knowledge Discovery Through AI

NOVA framework models iterative AI knowledge discovery as adaptive sampling, identifying conditions for genuine knowledge accumulation and failure modes.

Ax Jianbo Lin, Xiaomin Yu, Yi Xin, Yifu Guo, Zhuosong Jiang, Zhongqi Yue, Weishi Wang, Heqing Zou, Chengwei Qin, Hui Xiong 5/18/2026

ICRL: Learning to Internalize Self-Critique with Reinforcement Learning

ICRL uses reinforcement learning to help LLM agents internalize critique feedback, enabling improvement that persists without explicit critic guidance.

Ax Naruki Yoshikawa, Ryo Tamura 5/18/2026

NIMO Controller: a self-driving laboratory orchestrator based on the Model Context Protocol

NIMO Controller applies Model Context Protocol to self-driving laboratory orchestration, providing standardized agent interfaces for lab automation.

Ax Jun He, Deying Yu 5/18/2026

Verifiable Agentic Infrastructure: Proof-Derived Authorization for Sovereign AI Systems

Proof-derived authorization framework for sovereign AI systems addressing semantic safety of autonomous agent actions beyond credential-based identity checks.

Ax Han Li, Jinyu Tian, Rili Feng, Yuqiao Du, Chong Zheng, Chenyu Wang, Chenchen Liu, Shihao Li, Xinping Lei, Yifan Yao, Weihao Xie, Letian Zhu, Jiaheng Liu 5/18/2026

Solvita: Enhancing Large Language Models for Competitive Programming via Agentic Evolution

Solvita enables continuous learning in agentic systems for competitive programming by maintaining problem-solving experience across tasks rather than stateless execution.

Ax Jiachen Jiang, Huminhao Zhu, Zhihui Zhu 5/18/2026

SMCEvolve: Principled Scientific Discovery via Sequential Monte Carlo Evolution

SMCEvolve reframes LLM-driven program search as Sequential Monte Carlo sampling with principled convergence guarantees for automated scientific discovery.

Ax Jingjing Wang, Xiwen Chen, Wenhui Zhu, Huayu Li, Zhengxiao He, Feiyang Cai, Ana S. Carreon-Rascon, Xuanzhao Dong, Feng Luo 5/18/2026

Context Pruning for Coding Agents via Multi-Rubric Latent Reasoning

Context pruning technique for coding agents using multi-rubric latent reasoning to reduce irrelevant repository files in context, improving token efficiency.

Ax Kin Max Piamolini Gusm\~ao, Nathan Gavenski, Nir Oren, Felipe Meneguzzi 5/18/2026

Zero-Shot Goal Recognition with Large Language Models

First systematic study of zero-shot goal recognition using LLMs, showing LLMs better suited for abductive consistency evaluation than novel plan generation.

Ax Joshua C. Yang, Maurice Flechtner, Damian Dailisan, Michiel A. Bakker 5/18/2026

Belief Engine: Configurable and Inspectable Stance Dynamics in Multi-Agent LLM Deliberation

Belief Engine enables auditable stance dynamics in multi-agent LLM deliberation by treating beliefs as evidential states, making agent reasoning changes interpretable.

Ax Eugene Koran, Yejun Yun, Samantha Tetef, Benjamin Arnav, Pablo Bernabeu-P\'erez 5/18/2026

Ensemble Monitoring for AI Control: Diverse Signals Outweigh More Compute

Ensemble monitoring approach for AI agent safety that combines diverse monitoring signals to detect misaligned actions more reliably than individual monitors.

Ax Wei Sheng, Rohan Paleja 5/18/2026

Beyond Partner Diversity: An Influence-Based Team Steering Framework for Zero-Shot Human-Machine Teaming

Zero-shot human-machine teaming framework using influence-based steering to coordinate with unseen partners without requiring domain-specific training data.

Ax Ruobing Zuo, Hanrui Zhao, Gaolei He, Zhengfeng Yang, Jianlin Wang 5/18/2026

From LLM-Generated Conjectures to Lean Formalizations: Automated Polynomial Inequality Proving via Sum-of-Squares Certificates

Automated polynomial inequality proving combining LLM-generated conjectures with formal verification in Lean using sum-of-squares certificates for mathematical reasoning.

Ax Guruprasad Raghavan, George Nychis, Rohan Narayana Murthy 5/18/2026

X-SYNTH: Beyond Retrieval -- Enterprise Context Synthesis from Observed Human Attention

X-SYNTH proposes context synthesis for enterprise AI agents by analyzing human attention patterns across systems, moving beyond traditional retrieval-based approaches.

Ax Fangzhou Lin, Shuo Xing, Peiran Li, Siyuan Yang, Qianwen Ge, Kazunori Yamada, Ziming Zhang, Haichong Zhang, Zhengzhong Tu 5/18/2026

CAPS: Cascaded Adaptive Pairwise Selection for Efficient Parallel Reasoning

CAPS optimizes parallel reasoning in LLMs by adaptively selecting which pairwise verification judgments to perform, reducing computational cost while maintaining solution quality.

Ax Jing Wang, Shang Liu, Hangan Zhou, Zhiyao Xie 5/18/2026

RTL-BenchMT: Dynamic Maintenance of RTL Generation Benchmark Through Agent-Assisted Analysis and Revision

RTL-BenchMT agentic framework for dynamically maintaining RTL generation benchmarks through LLM-assisted analysis to address flawed cases and overfitting.

Ax Yichao Liu, Huawen Shen, Liu Yu, Shiyu Liu, Zeyu Chen, Yu Zhou 5/18/2026

DRS-GUI: Dynamic Region Search for Training-Free GUI Grounding

DRS-GUI dynamic region search method for training-free GUI grounding in MLLM-powered agents, focusing on high-resolution screenshot element localization.

Ax Sergei Chuprov, Richard D. Lange, Leon Reznik, Paulo Shakarian, Raman Zatsarenko, Dmitrii Korobeinikov 5/18/2026

Position: Artificial Intelligence Needs Meta Intelligence -- the Case for Metacognitive AI

Position paper advocating metacognition as design principle for AI systems to monitor internal states and allocate resources based on problem difficulty.

Ax Junle Wang, Xingchuang Liao, Wenjun Wu 5/18/2026

STAR: A Stage-attributed Triage and Repair framework for RCA Agents in Microservices

STAR framework for triaging and repairing LLM-based root cause analysis agents in microservices through stage-attributed error detection.

Ax Yuejia Li, Ke He, Junheng Li, Shutong Chen, Jingkang Xia, Zhiyue Su, Junchi Zhang, Mang Ye 5/18/2026

See Before You Code: Learning Visual Priors for Spatially Aware Educational Animation Generation

Visual priors approach for code generation of educational animations using render feedback to correct visual defects.

Ax Junle Wang, Xingchuang Liao, Wenjun Wu 5/18/2026

TopoEvo: A Topology-Aware Self-Evolving Multi-Agent Framework for Root Cause Analysis in Microservices

TopoEvo multi-agent framework for root cause analysis in microservices using topology-aware LLM agents to address cascading failures and drift.

Ax Lijie Ding, Changwoo Do 5/18/2026

ColPackAgent: Agent-Skill-Guided Hard-Particle Monte Carlo Workflows for Colloidal Packing

ColPackAgent framework using Model Context Protocol and agent skills for autonomous Monte Carlo simulations of colloidal packing.

Ax Keshava Chaitanya, Jahnavi Gundakaram 5/18/2026

PRISM: Prompt Reliability via Iterative Simulation and Monitoring for Enterprise Conversational AI

PRISM framework for LLM conversational agent reliability in enterprise through iterative simulation and monitoring of prompt regressions.

Ax Chanuk Lee, Sangwoo Park, Minki Kang, Sung Ju Hwang 5/18/2026

Nudging Beyond the Comfort Zone: Efficient Strategy-Guided Exploration for RLVR

Strategy-guided exploration method for reinforcement learning with verifiable rewards to improve LLM reasoning capabilities beyond brute-force scaling.

Ax Izabella Krzeminska, Michal Butkiewicz, Ewa Komkowska 5/18/2026

Can We Trust AI-Inferred User States. A Psychometric Framework for Validating the Reliability of Users States Classification by LLMs in Operational Environments

Psychometric framework validating reliability of LLM-inferred user state classifications in conversational systems.

Ax Xiang Li, Liping Yi, Mingze Kong, Min Zhang, Zhongxiang Dai, QingHua Hu 5/18/2026

ALSO: Adversarial Online Strategy Optimization for Social Agents

ALSO framework for adversarial online strategy optimization in social agents using LLMs to dynamically adapt strategies in multi-turn dialogues.

Ax Kean Shi, Zihang Li, Tianyi Ma, Zengji Tu, Jialong Wu, Xinbo Xu, Qingyao Yang, Ruoyu Wu, Weichu Xie, Ming Wu, Jason Zeng, Michael Heinrich, Elvis Zhang, Liang Chen, Kuan Li, Baobao Chang 5/18/2026

SaaS-Bench: Can Computer-Use Agents Leverage Real-World SaaS to Solve Professional Workflows?

SaaS-Bench benchmark evaluating computer-use agents on realistic professional workflows across real Software-as-a-Service platforms.

Ax Alberto Pepe, Chien-Yu Lin, Despoina Magka, Bilge Acun, Yannan Nellie Wu, Anton Protopopov, Carole-Jean Wu, Yoram Bachrach 5/18/2026

Agentic Discovery of Neural Architectures: AIRA-Compose and AIRA-Design

AIRA framework with dual LLM agents autonomously designing neural architectures beyond Transformers through architecture search and mechanistic implementation.

Ax Jingxuan Wei, Xi Bai, Shan Liu, Caijun Jia, Zheng Sun, Xinglong Xu, Siyuan Li, Linzhuang Sun, Bihui Yu, Conghui He, Cheng Tan 5/18/2026

PAGER: Bridging the Semantic-Execution Gap in Point-Precise Geometric GUI Control

PAGER framework for precise geometric GUI control in vision-language model agents, addressing pixel-accurate action placement in continuous canvas spaces.

Ax Fabio Rovai 5/18/2026

Deterministic Event-Graph Substrates as World Models for Counterfactual Reasoning

Event-graph world models representing agent state as append-only RDF triple logs for counterfactual reasoning and explanatory queries without learned components.

Ax Dillon Z. Chen, Till Hofmann, Toryn Q. Klassen, Sheila A. McIlraith 5/18/2026

Learning Bilevel Policies over Symbolic World Models for Long-Horizon Planning

Bilevel policy learning framework combining imitation learning over low-level continuous control with symbolic world models for long-horizon embodied AI agent planning.

Ax Michael Solodko, Justin Wagle 5/18/2026

ScreenSearch: Uncertainty-Aware OS Exploration

ScreenSearch system for GUI agents that explores desktop OS state under partial observability, reducing ambiguity before committing to actions.

Ax Parisa Kordjamshidi, Samer Aslan, Madhavan Seshadri, Leslie Barrett, Enrico Santus 5/18/2026

Reasoners or Translators? Contamination-aware Evaluation and Neuro-Symbolic Robustness in Tax Law

Study on LLM performance in tax law reasoning with contamination detection protocols to assess reliability and identify inflated performance from data contamination.

Ax Chinmay Savadikar, Mingyu Zhao, Yuanzheng Zhu, Han Li, Shuang Xie, Alberto Castelo, Tianfu Wu, Lingyun Wang 5/18/2026

ShopGym: An Integrated Framework for Realistic Simulation and Scalable Benchmarking of E-Commerce Web Agents

ShopGym: Realistic simulation framework and benchmark for evaluating e-commerce web agents with reproducible evaluation.

Ax Augusto B. Corr\^ea, Andr\'e G. Pereira, Jendrik Seipp 5/18/2026

Property-Guided LLM Program Synthesis for Planning

Property-guided LLM program synthesis using semantic constraints to reduce inference costs vs. random search.