HN shubhamoriginx 3/20/2026

Ask HN: How do you programmatically evaluate if an LLM sounds "too AI"?

Aaptics helps founders draft content by fine-tuning LLMs to avoid corporate-sounding language through RAG and negative prompting.

HN isaacsight 3/20/2026

Show HN: Kbot – terminal AI agent that learns from every user who uses it

kbot is an open-source terminal AI agent with 23 agents, 290 tools, and 20 providers. Multi-model, local-first, works with MCP-compatible IDEs.

Ax Yinghui Li, Jiayi Kuang, Peng Xing, Daixian Liu, Junnan Dong, Shu-Yu Guo, Yangning Li, Qingyu Zhou, Wenhao Jiang, Hai-Tao Zheng, Ying Shen, Liang Lin, Philip S. Yu 3/20/2026

Cognitive Mismatch in Multimodal Large Language Models for Discrete Symbol Understanding

Benchmark evaluating multimodal LLMs' ability to process discrete symbols like math formulas and chemical structures, addressing gap in symbol understanding.

Ax Zizhao Hu, Mohammad Rostami, Jesse Thomason 3/20/2026

Expert Personas Improve LLM Alignment but Damage Accuracy: Bootstrapping Intent-Based Persona Routing with PRISM

Introduces PRISM for intent-based persona routing in LLMs, improving both alignment and accuracy in multi-agent systems through selective persona application.

Ax Jungmyung Wi, Hyunsoo Kim, Donghyun Kim 3/20/2026

Correlation-Weighted Multi-Reward Optimization for Compositional Generation

Proposes correlation-weighted multi-reward optimization to improve compositional generation in text-to-image models by reducing concept interference.

Ax Enoch Hyunwook Kang 3/20/2026

Reasonably reasoning AI agents can avoid game-theoretic failures in zero-shot, provably

Studies how reasonably reasoning AI agents can avoid game-theoretic failures in interactive economic environments without post-training alignment methods.

Ax Yicheng Hu, Xinyu Lin, Shulin Li, Wenjie Wang, Fengbin Zhu, Fuli Feng 3/20/2026

CAPSUL: A Comprehensive Human Protein Benchmark for Subcellular Localization

Presents CAPSUL benchmark dataset for protein subcellular localization with 3D structural information for structure-based ML models.

Ax Jerome Ramos, Feng Xia, Xi Wang, Shubham Chatterjee, Xiao Fu, Hossein A. Rahmani, Aldo Lipani 3/20/2026

Interplay: Training Independent Simulators for Reference-Free Conversational Recommendation

Proposes Interplay, training independent simulators for conversational recommendation systems to generate reference-free dialogue data at scale.

Ax Zhihui Chen, Kai He, Qingyuan Lei, Bin Pu, Jian Zhang, Yuling Xu, Mengling Feng 3/20/2026

MedForge: Interpretable Medical Deepfake Detection via Forgery-aware Reasoning

Proposes MedForge for interpretable medical deepfake detection using MLLMs with explainable forgery-aware reasoning for healthcare applications.

Ax Wanjia Zhao, Ludwig Schmidt, James Zou, Vidhisha Balachandran, Lingjiao Chen 3/20/2026

ZEBRAARENA: A Diagnostic Simulation Environment for Studying Reasoning-Action Coupling in Tool-Augmented LLMs

Introduces ZebraArena, a procedurally generated diagnostic environment for evaluating reasoning-action coupling in tool-augmented LLMs with minimal dataset contamination.

Ax Ping Chen, Daoxuan Zhang, Xiangming Wang, Yungeng Liu, Haijin Zeng, Yongyong Chen 3/20/2026

Agentic Flow Steering and Parallel Rollout Search for Spatially Grounded Text-to-Image Generation

Presents AFS-Search for text-to-image generation using agentic flow steering and parallel rollout search to improve spatial reasoning and reduce error accumulation.

Ax Zhixing You, Jiachen Yuan, Jason Cai 3/20/2026

D-Mem: A Dual-Process Memory System for LLM Agents

Introduces D-Mem, a dual-process memory system for LLM agents enabling high-fidelity memory access for long-horizon reasoning and autonomous operation.

Ax Huansheng Ning, Jianguo Ding 3/20/2026

An Onto-Relational-Sophic Framework for Governing Synthetic Minds

Discusses governance frameworks for synthetic minds and AI regulation, focusing on conceptual foundations beyond tool-centric approaches.

Ax Shaked Perek, Ben Wiesel, Avihu Dekel, Nimrod Shabtay, Eli Schwartz 3/20/2026

Balanced Thinking: Improving Chain of Thought Training in Vision Language Models

Proposes SCALe method to improve chain-of-thought training in vision-language models by addressing token imbalance between reasoning traces and answer segments.

Ax Haokun Zhao, Wanshi Xu, Haidong Yuan, Songjun Cao, Long Ma, Yanghua Xiao 3/20/2026

Thinking with Constructions: A Benchmark and Policy Optimization for Visual-Text Interleaved Geometric Reasoning

Benchmark and policy optimization for visual-text geometric reasoning with dynamic construction. Addresses strategic diagram generation in multimodal LLM agents.

Ax Zuher Jahshan, Ben Ben Ishay, Leonid Yavits 3/20/2026

MANAR: Memory-augmented Attention with Navigational Abstract Conceptual Representation

Memory-augmented attention layer inspired by Global Workspace Theory for contextualization. Cognitive model-based improvements to multi-head attention mechanisms.

Ax Lei Gao, Hengda Bao, Jingfei Fang, Guangzheng Wu, Weihua Zhou, Yun Zhou 3/20/2026

Accurate and Efficient Multi-Channel Time Series Forecasting via Sparse Attention Mechanism

Sparse attention architecture for multi-channel time series forecasting. Machine learning for finance/supply chain, not LLM or agent-focused.

Ax Minhua Lin, Zhiwei Zhang, Hanqing Lu, Hui Liu, Xianfeng Tang, Qi He, Xiang Zhang, Suhang Wang 3/20/2026

MemMA: Coordinating the Memory Cycle through Multi-Agent Reasoning and In-Situ Self-Evolution

Multi-agent memory coordination framework optimizing construction, retrieval, and utilization cycles. Applies multi-agent reasoning to improve memory-augmented LLM agent performance.

Ax Martina Ullasci, Marco Rondina, Riccardo Coppola, Flavio Giobergia, Riccardo Bellanca, Gabriele Mancari Pasi, Luca Prato, Federico Spinoso, Silvia Tagliente 3/20/2026

Analysis Of Linguistic Stereotypes in Single and Multi-Agent Generative AI Architectures

Analysis of dialect-sensitive stereotypes in single and multi-agent LLM architectures. Studies bias variation across Standard American and African-American English inputs.

Ax Huichi Zhou, Siyuan Guo, Anjie Liu, Zhongwei Yu, Ziqin Gong, Bowen Zhao, Zhixun Chen, Menglong Zhang, Yihang Chen, Jinsong Li, Runyu Yang, Qiangbin Liu, Xinlei Yu, Jianmin Zhou, Na Wang, Chunyang Sun, Jun Wang 3/20/2026

Memento-Skills: Let Agents Design Agents

LLM agent system that autonomously designs task-specific agents through memory-based RL and stateful prompts. Meta-agent framework with skill-based continual learning.

Ax Duc Hao Pham, Van Duy Truong, Duy Khanh Dinh, Tien Cuong Nguyen, Dien Hy Ngo, Tuan Anh Bui 3/20/2026

A Concept is More Than a Word: Diversified Unlearning in Text-to-Image Diffusion Models

Method for concept unlearning in text-to-image diffusion models beyond keyword-based approaches. Addresses selective content removal from generative models.

Ax Nitay Alon, Joseph M. Barnby, Reuth Mirsky, Stefan Sarkadi 3/20/2026

Proceedings of the 2nd Workshop on Advancing Artificial Intelligence through Theory of Mind

Workshop proceedings on Theory of Mind in AI research. Collection of papers on cognitive modeling and AI understanding.

Ax Wenxuan Zhang, Lemeng Wu, Changsheng Zhao, Ernie Chang, Mingchen Zhuge, Zechun Liu, Andy Su, Hanxian Huang, Jun Chen, Chong Zhou, Raghuraman Krishnamoorthi, Vikas Chandra, Mohamed Elhoseiny, Wei Wen 3/20/2026

dTRPO: Trajectory Reduction in Policy Optimization of Diffusion Large Language Models

Policy optimization technique for diffusion LLMs reducing trajectory computation cost. Improves efficiency of preference alignment in generative language models.

Ax Xiaoyang Chen, Xiang Jiang 3/20/2026

Can LLM generate interesting mathematical research problems?

Evaluation of LLM capability to generate novel mathematical research problems. Studies mathematical creativity and problem generation in language models.

Ax Hao Zhang, Mingjie Liu, Shaokun Zhang, Songyang Han, Jian Hu, Zhenghui Jin, Yuchi Zhang, Shizhe Diao, Ximing Lu, Binfeng Xu, Zhiding Yu, Jan Kautz, Yi Dong 3/20/2026

ProRL Agent: Rollout-as-a-Service for RL Training of Multi-Turn LLM Agents

Service architecture for distributed RL training of multi-turn LLM agents. Decouples rollout orchestration from training for scalable agent development.

Ax Xiao Feng, Bo Han, Zhanke Zhou, Jiaqi Fan, Jiangchao Yao, Ka Ho Li, Dahai Yu, Michael Kwok-Po Ng 3/20/2026

RewardFlow: Topology-Aware Reward Propagation on State Graphs for Agentic RL with Large Language Models

Topology-aware reward propagation for RL training of LLM agents. Addresses sparse reward problem in agentic LLM reasoning with graph-based methods.

Ax Xuemian Wu, Shizhe Zhao, Zhongqiang Ren 3/20/2026

Conflict-Based Search for Multi Agent Path Finding with Asynchronous Actions

Multi-agent path finding algorithm with asynchronous action support. Graph search problem unrelated to LLMs or AI agents.

Ax Gaoxiang Cao, Wenke Yuan, Huasen He, Yunpeng Hou, Xiaofeng Jiang, Shuangwu Chen, Jian Yang 3/20/2026

Bridging Network Fragmentation: A Semantic-Augmented DRL Framework for UAV-aided VANETs

DRL framework for UAV network deployment in vehicular networks. Reinforcement learning application outside core AI/LLM focus areas.

Ax Krzysztof Janowicz, Gengchen Mai, Rui Zhu, Song Gao, Zhangyu Wang, Yingjie Hu, Lauren Bennett 3/20/2026

Geography According to ChatGPT -- How Generative AI Represents and Reasons about Geography

Study analyzing how ChatGPT represents and reasons about geographic knowledge. Evaluates factual reasoning and world modeling in LLMs.

Ax Pranjal Aggarwal, Marjan Ghazvininejad, Seungone Kim, Ilia Kulikov, Jack Lanchantin, Xian Li, Tianjian Li, Bo Liu, Graham Neubig, Anaelia Ovalle, Swarnadeep Saha, Sainbayar Sukhbaatar, Sean Welleck, Jason Weston, Chenxi Whitehouse, Adina Williams, Jing Xu, Ping Yu, Weizhe Yuan, Jingyu Zhang, Wenting Zhao 3/20/2026

Reasoning over mathematical objects: on-policy reward modeling and test time aggregation

Research on LLM mathematical reasoning with formal expression derivation. Addresses structured reasoning in STEM via language models.

Ax Nicolas Martorell 3/20/2026

Quantitative Introspection in Language Models: Tracking Internal States Across Conversation

Develops quantitative introspection methods inspired by psychology to track internal state changes in LLMs across conversations using numeric self-report.

Ax Vedanta S P, Ponnurangam Kumaraguru 3/20/2026

I Can't Believe It's Corrupt: Evaluating Corruption in Multi-Agent Governance Systems

Evaluates whether multi-agent LLM governance systems follow institutional rules when granted authority, finding integrity requires pre-deployment safeguards.

Ax Matt Gorbett, Suman Jana 3/20/2026

Secure Linear Alignment of Large Language Models

Studies cross-model alignment of LLM representations for downstream objectives with applications in privacy-preserving and security-constrained settings.

Ax Diego Calvanese, Angelo Casciani, Giuseppe De Giacomo, Marlon Dumas, Fabiana Fournier, Timotheus Kampik, Emanuele La Malfa, Lior Limonad, Andrea Marrella, Andreas Metzger, Marco Montali, Daniel Amyot, Peter Fettke, Artem Polyvyanyy, Stefanie Rinderle-Ma, Sebastian Sardi\~na, Niek Tax, Barbara Weber 3/20/2026

Agentic Business Process Management: A Research Manifesto

Research manifesto proposing Agentic Business Process Management paradigm extending BPM for governing autonomous agents executing organizational processes.

Ax Dario Compagno, Fabio Massimo Zennaro 3/20/2026

Teleological Inference in Structural Causal Models via Intentional Interventions

Extends structural causal models with intentional interventions operator for teleological inference about goal-directed agent behavior in causal systems.

Ax Peng Gang 3/20/2026

Evaluating 5W3H Structured Prompting for Intent Alignment in Human-AI Interaction

Evaluates PPS (5W3H-based structured prompting framework) for reducing intent transmission loss between users and LLMs across business, technical, and travel domains.

Ax Pronob Kumar Barman, Pronoy Kumar Barman 3/20/2026

Unmasking Algorithmic Bias in Predictive Policing: A GAN-Based Simulation Framework with Multi-City Temporal Analysis

GAN-based simulation framework measuring racial bias propagation in predictive policing systems across multiple US cities with temporal analysis.

Ax Chun-Jui Wang, Jian-Ting Guo, Hung Guei, Chung-Chin Shih, Ti-Rong Wu, I-Chen Wu 3/20/2026

Evaluating Game Difficulty in Tetris Block Puzzle

Uses Stochastic Gumbel AlphaZero to evaluate difficulty in Tetris Block Puzzle variants, applying game-playing AI as evaluator for puzzle design.

Ax Rui Chai 3/20/2026

Regret Bounds for Competitive Resource Allocation with Endogenous Costs

Analyzes online resource allocation among interacting modules with endogenous costs under uniform, gated, and competitive allocation paradigms with regret bounds.

Ax Jonah Leshin, Manish Shah, Ian Timmis, Daniel Kang 3/20/2026

Behavioral Fingerprints for LLM Endpoint Stability and Identity

Stability Monitor: behavioral fingerprinting system tracking LLM endpoint identity changes from model updates, quantization, inference engines beyond traditional uptime metrics.

Ax Qiawen Ella Liu, Marina Dubova, Henry Conklin, Takumi Harada, Thomas L. Griffiths 3/20/2026

Serendipity by Design: Evaluating the Impact of Cross-domain Mappings on Human and LLM Creativity

Evaluates whether cross-domain mapping interventions increase creativity equally in humans and LLMs through product feature generation experiments.

Ax Dana\'e Broustail, Anna Tegon, Thorir Mar Ingolfsson, Yawei Li, Luca Benini 3/20/2026

LuMamba: Latent Unified Mamba for Electrode Topology-Invariant and Efficient EEG Modeling

LuMamba: self-supervised Mamba architecture for EEG modeling with topology-invariant electrode handling and improved computational efficiency over Transformers.

Ax Maksym Del, Markus K\"angsepp, Marharyta Domnich, Ardi Tampuu, Lisa Yankovskaya, Meelis Kull, Mark Fishel 3/20/2026

How Uncertainty Estimation Scales with Sampling in Reasoning Models

Studies how uncertainty estimation scales with parallel sampling in reasoning models using self-consistency and verbalized confidence across mathematics and STEM tasks.

Ax Qiang Li, XiangRui Zhang, Haining Wang 3/20/2026

Implicit Patterns in LLM-Based Binary Analysis

Large-scale trace-level study showing multi-pass LLM reasoning in binary vulnerability analysis exhibits structured, token-level exploration patterns across hundreds of steps.

Ax Jonathan Lys, Vincent Gripon, Bastien Pasdeloup, Axel Marmoret, Lukas Mauch, Fabien Cardinaux, Ghouthi Boukli Hacene 3/20/2026

D5P4: Partition Determinantal Point Process for Diversity in Parallel Discrete Diffusion Decoding

D5P4: generalized beam-search framework using determinantal point processes for diverse parallel decoding in discrete diffusion text generation models.

Ax Yuyang Liu 3/20/2026

cuGenOpt: A GPU-Accelerated General-Purpose Metaheuristic Framework for Combinatorial Optimization

cuGenOpt: GPU-accelerated metaheuristic framework for combinatorial optimization balancing generality, performance, and usability across logistics and scheduling problems.

Ax Zou Qiang 3/20/2026

Box Maze: A Process-Control Architecture for Reliable LLM Reasoning

Box Maze framework enforces LLM reasoning integrity through process-control architecture to mitigate hallucination and unreliable reasoning under adversarial prompting.

Ax Zehao Li, Zhenyu Wu, Yibo Zhao, Bowen Yang, Jingjing Xie, Zhaoyang Liu, Zhoumianze Liu, Kaiming Jin, Jianze Liang, Zonglin Li, Feng Wu, Bowen Zhou, Zun Wang, Zichen Ding 3/20/2026

OS-Themis: A Scalable Critic Framework for Generalist GUI Rewards

OS-Themis: scalable multi-agent critic framework using decomposed trajectory milestones for training robust GUI agents with reinforcement learning.

Ax Sawsan Alqahtani, Garima Lalwani, Yi Zhang, Salvatore Romeo, Saab Mansour 3/20/2026

Using Optimal Transport as Alignment Objective for fine-tuning Multilingual Contextualized Embeddings

Uses optimal transport as alignment objective for fine-tuning multilingual contextualized embeddings to improve cross-lingual word representations.

Ax Anna Babarczy, Andras Lukacs, Peter Vedres, Zeteny Bujka 3/20/2026

Do Large Language Models Possess a Theory of Mind? A Comparative Evaluation Using the Strange Stories Paradigm

Comparative study evaluating whether LLMs demonstrate Theory of Mind capabilities using psychological paradigms.