2/18/2026
ARGUS: Adaptive Rotation-Invariant Geometric Unsupervised System
ARGUS detects distributional drift in high-dimensional data streams using local statistics over fixed spatial partitions of data manifold.
2/18/2026
Stratified Hazard Sampling: Minimal-Variance Event Scheduling for CTMC/DTMC Discrete Diffusion and Flow Models
Stratified hazard sampling reduces variance in discrete diffusion/flow models by optimizing event scheduling in CTMC/DTMC processes.
2/18/2026
PROMA: Projected Microbatch Accumulation for Reference-Free Proximal Policy Updates
PROMA: reference-free proximal policy method for LLM training that controls KL divergence via gradient projection without reference model.
2/18/2026
Orthogonalized Policy Optimization:Decoupling Sampling Geometry from Optimization Geometry in RLHF
OPO: theoretical framework for LLM alignment using constrained proximal policy optimization with work-dissipation principle and chi-square geometry.
2/18/2026
Improving Policy Exploitation in Online Reinforcement Learning with Instant Retrospect Action
Instant Retrospect Action algorithm improves policy exploitation in online RL through Q-network representation learning.
2/18/2026
Green-NAS: A Global-Scale Multi-Objective Neural Architecture Search for Robust and Efficient Edge-Native Weather Forecasting
Green-NAS multi-objective neural architecture search optimizes weather forecasting models for efficiency and carbon footprint.
2/18/2026
Don't Forget Its Variance! The Minimum Path Variance Principle for Accurate and Stable Score-Based Models
MinPV Principle minimizes path variance in score-based models to improve accuracy and stability.
2/18/2026
On the Role of Iterative Computation in Reinforcement Learning
Analyzes role of iterative computation in RL, showing policies benefit from additional compute beyond fixed parameters.
2/18/2026
Achieving Optimal Static and Dynamic Regret Simultaneously in Bandits with Deterministic Losses
Algorithm achieving simultaneous optimal static and dynamic regret in adversarial multi-armed bandits.
2/18/2026
Horizon Imagination: Efficient On-Policy Rollout in Diffusion World Models
Horizon Imagination improves efficiency of diffusion-based world models for RL by denoising multiple future observations.
2/18/2026
How Well Do Large-Scale Chemical Language Models Transfer to Downstream Tasks?
Systematic evaluation of chemical language model scaling on molecular property prediction downstream tasks.
2/18/2026
Safe Reinforcement Learning via Recovery-based Shielding with Gaussian Process Dynamics Models
Recovery-based shielding framework integrates Gaussian process models with RL for provably safe control in continuous systems.
2/18/2026
Online GPU Energy Optimization with Switching-Aware Bandits
Online GPU energy optimization using bandit algorithms to reduce power consumption in HPC systems.
2/18/2026
Linear Bandits beyond Inner Product Spaces, the case of Bandit Optimal Transport
Extends linear bandits theory beyond inner product spaces using optimal transport for recommendation and clinical systems.
2/18/2026
Can Multimodal LLMs Perform Time Series Anomaly Detection?
Evaluates multimodal LLMs and vision-language models for time series anomaly detection in systems monitoring.
2/18/2026
Multimodal Integrated Knowledge Transfer to Large Language Models through Preference Optimization with Biomedical Applications
MINT framework aligns LLMs with biomedical knowledge using preference optimization on multimodal data.
2/18/2026
cadrille: Multi-modal CAD Reconstruction with Reinforcement Learning
Cadrille uses reinforcement learning for multi-modal CAD reconstruction from point clouds, images, and text inputs.
2/18/2026
Prover Agent: An Agent-Based Framework for Formal Mathematical Proofs
AI agent framework integrating LLMs with Lean formal proof assistant for automated theorem proving.
2/18/2026
NPG-Muse: Scaling Long Chain-of-Thought Reasoning with NP-Hard Graph Problems
Scaling long chain-of-thought reasoning in LLMs using NP-hard graph problems for cost-effective training.
2/18/2026
Agents of Discovery
Autonomous AI agents for physics data analysis using machine learning in particle physics research.
2/18/2026
Improving User Interface Generation Models from Designer Feedback
RLHF approaches for improving LLM-based UI generation using designer feedback and rationale.
2/18/2026
VLCE: A Knowledge-Enhanced Framework for Image Description in Disaster Assessment
Vision-language model enhancement framework for improved disaster assessment image descriptions.
2/18/2026
Toward Safer Diffusion Language Models: Discovery and Mitigation of Priming Vulnerability
Safety vulnerability analysis of diffusion language models and mitigation strategies for jailbreak attacks.
2/18/2026
Generalized Parallel Scaling with Interdependent Generations
Method for scaling parallel LLM inference by enabling interdependent token generation across multiple responses.
2/18/2026
PolySHAP: Extending KernelSHAP with Interaction-Informed Polynomial Regression
arXiv paper extending KernelSHAP with interaction-informed polynomial regression for efficient Shapley value approximation.
2/18/2026
FlowSteer: Interactive Agentic Workflow Orchestration via End-to-End Reinforcement Learning
arXiv paper on FlowSteer, an end-to-end reinforcement learning framework for interactive agentic workflow orchestration.
2/18/2026
Grappa: Gradient-Only Communication for Scalable Graph Neural Network Training
arXiv paper on Grappa, a distributed GNN training framework using gradient-only communication for scalability.
2/18/2026
Online Fine-Tuning of Pretrained Controllers for Autonomous Driving via Real-Time Recurrent RL
arXiv paper on online fine-tuning pretrained policies for autonomous driving using recurrent reinforcement learning.
2/18/2026
Amortised and provably-robust simulation-based inference
arXiv paper on robust simulation-based inference using generalized Bayesian inference and neural network approximation.
2/18/2026
Curriculum Learning and Pseudo-Labeling Improve the Generalization of Multi-Label Arabic Dialect Identification Models
arXiv paper on curriculum learning and pseudo-labeling for multi-label Arabic dialect identification.
2/18/2026
Causally constrained reduced-order neural models of complex turbulent dynamical systems
arXiv paper on causal constraints in neural emulators of turbulent systems using response theory and score matching.
2/18/2026
Tabular Foundation Models Can Learn Association Rules
arXiv paper demonstrating tabular foundation models effectively learn association rules for knowledge discovery in tabular data.
2/18/2026
Pentagon might ask contractors to certify they don't use Anthropic's Claude
Report that US Pentagon may require contractors to certify non-use of Anthropic's Claude API.
2/18/2026
If AI Agents Do the Work, Who Pays for the Seat?
Analysis of how AI agent adoption disrupts seat-based SaaS pricing models and software business economics.
2/18/2026
Agent Skills Hub – Security first directory for AI agent skills and MCP
Security-audited directory of 458+ MCP skills and integrations for AI agents; graded on adoption and audit scores.
2/18/2026
Stop building generic AI chatbots: 45% of support leaders are ahead
Business insight that 45% of support leaders are moving beyond generic AI chatbots to specialized implementations.
2/18/2026
Show HN: Rot – Financial Intelligence MCP Server
MCP server converting Reddit sentiment into options trading signals via Claude; 9-stage pipeline achieving 52% win rate on live trades.
2/18/2026
Show HN: Agent Audit Kit v0.1 – deterministic replay + stress for LLM agents
Agent Audit Kit v0.1 provides deterministic replay and stress testing for LLM agents. Minimal details in title-only post.
2/18/2026
Show HN: DevDay – End-of-day recap for AI coding sessions
DevDay CLI tool aggregates AI coding session data from multiple tools (Cursor, Claude Code, OpenCode), syncs with git, generates standup summaries locally.
2/18/2026
Show HN: AIBenchy – Independent AI Leaderboard
Independent AI leaderboard with custom benchmarks testing real end-user and developer scenarios beyond saturated MMLU/HumanEval tests.
2/18/2026
A Guide to Which AI to Use in the Agentic Era
Guide explaining shift from chatbot interactions to agentic AI usage where models autonomously accomplish tasks with tools.
2/18/2026
Show HN: Conduit: One Swift interface for every AI provider, on-device and cloud
Conduit: Swift library providing unified interface across multiple AI providers (cloud and on-device). Actor-based architecture prevents vendor lock-in with one-line provider switching.
2/18/2026
Show HN: AgentForge – Multi-LLM Orchestrator in 15KB
AgentForce: 15KB Python multi-LLM orchestrator replacing LangChain. 6.5x faster latency, 75% memory reduction, 1000x smaller footprint, 89% LLM cost reduction via Redis caching.
2/18/2026
AI agents with wallets and self-update loops
Incomplete submission title only, no content provided.
2/18/2026
I got tired of on-device LLMs crashing my apps, so I built a managed runtime
Edge-Veda: Managed on-device Flutter runtime for LLMs (text, vision, speech, RAG) with stability guarantees and privacy. Supports Llama 3.2, Qwen, SmolVLM models.
2/18/2026
Show HN: GreedyPhrase – 1.21x better compression than GPT-4o tiktoken, 6x faster
GreedyPhrase tokenizer achieves 1.21x better compression than GPT-4o tiktoken with 65K vocabulary and 6x faster throughput.
2/18/2026
Practical Guide to Building Reliable AI Agents
Guide covering AI agent design principles, patterns, and techniques including prompt engineering, specialist coordination, and iteration.
2/18/2026
The most practical, fast, tiny command sandboxing for AI agents
Command sandboxing solution designed for safe execution of code in AI agent environments.
2/18/2026
First Agent Skills Hackathon by the Authors of SkillsBench
First hackathon event for agent skills organized by SkillsBench authors.
2/18/2026
Rathbun's Operator
First-person account of user operating OpenClaw agent over two weeks, describing agent capabilities and behaviors observed during extended use.