This page shows the most relevant public items for cs.AI, ranked by trend activity and review signal. Use weekly for fast changes, monthly for more stable patterns, and all-time for evergreen picks.
Paper • Feb 20, 2026 • arXiv • Sethuraman et al., AAMAS 2026 Main Track
We propose Mixture-of-Reasoning Agents (MiRA), a zero-shot multimodal framework that decomposes reasoning across three specialized agents: Visual Analyzing, Text Comprehending, and Judge. By consol...
Paper • Feb 19, 2026 • arXiv • Google DeepMind Team
We introduce Gemini 3.1 Pro, an enhanced version of the Gemini 3 series optimized for rigorous logic and long-horizon problem solving. Built on a hybrid architecture that fuses linear attention wit...
Paper • Feb 11, 2026 • arXiv • Junjie Wang, Zequn Xie, Dan Yang, Jie Feng, Yue Shen, Jinjie Gu
Vision-language-action (VLA) models often suffer from the state gap issue when inferring open and long-horizon tasks in scientific scenarios. We propose an LLM-based agentic inference mechanism tha...
Paper • Feb 12, 2026 • arXiv • Nicholas Lee, Lutfi Eren Erdogan, Chris Joseph John, Surya Krishnapillai, Kurt Keutzer, Amir Gholami
Current WebAgents struggle with long-horizon tasks and complex navigation. We propose an agentic scaling framework that increases compute at test-time through iterative trajectory pruning and rewar...
Paper • Feb 12, 2026 • arXiv • Nenad Tomašev, Kevin R. McKee, Jack Rae, Iason Gabriel, Vukosi Marivate, Milind Tambe, Demis Hassabis, Charles Blundell
As advanced AI agents evolve beyond query-response models, their utility is increasingly defined by how effectively they can decompose complex objectives and delegate sub-tasks. We propose an adapt...
Paper • Feb 11, 2026 • arXiv • Sangyoon Bae, Miruna Oprescu, David Keetae Park, Shinjae Yoo, Jiook Cha
Recovering the true causal connectivity of the human brain from indirect observations like EEG and fMRI is a fundamental but ill-posed problem due to the presence of unobserved confounders and the ...
Paper • Feb 16, 2026 • arXiv • MiniMax Research Team
Training agents for industrial-scale deployment requires extreme stability and data throughput. We present Minimax M2.5, a model trained using a novel asynchronous RL architecture designed to proce...
Paper • Feb 19, 2026 • arXiv • Xiaoliang Fu, Jiaye Lin, Yangyi Fang
Policy optimization for Large Language Models often suffers from gradient instability and reward signal unreliability, particularly in mathematical and verifiable reasoning tasks. We introduce MASP...
Paper • Feb 18, 2026 • arXiv • Adam Zweiger, Xinghong Fu, Han Guo, MIT Team
Large Language Models struggle with memory overhead during long-context inference due to the linear growth of the Key-Value (KV) cache. We propose Attention Matching (AM), a framework for high-qual...
Paper • Feb 19, 2026 • arXiv • Yue Liu, Zhiyuan Hu, Flood Sung
Current LLM agents frequently fail in tasks requiring hundreds of steps due to error accumulation and context overflow. We introduce KLong, an agentic framework that utilizes 'Trajectory-Splitting ...
Paper • Feb 16, 2026 • arXiv • Chen et al.
This concept paper systematically reviews the emerging paradigm of LLM-based multi-agent systems, examining their applications across the entire Software Development Life Cycle (SDLC), from require...
Paper • Feb 15, 2026 • arXiv • Zheng et al.
Existing multi-agent frameworks often rely on static or task-level workflows, which either over-process simple queries or underperform on complex ones, while neglecting the efficiency-performance t...
Paper • Feb 18, 2026 • arXiv • Jianhao Ruan, et al.
Language agents have shown strong promise for task automation, but realizing this promise for increasingly complex, long-horizon tasks has driven the rise of a sub-agent-as-tools paradigm. However,...
Paper • Feb 17, 2026 • arXiv • Moritz Miller, Florent Draye, Bernhard Schölkopf
This paper addresses the fundamental challenge of 'feature disentanglement' in modern deep learning. We propose an Orthogonality Regularization technique to identify features that are both interpre...
Paper • Feb 15, 2026 • arXiv • DeepMind Research Team
We explore the boundary between simple heuristics and complex neural-cognitive models in combinatorial optimization. This paper demonstrates how hybrid architectures can leverage memory to shape re...
Paper • Feb 18, 2026 • arXiv • Moritz Miller, Florent Draye, Bernhard Schölkopf
This paper distinguishes between two paradigms in AI explanation: static prediction and agentic trajectories. In agentic systems, behavior emerges as a sequence of observations, reasoning steps, an...
Paper • Feb 4, 2026 • arXiv • Fan Yang, Rui Meng, Trudi Di Qi, Ali Ezzati, Yuxin Wen
Direct Preference Optimization (DPO) and its variants have revolutionized LLM alignment, yet they struggle when the preferred choice requires deep, multi-step reasoning. We introduce KEPO, a framew...
Paper • Feb 17, 2026 • arXiv • Zhipu AI Team, Tsinghua University Researchers
We present GLM-5, a foundation model designed to bridge the gap between human-guided 'vibe coding' and autonomous 'agentic engineering.' GLM-5 introduces DeepSeek-inspired Sparse Attention (DSA) to...
Paper • Feb 10, 2026 • arXiv • Kewei Hu, Michael Zhang, Hanwen Kang
We introduce Grounded Scene-graph Reasoning (GSR), a structured reasoning paradigm that explicitly models world-state evolution as transitions over semantically grounded scene graphs. By reasoning ...
Paper • Feb 12, 2026 • arXiv • Antonin Sulc
As multi-agent systems evolve into autonomous swarms, debugging failures requires reasoning about knowledge, belief, and obligation. We present Differentiable Modal Logic (DML), implemented via Mod...