Author

SmartApe AI

SmartApe's AI technology content account, focusing on cutting-edge AI developments and open-source frontlines.

Articles

Likes

Views

Comments

Latest from SmartApe AI

70 recent articles

SmartApe AI

Nov 3, 2026 · Artificial Intelligence

DexHoldem: HKU's Poker-Table Benchmark Unifies Robot Dexterity and Reasoning Evaluation

Hong Kong University and Yisheng Technology introduce DexHoldem, a real-robot benchmark using Texas Hold'em to test both dexterous manipulation (1,470 demos, 14 skills, UR10e+Shadow hand) and embodied perception/reasoning (36 questions, 14 model combos), with metrics TCR and SPSR revealing gaps between action success and scene preservation.

DexHoldemNeurIPS 2026Sim-to-Real

0 likes · 10 min read

DexHoldem: HKU's Poker-Table Benchmark Unifies Robot Dexterity and Reasoning Evaluation

SmartApe AI

Nov 2, 2026 · Artificial Intelligence

Proactive AI for Superintelligent Agents: Google & UIUC's 6A Framework

This article summarizes a 203-page survey from Google, UIUC, Stanford, UC Berkeley, and CMU on proactive AI for superintelligent agents, detailing the 6A framework (Awareness, Anticipation, Agenda, Arbitration, Action, Adaptation), evaluation methods across task, mechanism, and user levels, and applications in 13 domains.

6A FrameworkAI evaluationGoogle

0 likes · 13 min read

Proactive AI for Superintelligent Agents: Google & UIUC's 6A Framework

SmartApe AI

Nov 2, 2026 · Artificial Intelligence

RSI-Nav: Self-Improving Robot Navigation with Unlabeled Trajectories, +17.7% Success

Fudan University's RSI-Nav uses a single Qwen3.5-9B model to both follow language instructions and generate them from unlabeled trajectories, then verifies generated instructions by executing them in simulation, filtering successful ones to iteratively improve navigation performance, achieving 68.6% and 72.4% success on R2R-CE and RxR-CE benchmarks, a 15.7 and 17.7 point gain, and demonstrates real-world deployment on a DEEP Robotics M20 quadruped.

Data-Centric AIInstruction GenerationQwen3.5-9B

0 likes · 14 min read

RSI-Nav: Self-Improving Robot Navigation with Unlabeled Trajectories, +17.7% Success

SmartApe AI

Oct 31, 2026 · Artificial Intelligence

Trustworthy Recursive Self-Improvement in Large Models: Foundations, Risks & Evaluation

A survey from Renmin University and UCAS outlines a four-stage framework for recursive self-improvement (RSI) in large models, defines four trustworthiness themes, three evaluation levels, and seven open challenges including verifying self-modifying improvement mechanisms and tracking long-term risks.

AI AlignmentAI SafetyAI evaluation

0 likes · 13 min read

Trustworthy Recursive Self-Improvement in Large Models: Foundations, Risks & Evaluation

SmartApe AI

Oct 31, 2026 · Artificial Intelligence

MIPU: New RL Method Bridges LLM Training-Inference Gap

Researchers from Tianjin University and Alibaba propose MIPU, a reinforcement learning method that addresses training-inference mismatch in LLMs by directly optimizing inference-time performance, achieving higher accuracy and more stable training on Qwen3 models across math benchmarks.

LLM TrainingMIPI PrincipleMIPU

0 likes · 11 min read

MIPU: New RL Method Bridges LLM Training-Inference Gap

SmartApe AI

Oct 30, 2026 · Artificial Intelligence

ZwZ2: Single-Image Evidence Connection Boosts Multimodal Reasoning Across Tasks

Shanghai Jiao Tong University researchers introduce ZwZ2 and the EvidenceConnect framework, which synthesizes 20.4K single-image training samples requiring models to connect evidence across multiple regions, yielding 4B, 9B, and 27B models that surpass base models on 25 benchmarks and transfer gains to multi-image, spatial reasoning, and agent tasks.

EvidenceConnectGRPOMultimodal Models

0 likes · 14 min read

ZwZ2: Single-Image Evidence Connection Boosts Multimodal Reasoning Across Tasks

SmartApe AI

Oct 30, 2026 · Artificial Intelligence

Oxford Survey: One World, Four Operations for Video Foundation Models

Oxford University, NUS, and NTU researchers present a unified survey framing video foundation models around four operations—Reading, Writing, Sharing, and Interacting—distinguishing temporal coherence from world consistency, proposing an evidence ladder, and identifying seven open challenges for world modeling.

Oxford surveyVideo GenerationWorld Models

0 likes · 16 min read

Oxford Survey: One World, Four Operations for Video Foundation Models

SmartApe AI

Oct 30, 2026 · Artificial Intelligence

ClawLoop: Harness-Native Agent Self-Improvement Loop Without Stronger Teacher Models

Meituan and Peking University propose ClawLoop, a Harness-native framework closing the loop of task generation, model training, and diagnostic feedback, enabling agents to self-improve without stronger teacher models; experiments on Qwen3.6-27B show consistent gains across three benchmarks after three iterations, with SkillsBench rising from 46.3 to 57.2.

BenchmarkingClawLoopGRPO

0 likes · 11 min read

ClawLoop: Harness-Native Agent Self-Improvement Loop Without Stronger Teacher Models

SmartApe AI

Oct 30, 2026 · Artificial Intelligence

DeepSeek & Kimi Face Benchmark Gap & GPT-6 Distillation Allegations

Chinese AI models DeepSeek and Kimi face widening benchmark gaps against top closed-source models per Artificial Analysis v5, while The Information alleges they distilled OpenAI's GPT-6 Astra; the article also highlights their technical achievements and argues against dismissing contributions based on unverified claims.

AI benchmarksChinese AIDeepSeek

0 likes · 9 min read

DeepSeek & Kimi Face Benchmark Gap & GPT-6 Distillation Allegations

SmartApe AI

Oct 29, 2026 · Artificial Intelligence

DeepSeek Harness Roadmap Unveils Mobile Access, Persistent Memory & Multi-Agent Teams

DeepSeek Harness publishes its six-direction product roadmap adding remote and mobile access, personalized long-term memory, browser and GUI automation, agent team collaboration, session sharing for multi-user workflows, and enhanced sandbox security, while revealing it leads as the most-used coding agent among DeepSeek API users.

AI coding agentDeepSeek HarnessGUI automation

0 likes · 9 min read

DeepSeek Harness Roadmap Unveils Mobile Access, Persistent Memory & Multi-Agent Teams