BestHub
Sign in

All Articles

140553 articles · Page 71 of 7028
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 18, 2026 · Artificial Intelligence

UniRL: Tencent Hunyuan’s Open‑Source Framework Unifying Multimodal RL Training

UniRL is an open‑source, distributed reinforcement‑learning post‑training framework that consolidates fragmented pipelines for image, video, and language‑vision models, offering a unified rollout‑reward‑advantage‑train‑sync contract, extensive model support, built‑in algorithms, and multi‑modal reward components to lower engineering barriers in AIGC research.

LLMMultimodal RLReward Engineering
0 likes · 10 min read
UniRL: Tencent Hunyuan’s Open‑Source Framework Unifying Multimodal RL Training
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 18, 2026 · Artificial Intelligence

Can a 3B Model Rival Claude Opus 4.5? Benchmark Gaps or Aggressive Post‑Training?

VibeThinker‑3B, a 3‑billion‑parameter language model built on Qwen2.5‑Coder‑3B, achieves scores within the range of 671 B‑parameter models on benchmarks such as LiveCodeBench, AIME26, IMO‑AnswerBench and GPQA, thanks to a two‑stage SFT, multi‑domain reinforcement learning, offline self‑distillation and a claim‑reliability (CLR) evaluator that together push its reasoning ability to the frontier.

Large Language ModelsParameter EfficiencyVibeThinker-3B
0 likes · 9 min read
Can a 3B Model Rival Claude Opus 4.5? Benchmark Gaps or Aggressive Post‑Training?
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 18, 2026 · Artificial Intelligence

TNT: Dynamic Token Limits Slash Reward Hacking in Mixed Inference Models Below 10%

The paper introduces Thinking‑Based Non‑Thinking (TNT), a reinforcement‑learning approach that sets a per‑question dynamic token ceiling for non‑thinking mode using the answer length from thinking mode, cutting reward‑hacking incidence to under 10% while boosting accuracy and cutting token usage by nearly half across several math benchmarks.

ACL 2026NLPTNT
0 likes · 10 min read
TNT: Dynamic Token Limits Slash Reward Hacking in Mixed Inference Models Below 10%
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 18, 2026 · Industry Insights

Anthropic Reverses Claude Code Ban, Unlocks Third‑Party Apps – Community Boils Over

Anthropic announced it will no longer block programmatic calls to Claude Code, restoring quota usage for third‑party tools like Claude‑p, OpenClaw and Hermes, while simultaneously facing a US lawsuit accusing the company of false advertising its Max 5x and Max 20x subscription plans.

AI subscriptionAnthropicClaude
0 likes · 7 min read
Anthropic Reverses Claude Code Ban, Unlocks Third‑Party Apps – Community Boils Over
macrozheng
macrozheng
Jun 18, 2026 · Artificial Intelligence

Why AgentScope Java 2.0 Is Needed to Bridge the Demo‑to‑Production Gap in AI Agent Development

AgentScope Java 2.0, released in June 2026, adds native distributed deployment, multi‑tenant isolation, fine‑grained permission control, workspace‑driven state management, middleware extensibility, model fault‑tolerance and event‑stream APIs, turning demo‑only agents into production‑ready, observable, and secure AI services for enterprise Java environments.

AI agentsAgentScopeJava
0 likes · 28 min read
Why AgentScope Java 2.0 Is Needed to Bridge the Demo‑to‑Production Gap in AI Agent Development
Meituan Technology Team
Meituan Technology Team
Jun 18, 2026 · Artificial Intelligence

Meituan’s End‑to‑End AIGC Poster Generation: The Generate‑Edit‑Judge System

The article details Meituan’s AIGC poster‑creation pipeline, which tackles design‑resource scarcity, rapid turnaround, and quality control by integrating three open‑source models—PosterCraft for end‑to‑end generation, PosterOmni for unified multi‑task editing, and PosterReward for automated quality assessment—forming a self‑evolving generate‑edit‑judge loop.

AIGCMeituanMulti-Task Learning
0 likes · 27 min read
Meituan’s End‑to‑End AIGC Poster Generation: The Generate‑Edit‑Judge System
Linux Tech Enthusiast
Linux Tech Enthusiast
Jun 18, 2026 · Fundamentals

Beyond Windows and Linux: A Comprehensive Guide to Lesser‑Known Network Operating Systems

This article explains what a network operating system (NOS) is, traces its evolution from early batch systems to modern UNIX, Linux, NetWare and Windows Server variants, and details the core functions, architecture, and management capabilities that enable resource sharing, communication, security, and interoperability across networks.

LinuxNetWareNetwork Management
0 likes · 9 min read
Beyond Windows and Linux: A Comprehensive Guide to Lesser‑Known Network Operating Systems
ZhiKe AI
ZhiKe AI
Jun 18, 2026 · Industry Insights

How Reframing Information Can Flip Your Decisions: 3 Ways to Spot the Framing Effect

The article explains how the same data presented in a positive or negative frame can dramatically change people's choices, illustrates the classic Tversky‑Kahneman experiment, and offers three practical techniques to recognize and counteract the framing effect in everyday decisions.

Prospect Theorybehavioral economicscognitive bias
0 likes · 5 min read
How Reframing Information Can Flip Your Decisions: 3 Ways to Spot the Framing Effect
MaGe Linux Operations
MaGe Linux Operations
Jun 18, 2026 · Artificial Intelligence

How to Pick the Right Parallelism for 7B‑70B Models: DP, TP, PP, ZeRO & FSDP

This guide walks engineers through the memory, compute and bandwidth limits of training 7B‑70B models, compares data parallel (DP/DDP), tensor parallel (TP), pipeline parallel (PP), ZeRO stages and FSDP, shows how to calculate GPU memory, estimate communication overhead, configure each strategy, and avoid common pitfalls, enabling you to decide which parallelism to use on multi‑GPU or multi‑node clusters.

DeepSpeedFSDPZeRO
0 likes · 24 min read
How to Pick the Right Parallelism for 7B‑70B Models: DP, TP, PP, ZeRO & FSDP
Machine Heart
Machine Heart
Jun 18, 2026 · Industry Insights

How LiblibAI’s $300M ARR Fueled a Near‑$300M Funding Round Valuing It at Over $2B

The article analyzes LiblibAI’s record‑breaking financing, its $300 million ARR, rapid revenue growth, and three concrete product‑market‑fit signals—organic viral spread, premium user willingness to pay, and large‑scale B‑side adoption—that together explain its $2 billion valuation.

AI ApplicationsAI startup ecosystemLibTV
0 likes · 12 min read
How LiblibAI’s $300M ARR Fueled a Near‑$300M Funding Round Valuing It at Over $2B
Programmer XiaoFu
Programmer XiaoFu
Jun 18, 2026 · Cloud Native

Why Use Service Registration When Nginx Already Handles Load Balancing?

The article explains that Nginx’s static upstream configuration and passive health checks cannot keep up with dynamic microservice environments, while a service registry provides real‑time instance awareness, automatic failure detection, and metadata‑driven routing, making both tools complementary rather than interchangeable.

KubernetesNginxeureka
0 likes · 9 min read
Why Use Service Registration When Nginx Already Handles Load Balancing?
IoT Full-Stack Technology
IoT Full-Stack Technology
Jun 18, 2026 · Fundamentals

How to Make var [a, b] = {a:1, b:2} Work in JavaScript?

This article explains why the destructuring statement var [a, b] = {a:1, b:2} throws a TypeError, explores the iterator protocol and Symbol.iterator, and shows how redefining Object.prototype[Symbol.iterator] enables the object to be iterable so the destructuring succeeds.

DestructuringInterview QuestionIterator Protocol
0 likes · 6 min read
How to Make var [a, b] = {a:1, b:2} Work in JavaScript?
Architect's Tech Stack
Architect's Tech Stack
Jun 18, 2026 · Backend Development

Why Ditch Spring AI? 3 Java AI Frameworks That Supercharge Spring Boot

The article compares three Java‑centric AI frameworks—LangChain4j, Spring AI Alibaba, and LangGraph4j—showing how each integrates with Spring Boot to turn ordinary services into stateful, workflow‑driven AI agents, and offers guidance on picking the right tool for different project needs.

Agent frameworkJava AILangChain4j
0 likes · 11 min read
Why Ditch Spring AI? 3 Java AI Frameworks That Supercharge Spring Boot
Frontend AI Walk
Frontend AI Walk
Jun 18, 2026 · Artificial Intelligence

Master AI Coding with a 7‑Step Loop: From Task Cards to Release Checks

This tutorial shows how to replace one‑off prompts with a repeatable Loop Engineering workflow—task cards, clarification, task breakdown, TDD cycles, verification records, progress snapshots, and release checks—so AI‑generated code stays stable, testable, and easy to resume across development sessions.

AI codingLoop EngineeringTDD
0 likes · 17 min read
Master AI Coding with a 7‑Step Loop: From Task Cards to Release Checks
Java Tech Workshop
Java Tech Workshop
Jun 18, 2026 · Backend Development

Understanding @Lazy in Spring: Solving Circular Dependencies and Accelerating Startup

The article explains that Spring's @Lazy annotation has two distinct mechanisms—class‑level lazy bean creation and dependency‑level proxy placeholders—detailing how each works, when to apply them, common pitfalls such as transaction and aspect failures, and best‑practice scenarios for reducing startup time while safely handling circular dependencies.

@LazyBean ProxyLazy Initialization
0 likes · 12 min read
Understanding @Lazy in Spring: Solving Circular Dependencies and Accelerating Startup
Architecture & Thinking
Architecture & Thinking
Jun 18, 2026 · Backend Development

How to Scale a Flash‑Sale System from Zero to 1 Million QPS: A Step‑by‑Step Architecture Guide

This article dissects the evolution of a flash‑sale system from a simple monolithic controller to a cloud‑native, micro‑service architecture that can handle over one million requests per second, detailing traffic‑shaping, multi‑level caching, async processing, and inventory‑consistency techniques.

CachingFlash SaleHigh concurrency
0 likes · 18 min read
How to Scale a Flash‑Sale System from Zero to 1 Million QPS: A Step‑by‑Step Architecture Guide
Black & White Path
Black & White Path
Jun 18, 2026 · Information Security

ADPulse: Open‑Source Read‑Only AD Security Scanner with 35 Checks

ADPulse is an open‑source, read‑only Active Directory security scanner that runs 35 built‑in checks, provides a 100‑point risk score, supports Pass‑the‑Hash authentication, and generates console, JSON, or self‑contained HTML reports with a single command, making it suitable for quick AD health assessments and penetration‑test reconnaissance.

AD auditActive DirectoryPass-the-Hash
0 likes · 8 min read
ADPulse: Open‑Source Read‑Only AD Security Scanner with 35 Checks
Black & White Path
Black & White Path
Jun 18, 2026 · Information Security

Inside the AI‑Powered Hack: Full Claude & Codex Attack Log Exposed

OALABS recovered over 1,000 Claude and Codex session logs from a compromised server, revealing how the attackers duplicated AI agents, used them for reconnaissance, vulnerability exploitation, data theft, and even attempted cryptocurrency cracking across at least 14 companies, demonstrating that AI agents can dramatically lower the technical barrier for sophisticated cyber‑attacks.

AI securityClaudeCodex
0 likes · 49 min read
Inside the AI‑Powered Hack: Full Claude & Codex Attack Log Exposed