BestHub
Sign in
Tagged articles

large language models

1419 articles · Page 3 of 15
21CTO
21CTO
Jun 27, 2026 · Artificial Intelligence

Lilian Weng’s Deep Dive Overturns Three Years of Large‑Model Scaling Law Assumptions

In a ten‑thousand‑word analysis, former OpenAI safety VP Lilian Weng retraces the history of model scaling laws from Kaplan’s 2020 formulation, demonstrates how DeepMind’s Chinchilla overturns the original parameter‑to‑data ratio, uncovers two critical bugs in the Chinchilla paper, and warns that the impending 2026‑2028 data wall makes naïve scaling of parameters and compute unsustainable.

ai-trainingchinchilladata wall
0 likes · 10 min read
Lilian Weng’s Deep Dive Overturns Three Years of Large‑Model Scaling Law Assumptions
DataFunSummit
DataFunSummit
Jun 26, 2026 · Artificial Intelligence

Why Memory Is the Bottleneck for AI Agents and How MemOS Boosts Performance by Over 200%

The article explains how memory has become the decisive factor for AI agents, details the MemOS framework’s five‑layer architecture and three‑layer memory coordination, compares model‑driven and application‑driven approaches, and shows how MemOS‑powered cloud services achieved 100‑200% monthly growth, 45‑72% token savings, and up to 50% reduction in overall token consumption.

AI memoryAgent ArchitectureMemOS
0 likes · 18 min read
Why Memory Is the Bottleneck for AI Agents and How MemOS Boosts Performance by Over 200%
PaperAgent
PaperAgent
Jun 26, 2026 · Artificial Intelligence

Lilian Weng’s Deep Dive into Scaling Laws for Large‑Model Training

The article explains how scaling laws serve as a budget guide for training large language models, comparing Kaplan’s and Chinchilla’s findings, illustrating optimal parameter‑token trade‑offs, and highlighting the impact of data quality and duplication on model performance.

Compute BudgetData QualityKaplan
0 likes · 9 min read
Lilian Weng’s Deep Dive into Scaling Laws for Large‑Model Training
JD Retail Technology
JD Retail Technology
Jun 25, 2026 · Artificial Intelligence

JD Donates Oxygen xLLM Inference Engine to OpenAtom Foundation to Accelerate Domestic AI Infra

JD announced the donation of its self‑developed Oxygen xLLM large‑model inference engine to the OpenAtom Open Source Foundation, detailing its service‑engine decoupled architecture, performance breakthroughs, multi‑chip support, and early industrial validations that aim to foster a collaborative domestic AI infrastructure ecosystem.

AI inferenceDomestic AI ecosystemEngineering Intelligence
0 likes · 8 min read
JD Donates Oxygen xLLM Inference Engine to OpenAtom Foundation to Accelerate Domestic AI Infra
Shuge Unlimited
Shuge Unlimited
Jun 25, 2026 · Artificial Intelligence

Superpowers 6.0 vs spec‑kit: Is spec a Scaffold or the Sole Truth?

An in‑depth analysis of the Superpowers 6.0 rewrite and spec‑kit reveals how the two star‑studded AI‑coding frameworks diverge on the role of specifications—Superpowers treats specs as design scaffolds for execution, while spec‑kit elevates specs to executable artifacts—and evaluates their mechanisms, disciplinary models, compatibility, and relevance as large language models grow more capable.

AI programmingSpec-Driven DevelopmentSpec-Kit
0 likes · 22 min read
Superpowers 6.0 vs spec‑kit: Is spec a Scaffold or the Sole Truth?
DataFunSummit
DataFunSummit
Jun 24, 2026 · Artificial Intelligence

Three Forms of Large Model Memory – Parameter, Token, and Latent – Why Top Companies Are All‑In

A new paper unifies AI memory research with a three‑dimensional framework (Forms, Functions, Dynamics), classifies memory as parameter‑level, token‑level, or latent‑space, and evaluates real‑world implementations from OpenAI, Google, Amazon and dozens of open‑source frameworks, highlighting trade‑offs such as retrieval quality, catastrophic forgetting and forgetting mechanisms.

AI memoryAgent Architectureframework comparison
0 likes · 19 min read
Three Forms of Large Model Memory – Parameter, Token, and Latent – Why Top Companies Are All‑In
Machine Heart
Machine Heart
Jun 24, 2026 · Industry Insights

Karpathy Backs Engram: AI Memory Startup Aiming for Persistent Enterprise Knowledge

Engram, a newly announced AI memory startup backed by investors such as General Catalyst, Kleiner Perkins, Sequoia and advisors including Andrej Karpathy, aims to move beyond temporary context retrieval by building a continuous‑learning memory layer that lets models absorb and recall enterprise‑specific knowledge, contrasting with typical RAG or long‑context methods.

AI memoryContinuous LearningEnterprise AI
0 likes · 6 min read
Karpathy Backs Engram: AI Memory Startup Aiming for Persistent Enterprise Knowledge
DataFunTalk
DataFunTalk
Jun 23, 2026 · Artificial Intelligence

Can a Fable‑Level AI Model That Evades Export Controls Beat Claude Mythos?

Amid the sudden shutdown of Anthropic's Claude Fable 5, Sakana AI unveils Fugu—an orchestration‑based, Fable‑level model that sidesteps export restrictions, matches or exceeds Fable 5 and Mythos on engineering, scientific, and reasoning benchmarks, and demonstrates a new trend toward model scheduling over sheer scale.

AI OrchestrationBenchmarkingSakana AI
0 likes · 8 min read
Can a Fable‑Level AI Model That Evades Export Controls Beat Claude Mythos?
DeepHub IMBA
DeepHub IMBA
Jun 23, 2026 · Artificial Intelligence

Parallel Training of 100B‑Parameter Models: Intra‑Node Tensor Parallelism and Inter‑Node Data Parallelism

Training 100‑billion‑parameter Transformers is limited by GPU memory rather than compute, requiring a mix of tensor parallelism within nodes and data parallelism across nodes, along with pipeline parallelism, gradient accumulation, and careful framework choices to balance memory, bandwidth, and compute overheads.

Distributed TrainingGPU memorydata parallelism
0 likes · 14 min read
Parallel Training of 100B‑Parameter Models: Intra‑Node Tensor Parallelism and Inter‑Node Data Parallelism
DataFunSummit
DataFunSummit
Jun 23, 2026 · Artificial Intelligence

Financial Large Language Models: Architecture Shifts, Engineering Lessons, and Cutting‑Edge Agent Strategies

The article analyzes how strict compliance, data‑security, and rigorous business requirements reshape financial large‑model deployments, detailing a PageIndex‑based retrieval architecture, engineering pitfalls such as rule explosion and prompt bloat, model‑selection trade‑offs, and forward‑looking agent‑centric designs.

Agentic AIRetrieval Augmentationfinancial AI
0 likes · 11 min read
Financial Large Language Models: Architecture Shifts, Engineering Lessons, and Cutting‑Edge Agent Strategies
AI Architecture Hub
AI Architecture Hub
Jun 23, 2026 · Artificial Intelligence

Top AI Papers This Week (June 14‑21): SpatialClaw, SkillWeaver, PreAct, and More

This article reviews seven recent AI research papers, detailing how SpatialClaw enables code‑based spatial reasoning for vision‑language models, SkillWeaver introduces compositional skill routing, PreAct compiles agent actions into reusable state‑machines, and other works advance world‑model inference, self‑designing RL environments, collective skill‑tree search, and process‑aligned reinforcement learning for diffusion LLMs.

Diffusion Modelsagent reasoninglarge language models
0 likes · 15 min read
Top AI Papers This Week (June 14‑21): SpatialClaw, SkillWeaver, PreAct, and More
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 22, 2026 · Artificial Intelligence

Why Large Language Models Need Not Run CoT on Every Question: Tencent Hunyuan’s On‑Demand CoT Trigger

The paper analyzes the efficiency and reward‑signal shortcomings of conventional generative reward models (GRM) and presents the E‑GRM framework, which uses model‑internal uncertainty to dynamically trigger chain‑of‑thought reasoning, employs a consensus‑based routing decision and a mixed‑loss discriminative scorer, achieving significant speed‑up and accuracy gains on benchmarks such as MATH, RM‑Bench and RewardBench.

Generative Reward ModelingMixed Loss ScorerModel Uncertainty
0 likes · 15 min read
Why Large Language Models Need Not Run CoT on Every Question: Tencent Hunyuan’s On‑Demand CoT Trigger
DaTaobao Tech
DaTaobao Tech
Jun 22, 2026 · Artificial Intelligence

Breaking AI Coding Bottlenecks: How Specflow Agent Separates Deep Analysis from Code Execution

The article dissects why AI‑generated code often fails to boost productivity—highlighting attention‑mechanism limits, context collapse, and mismatched developer workflows—and proposes a Specflow Agent that isolates deep requirement analysis from coding to dramatically cut manual intervention.

AI codingAgent ArchitectureContext Management
0 likes · 40 min read
Breaking AI Coding Bottlenecks: How Specflow Agent Separates Deep Analysis from Code Execution
Machine Heart
Machine Heart
Jun 22, 2026 · Artificial Intelligence

Building the First Real‑World CLI Workflow Benchmark from 80K Human Terminal Recordings

TerminalWorld leverages over 80,000 developer‑recorded terminal sessions to automatically generate 1,530 verified CLI tasks across 18 workflow categories, and its evaluation of leading LLMs and agent frameworks reveals modest success rates, capability gaps, and the shortcomings of expert‑crafted benchmarks.

AI Agentsasciinemaevaluation
0 likes · 13 min read
Building the First Real‑World CLI Workflow Benchmark from 80K Human Terminal Recordings
Machine Heart
Machine Heart
Jun 21, 2026 · Artificial Intelligence

Why the Once‑Rejected PPO Algorithm Became a Pillar of Modern LLM Training

The article recounts how Proximal Policy Optimization, initially dismissed by NeurIPS 2017 for limited novelty, later became a cornerstone of RLHF and large‑language‑model training, illustrating how academic evaluation can miss long‑term impact, with parallels to other once‑rejected breakthroughs such as LSTM, SIFT and Dropout.

Algorithm RejectionNeurIPSPPO
0 likes · 5 min read
Why the Once‑Rejected PPO Algorithm Became a Pillar of Modern LLM Training
PaperAgent
PaperAgent
Jun 21, 2026 · Artificial Intelligence

What Drives AI Model Evolution? OpenAI’s New Findings on Beneficial Traits

OpenAI’s latest study shows that injecting just 5% of beneficial‑trait data into reinforcement‑learning training yields over 80% improvement across more than 50 alignment evaluations, revealing that a few underlying personality traits drive cross‑domain alignment and persist under adversarial pressure.

AI Alignmentadversarial robustnessbeneficial traits
0 likes · 12 min read
What Drives AI Model Evolution? OpenAI’s New Findings on Beneficial Traits
Machine Heart
Machine Heart
Jun 21, 2026 · Artificial Intelligence

Is GRPO Obsolete? Why GLM‑5.2 Dropped It and What It Means for RL

GLM‑5.2 replaces the Group Relative Policy Optimization (GRPO) algorithm with a critic‑based PPO approach for long‑horizon tasks, arguing that GRPO’s group comparison breaks down on variable‑length trajectories, a shift that has sparked vigorous debate across the reinforcement‑learning community.

DeepSeekGLM-5.2GRPO
0 likes · 10 min read
Is GRPO Obsolete? Why GLM‑5.2 Dropped It and What It Means for RL
Data Party THU
Data Party THU
Jun 20, 2026 · Artificial Intelligence

Can Large Language Models Fall into a Silent Spiral? Uncovering AI Opinion Monopoly and Governance Solutions

This article examines how large language models can autonomously generate a digital “silence spiral,” suppressing minority viewpoints and creating opinion monopolies, outlines empirical evidence from recent ACL and arXiv studies, and proposes a three‑dimensional governance framework spanning technical, regulatory, and research interventions.

Governance FrameworkRAGinformation ecology
0 likes · 17 min read
Can Large Language Models Fall into a Silent Spiral? Uncovering AI Opinion Monopoly and Governance Solutions
MaGe Linux Operations
MaGe Linux Operations
Jun 20, 2026 · Artificial Intelligence

LoRA vs QLoRA vs Full Fine‑Tuning: Which Method Wins for Large‑Model Adaptation?

This article provides a practical, data‑driven comparison of Full Fine‑Tuning, LoRA, and QLoRA for adapting 7B‑70B open‑source LLMs, detailing memory requirements, training speed, cost, performance trade‑offs, step‑by‑step workflows, code examples, evaluation metrics, common pitfalls, and optimization tips to help engineers choose the most suitable fine‑tuning approach for their data and budget.

Full Fine-tuningGPU memoryLoRA
0 likes · 24 min read
LoRA vs QLoRA vs Full Fine‑Tuning: Which Method Wins for Large‑Model Adaptation?
Digital Planet
Digital Planet
Jun 20, 2026 · Industry Insights

AI Landscape June 2024: GLM‑5.2 Release, SpaceX’s $600B Cursor Deal, and Rising Regulatory Scrutiny

In June 2024 the AI sector saw major technical breakthroughs and regulatory actions, including the open‑source launch of Zhipu AI's GLM‑5.2, SpaceX's $600 billion acquisition of Cursor, US state investigations of OpenAI, Salesforce's Fin purchase, Cloudflare's AI‑agent traffic surge, Meta's AI Mode search, and new policies supporting large models in China.

AIGLM-5.2SpaceX
0 likes · 10 min read
AI Landscape June 2024: GLM‑5.2 Release, SpaceX’s $600B Cursor Deal, and Rising Regulatory Scrutiny
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 19, 2026 · Artificial Intelligence

Can Post‑Training Close the Gap to Mythos‑Level AI? Musk Says 9 Months, Tang Says Faster

The article analyzes whether post‑training on GLM‑5.1/5.2 can bridge the gap to the banned Mythos model, citing Musk’s nine‑month claim, Tang’s rebuttal, Mind Lab’s benchmark gains, architectural adaptations, and the high barriers that make post‑training a critical yet scarce capability in China.

AI developmentGLM-5.2IndexCache
0 likes · 9 min read
Can Post‑Training Close the Gap to Mythos‑Level AI? Musk Says 9 Months, Tang Says Faster
Architect
Architect
Jun 19, 2026 · Artificial Intelligence

From Harness to Environment: The Next Engineering Layer for LLM Agents

The article argues that while Harness engineering still controls how agents run, the emerging focus on Environment engineering determines whether agents receive reliable, verifiable feedback, shaping their long‑term learning and safety in real‑world tasks.

AI SystemsAgent EngineeringEnvironment Engineering
0 likes · 21 min read
From Harness to Environment: The Next Engineering Layer for LLM Agents
Machine Heart
Machine Heart
Jun 19, 2026 · Artificial Intelligence

Who Is Quietly Building China’s Mythos‑Level AI? Musk Says 9 Months, Tang Says It’s Not That Fast

The article analyzes China’s race to achieve Mythos‑level intelligence, contrasting Musk’s nine‑month claim with Tang’s skepticism, and highlights Mind Lab’s unique post‑training work on GLM‑5.1/5.2 that has already delivered significant benchmark gains, while outlining the technical hurdles and timeline uncertainties.

AI development in ChinaGLM-5.2Mind Lab
0 likes · 8 min read
Who Is Quietly Building China’s Mythos‑Level AI? Musk Says 9 Months, Tang Says It’s Not That Fast
DataFunSummit
DataFunSummit
Jun 19, 2026 · Artificial Intelligence

Why Memory Bottlenecks AI Agents: Inside MemOS Architecture and 200% Cloud Usage Surge

The article analyzes how memory has become the critical bottleneck for AI agents, compares model‑driven and application‑driven memory approaches, details the five‑layer MemOS framework, reports cloud service call growth of over 200% and token‑cost reductions of up to 72%, and shows real‑world enterprise deployments such as OpenClaw and ClawForce.

AI AgentsAI memoryMemOS
0 likes · 16 min read
Why Memory Bottlenecks AI Agents: Inside MemOS Architecture and 200% Cloud Usage Surge
ThinkingAgent
ThinkingAgent
Jun 19, 2026 · Industry Insights

The AI Trifecta: How Tokens, Power, and Data Define AI’s Limits

The article breaks down the three core factors—tokens, electricity, and data—that together determine AI model cost, speed, capability, and industry competition, illustrating the trade‑offs with concrete numbers, examples, and future outlooks.

AIAI industryEnergy Consumption
0 likes · 10 min read
The AI Trifecta: How Tokens, Power, and Data Define AI’s Limits
DataFunTalk
DataFunTalk
Jun 19, 2026 · Artificial Intelligence

Best Model Combo Guide: GLM 5.2, Kimi 2.7, DeepSeek V4 & MiniMax M3

The author compares four Chinese large‑language models—GLM 5.2, Kimi 2.7, DeepSeek V4 and MiniMax M3—detailing their strengths, pricing, and ideal use‑cases for writing, coding, multimodal processing and high‑throughput batch tasks, and shares personal trust insights.

AI Model ComparisonDeepSeek V4GLM-5.2
0 likes · 10 min read
Best Model Combo Guide: GLM 5.2, Kimi 2.7, DeepSeek V4 & MiniMax M3
Design Hub
Design Hub
Jun 19, 2026 · Artificial Intelligence

AI Tools Gain ‘Hands and Feet’: 7 Must‑Watch Updates This Week

This week’s AI roundup shows a shift from pure content generation to workflow execution, with Cursor’s natural‑language automation, Design Arena’s open‑weight GLM‑5.2 topping leaderboards, Figma’s web‑search‑enabled design agent, OpenAI Codex’s record‑and‑replay skills, Claude’s shareable artifacts and brand‑aware design, and Unreal Engine 5.8’s LLM‑driven pipelines, highlighting new capabilities, risks, and management challenges for developers and designers.

AI toolsClaudeCursor
0 likes · 25 min read
AI Tools Gain ‘Hands and Feet’: 7 Must‑Watch Updates This Week
Machine Heart
Machine Heart
Jun 19, 2026 · Artificial Intelligence

LLMs Finally Derive Formulas: FunctionEvolve Boosts LLM‑SRBench Accuracy 3.6× and Scores Perfect on AI‑Feynman

The FunctionEvolve framework represents formulas as abstract syntax trees, letting large language models guide symbolic regression; this yields a 3.6‑fold improvement on the LLM‑SRBench benchmark (55.8% SA@1) and a perfect 120/120 score on AI‑Feynman, with detailed component ablations confirming the value of structure‑aware generation, selection, mutation, and optimization.

AI‑FeynmanASTFunctionEvolve
0 likes · 15 min read
LLMs Finally Derive Formulas: FunctionEvolve Boosts LLM‑SRBench Accuracy 3.6× and Scores Perfect on AI‑Feynman
dbaplus Community
dbaplus Community
Jun 19, 2026 · Industry Insights

Why Software Engineering Has Never Been Truly Engineered – How Large AI Models May Finally Deliver Real Engineering

The article argues that software engineering has spent the past fifty years merely managing human uncertainty rather than true engineering, and that large language models now make it possible to replace low‑level cognition with energy‑driven intelligence, demanding a shift to an AI‑centered paradigm, closed‑loop automation, and a new focus on scenario‑driven knowledge distillation.

AIEngineering Processautomation
0 likes · 50 min read
Why Software Engineering Has Never Been Truly Engineered – How Large AI Models May Finally Deliver Real Engineering
Linyb Geek Road
Linyb Geek Road
Jun 19, 2026 · Artificial Intelligence

Agent Skills Review: How New AI Skills Are Redefining Large‑Model Operating Systems

The article surveys the rapid emergence of Agent Skills, outlines a six‑layer framework that defines their ontology, representation, lifecycle, runtime integration, governance, and applications, highlights severe security vulnerabilities revealed in large‑scale studies, and discusses the open research challenges ahead.

AI Agent ApplicationsAI safetyAgent Governance
0 likes · 16 min read
Agent Skills Review: How New AI Skills Are Redefining Large‑Model Operating Systems
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 18, 2026 · Artificial Intelligence

From Imitation to Optimization: Recent Advances in On-Policy Distillation

This article surveys the latest research on On-Policy Distillation for large language models, covering methods that improve training stability, self‑distillation frameworks, and detailed analyses of when and why OPD succeeds or fails, with concrete experimental results and practical insights.

Entropy-AwareOn-Policy DistillationSelf-Distillation
0 likes · 19 min read
From Imitation to Optimization: Recent Advances in On-Policy Distillation
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 18, 2026 · Artificial Intelligence

Why LLMs Miss Simple Addition: Geometric Mechanism Behind Arithmetic Errors

A recent ICML 2026 paper from Nanjing University reveals that large language models encode correct arithmetic information in structured geometric manifolds, yet errors arise from noisy quantization at decision boundaries, and proposes probing, Iso‑Raw‑Sum Trajectory, and a dual‑stream consistency check to diagnose and correct these mistakes.

Arithmetic ErrorsDual-Stream ConsistencyIso-Raw-Sum Trajectory
0 likes · 11 min read
Why LLMs Miss Simple Addition: Geometric Mechanism Behind Arithmetic Errors
Data Party THU
Data Party THU
Jun 18, 2026 · Artificial Intelligence

Why Large Language Models Are Short‑Sighted and How Next‑ToBE Unlocks Anticipatory Reasoning

The article examines the short‑sighted nature of current next‑token prediction in LLMs, presents the Next‑ToBE (Next Token‑Bag Exploitation) method that reshapes the training objective to expose latent future‑token awareness, and shows through extensive experiments that this approach improves anticipatory reasoning and downstream task performance.

Anticipatory ReasoningFuture Token PredictionLLM evaluation
0 likes · 12 min read
Why Large Language Models Are Short‑Sighted and How Next‑ToBE Unlocks Anticipatory Reasoning
Kuaishou Tech
Kuaishou Tech
Jun 18, 2026 · Artificial Intelligence

Kuaishou Tech Team Highlights Multiple ICML 2026 Papers Across AI Domains

The Kuaishou technology team reports that several of its papers were accepted at the prestigious ICML 2026 conference—including a spotlight paper on metaphor video understanding, works on causal discovery for irregular time series, image super‑resolution, large‑scale notification dispatch, full‑order ranking, phase‑aware MoE for RL, end‑to‑end e‑commerce search, spatial‑reasoning rewards, a unified SWE benchmark, video temporal grounding, and interpretable transformers—while also inviting attendees to visit their booth B101 in Seoul.

Agentic AIICML 2026Kuaishou
0 likes · 18 min read
Kuaishou Tech Team Highlights Multiple ICML 2026 Papers Across AI Domains
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.

Post-TrainingVibeThinker-3Bbenchmark performance
0 likes · 9 min read
Can a 3B Model Rival Claude Opus 4.5? Benchmark Gaps or Aggressive Post‑Training?
TechVision Expert Circle
TechVision Expert Circle
Jun 17, 2026 · Information Security

AI Agents Ignite an Automated War: How Hackers and Defenders Are Racing with Machines

In early 2026, multiple breach investigations revealed AI agents capable of autonomous decision‑making that complete reconnaissance to lateral movement in minutes, while traditional SOCs still need hours, prompting both attackers and defenders to adopt fast, AI‑driven automation for enterprise security.

AI AgentsAI-driven defenseSOC
0 likes · 13 min read
AI Agents Ignite an Automated War: How Hackers and Defenders Are Racing with Machines
Baidu Intelligent Cloud Tech Hub
Baidu Intelligent Cloud Tech Hub
Jun 17, 2026 · Artificial Intelligence

The Multimodal Model Battlefield Is Going Rogue – LoongForge’s ‘Dark Arts’ Framework

Facing mounting challenges of heterogeneous models, data, and hardware in multimodal training, Baidu’s open‑source LoongForge framework unifies LLM, VLM, VLA and diffusion workloads, delivering 1.15‑2.31× speedups and over 5× gains for DSA models while scaling linearly across thousands of GPUs and Kunlun XPU cards.

Kunlun XPULoongForgeMultimodal Training
0 likes · 8 min read
The Multimodal Model Battlefield Is Going Rogue – LoongForge’s ‘Dark Arts’ Framework
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 17, 2026 · Artificial Intelligence

Is More Chain‑of‑Thought Always Better? Introducing E‑GRM for On‑Demand LLM Reasoning

The article critically examines the assumption that longer chain‑of‑thought reasoning always improves large language model performance, presents the E‑GRM framework that dynamically decides when to invoke full CoT based on model‑internal uncertainty, and validates its efficiency and accuracy gains through extensive experiments and ablations.

Generative Reward Modelablation studychain-of-thought
0 likes · 16 min read
Is More Chain‑of‑Thought Always Better? Introducing E‑GRM for On‑Demand LLM Reasoning
Kuaishou Tech
Kuaishou Tech
Jun 16, 2026 · Artificial Intelligence

Generative Recommendation at Kuaishou: Systematic Evolution and Salon Highlights

The article recaps Kuaishou's June 13 technical salon, detailing the systematic evolution of generative recommendation—from scaling models to reasoning enhancements—through core projects OneReason, Pool‑Rec, OneSearch V2, and GR4AD, and announces the industry‑wide LLM‑Rec challenge for students.

GR4ADGenerative RecommendationKuaishou
0 likes · 10 min read
Generative Recommendation at Kuaishou: Systematic Evolution and Salon Highlights
Machine Heart
Machine Heart
Jun 16, 2026 · Artificial Intelligence

From AI Scientists to Social Science: Tsinghua Unveils AgentSociety² Silicon‑Based Social Lab

AgentSociety² combines large‑model agents, a scalable simulation engine, and a unified research workflow to turn AI‑driven social simulations into an executable laboratory for computational social science, enabling hypothesis testing, intervention design, and reproducible experiments across micro, meso, and macro scales.

AI Social ScientistsAgentSocietyExecutable Social Science
0 likes · 12 min read
From AI Scientists to Social Science: Tsinghua Unveils AgentSociety² Silicon‑Based Social Lab
ZhiKe AI
ZhiKe AI
Jun 15, 2026 · Artificial Intelligence

Why AI Hallucinates and How Retrieval-Augmented Generation Gives It a Research Assistant

Retrieval-Augmented Generation (RAG) equips large language models with a three‑step "retrieve‑augment‑generate" workflow, turning closed‑book AI into an open‑book system that lowers hallucinations, updates knowledge in real time, and improves answer accuracy, though it still faces retrieval errors and reasoning limits.

AI hallucinationEnterprise AIKnowledge Retrieval
0 likes · 5 min read
Why AI Hallucinates and How Retrieval-Augmented Generation Gives It a Research Assistant
ThinkingAgent
ThinkingAgent
Jun 14, 2026 · Artificial Intelligence

Ontology: The Overlooked Knowledge Infrastructure Driving AI Understanding

The article explains how ontology—a 2,500‑year‑old philosophical concept—provides the structured knowledge backbone that large language models lack, detailing its definition, differences from databases and knowledge graphs, its role in reducing hallucinations, defining knowledge boundaries, enabling reasoning, and four practical AI application scenarios.

AI AgentsKnowledge GraphRAG
0 likes · 17 min read
Ontology: The Overlooked Knowledge Infrastructure Driving AI Understanding
JavaGuide
JavaGuide
Jun 13, 2026 · Industry Insights

Claude Fable 5 Banned While GLM‑5.2 Opens to All Users

The article analyzes the sudden suspension of Anthropic's Claude Fable 5 due to U.S. export controls, contrasts it with Zhipu's rapid full release of GLM‑5.2 for the GLM Coding Plan, and discusses the models' capabilities, competitive landscape, and the surprising market valuations driving developer choices.

AI model competitionClaudeGLM-5.2
0 likes · 6 min read
Claude Fable 5 Banned While GLM‑5.2 Opens to All Users
DataFunSummit
DataFunSummit
Jun 13, 2026 · Artificial Intelligence

Beyond General LLMs: Efficient Adaptation and Data Value Mining for Finance

The article details a systematic practice—starting from the “iceberg” challenges of finance, through data and knowledge engineering, reverse knowledge extraction with REER, multi‑dimensional synthetic data generation, prompt engineering (APO), cost‑aware fine‑tuning, inference acceleration, and emotion‑value evaluation—culminating in actionable guidelines for deploying large models in banking scenarios.

Inference AccelerationKnowledge Engineeringemotion evaluation
0 likes · 14 min read
Beyond General LLMs: Efficient Adaptation and Data Value Mining for Finance
DataFunSummit
DataFunSummit
Jun 13, 2026 · Artificial Intelligence

Ontology: The Semantic Operating System Powering Large‑Model AI

The article argues that in the era of large language models the missing layer for enterprises is not more model capability but a unified, computable, and evolvable semantic structure—an ontology that acts as a semantic operating system, and it examines why this is needed, how it can be built, and the organizational and open‑source challenges involved.

Enterprise AIKnowledge GraphOpenKG
0 likes · 17 min read
Ontology: The Semantic Operating System Powering Large‑Model AI
Top Architect
Top Architect
Jun 13, 2026 · Artificial Intelligence

What Is an Inference Large Language Model? A Visual Guide

The article explains inference‑type large language models, how they differ from traditional models by breaking questions into reasoning steps, the shift from training‑time to test‑time compute, scaling‑law insights, validation techniques, proposal‑distribution tricks, and the detailed training pipeline of DeepSeek‑R1, while also discussing failed experiments and future directions.

DeepSeek-R1inference modelslarge language models
0 likes · 20 min read
What Is an Inference Large Language Model? A Visual Guide
Data Party THU
Data Party THU
Jun 13, 2026 · Artificial Intelligence

How Subconscious Learning in Large Language Models Can Transfer Behavioral Biases

A recent Nature paper reveals that large language models can inherit hidden behavioral preferences from teacher models through subconscious learning, even when training data lack explicit semantic signals, leading to significant misalignment risks demonstrated across numeric, code, and chain‑of‑thought experiments.

AI safetyemergent behaviorknowledge distillation
0 likes · 9 min read
How Subconscious Learning in Large Language Models Can Transfer Behavioral Biases
DataFunSummit
DataFunSummit
Jun 12, 2026 · Artificial Intelligence

How Agentic Architectures Power Next‑Gen Recommendation and Search Systems

This article analyzes cutting‑edge AI search and recommendation technologies, covering Alibaba Cloud's Agentic RAG architecture, Huawei Noah's LLM‑enhanced recommender evolution, and Baidu's generative ranking model GRAB, each with detailed designs, performance metrics, and real‑world deployment insights.

AI AgentsGenerative RankingMultimodal Retrieval
0 likes · 6 min read
How Agentic Architectures Power Next‑Gen Recommendation and Search Systems
Machine Heart
Machine Heart
Jun 12, 2026 · Artificial Intelligence

Breaking Fable 5’s Safety in Under 5 Seconds with a Single Dialogue

A multinational research team demonstrated that the new safety classifier of Anthropic’s Fable 5 can be bypassed in less than five seconds with just one conversation, revealing an internal safety collapse (ISC) flaw that lets agents generate harmful content despite external defenses.

AI safetyAgent SecurityInternal Safety Collapse
0 likes · 11 min read
Breaking Fable 5’s Safety in Under 5 Seconds with a Single Dialogue
DataFunTalk
DataFunTalk
Jun 12, 2026 · Artificial Intelligence

How Ontology + Large Models Enable Knora to Tackle Hallucinations and Execution Gaps in Enterprise AI

The article explains how Knora 4.0 combines ontology with large‑model AI to move enterprise applications from isolated chat bots to autonomous, end‑to‑end systems, addressing six major challenges such as hallucinations, unstable outputs, weak planning, poor responsiveness, data integration difficulty, and long cold‑start cycles, and demonstrates the approach with real LED‑line use cases, architectural details, and a roadmap for future autonomous agents.

AI platformEnterprise AIKnowledge Graph
0 likes · 17 min read
How Ontology + Large Models Enable Knora to Tackle Hallucinations and Execution Gaps in Enterprise AI
Machine Heart
Machine Heart
Jun 12, 2026 · Artificial Intelligence

Can Transformers Solve Any Computable Problem? RUC Study Shows Context Management Sets the Upper Bound

A recent ICML 2026 position paper clarifies that the computational power of a fixed Transformer model is limited by its context‑management strategy, distinguishing fixed‑system and scaling‑family settings and showing how five concrete management approaches span from constant‑space to full Turing‑completeness.

Computational theoryContext ManagementTransformer
0 likes · 16 min read
Can Transformers Solve Any Computable Problem? RUC Study Shows Context Management Sets the Upper Bound
AI Large-Model Wave and Transformation Guide
AI Large-Model Wave and Transformation Guide
Jun 11, 2026 · Artificial Intelligence

How a 4B Ontology Model Beats Trillion-Parameter LLMs with 89.47% Enterprise Inference Accuracy

A 4‑billion‑parameter Large Ontology Model (LOM) outperforms the trillion‑parameter DeepSeek‑V3.2 on complex enterprise reasoning tasks, achieving 89.47% accuracy by embedding a dual‑layer ontology into the model through a three‑stage Build‑Align‑Reason framework, dramatically lowering deployment cost and latency.

Enterprise AIKnowledge GraphLOM
0 likes · 12 min read
How a 4B Ontology Model Beats Trillion-Parameter LLMs with 89.47% Enterprise Inference Accuracy
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 11, 2026 · Artificial Intelligence

Anthropic Announces Recursive Self‑Improvement Era: How LLMs Achieve Self‑Evolution

The article surveys the emerging LLM self‑improvement paradigm, citing Anthropic's internal data that 80% of its code is now generated by Claude and engineers are eight times more productive, and detailing the SUNY Stony Brook paper that defines a closed‑loop system of data acquisition, selection, model optimization, inference refinement and autonomous evaluation, while outlining its challenges, applications, and future research directions.

AI safetyAutonomous EvaluationLLM
0 likes · 14 min read
Anthropic Announces Recursive Self‑Improvement Era: How LLMs Achieve Self‑Evolution
DataFunSummit
DataFunSummit
Jun 11, 2026 · Artificial Intelligence

Designing Next‑Gen Recommendation and Search with Agentic Architectures

This article reviews cutting‑edge AI search and recommendation techniques—including Alibaba Cloud's Agentic RAG, Huawei Noah's LLM‑enhanced recommender, and Baidu's generative ranking model GRAB—detailing their architectures, multi‑modal retrieval strategies, performance gains, and practical deployment insights.

AI SearchAgentic RAGBaidu GRAB
0 likes · 6 min read
Designing Next‑Gen Recommendation and Search with Agentic Architectures
Machine Heart
Machine Heart
Jun 11, 2026 · Artificial Intelligence

Can Agents Search Without a Vector Database? A Simple Grep Is Enough

The paper introduces Direct Corpus Interaction (DCI), letting LLM agents bypass vector indexes and use command‑line tools like grep to directly search raw text, achieving higher accuracy and lower cost on complex multi‑hop QA and retrieval benchmarks.

Agentic SearchCost EfficiencyDirect Corpus Interaction
0 likes · 12 min read
Can Agents Search Without a Vector Database? A Simple Grep Is Enough
Machine Heart
Machine Heart
Jun 11, 2026 · Artificial Intelligence

Do Large Language Models Truly Grasp Phrase Semantics? Findings from ACL 2026 Oral

The SemanticQA benchmark breaks phrase‑level semantic understanding into extraction, categorization and interpretation tasks, evaluates over ten models—including GPT‑5, Claude Sonnet and Gemini 2.5 Pro—and reveals systematic gaps, performance drops with finer categories, and error propagation in multi‑step pipelines.

SemanticQAevaluation benchmarklarge language models
0 likes · 18 min read
Do Large Language Models Truly Grasp Phrase Semantics? Findings from ACL 2026 Oral
AI Architecture Hub
AI Architecture Hub
Jun 11, 2026 · Artificial Intelligence

Why Every AI Engineer Must Master Agent Loops by 2026

The article explains how AI engineers should shift from single‑prompt interactions to designing autonomous agent loops, outlines the token‑cost challenges of open‑ended cycles, presents closed‑loop and multi‑agent architectures, and details six essential components and practical examples for building cost‑effective, scalable automation.

AI AgentsLoop Engineeringautomation
0 likes · 18 min read
Why Every AI Engineer Must Master Agent Loops by 2026
SuanNi
SuanNi
Jun 10, 2026 · Artificial Intelligence

Anthropic’s Claude Fable 5 and Mythos 5: 50 M‑Line Code Migration in One Day

Anthropic released two new Claude models—Fable 5, open to all users with a safety classifier, and Mythos 5, a restricted, high‑security version—both achieving record‑breaking performance on software‑engineering, research, vision, and long‑context tasks, while offering a pricing model of $10 per M input tokens and $50 per M output tokens.

AI benchmarksClaude Fable 5Mythos 5
0 likes · 11 min read
Anthropic’s Claude Fable 5 and Mythos 5: 50 M‑Line Code Migration in One Day
Design Hub
Design Hub
Jun 10, 2026 · Artificial Intelligence

Claude Fable 5 & Mythos 5: Anthropic’s New High‑Capability AI Distribution System Explained

Anthropic’s June 9 launch of Claude Fable 5 and Claude Mythos 5 introduces a Mythos‑class model split into a public‑ready “Fable” version and a trusted‑partner “Mythos” version, highlighting stronger coding, long‑task, vision, and research abilities, a safety‑first distribution framework, and the shifting focus from raw model power to controlled, low‑friction AI deployment.

AI Product StrategyAI safetyAnthropic
0 likes · 30 min read
Claude Fable 5 & Mythos 5: Anthropic’s New High‑Capability AI Distribution System Explained
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 10, 2026 · Artificial Intelligence

Anthropic Unleashes Mythic‑Level Claude 5 and Claude Fable 5 – A Massive Performance Leap

Anthropic has just released Claude Fable 5 and Claude Mythos 5, two new LLMs that outperform all prior models on a wide range of benchmarks—from coding and agent tasks to visual reasoning and protein design—while introducing a safety classifier in Fable 5, offering comparable pricing to Opus 4.8, and showcasing dramatic real‑world demos such as autonomous Factorio building, 3D CAD generation, and a full Pokémon playthrough.

AI benchmarksAI safetyAnthropic
0 likes · 11 min read
Anthropic Unleashes Mythic‑Level Claude 5 and Claude Fable 5 – A Massive Performance Leap
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 10, 2026 · Artificial Intelligence

Can Code Really Boost Large‑Model Reasoning? A Re‑Examination of New Experiments

The paper “What Really Improves Mathematical Reasoning: Structured Reasoning Signals Beyond Pure Code” (ICML 2026) shows that while pure code data clearly enhances programming ability, it does not consistently improve complex mathematical reasoning and can even compete with math data, whereas structured reasoning signals—cognitive scaffolds—significantly lift difficult math benchmarks without harming coding performance.

code datalarge language modelsmathematical reasoning
0 likes · 13 min read
Can Code Really Boost Large‑Model Reasoning? A Re‑Examination of New Experiments
Machine Heart
Machine Heart
Jun 9, 2026 · Artificial Intelligence

Can a $10 Million Inference Budget Uncover AI’s Real Upper Limit?

The article argues that as large language models grow more capable, single‑score benchmarks no longer capture true performance; instead, evaluating models across varying inference budgets—measured in tokens, cost, or time—reveals their real capabilities and safety risks, prompting a shift toward performance‑cost curves and new industry standards.

AI evaluationAI safetyBenchmarking
0 likes · 13 min read
Can a $10 Million Inference Budget Uncover AI’s Real Upper Limit?
DataFunSummit
DataFunSummit
Jun 9, 2026 · Artificial Intelligence

From Gut Feelings to Measurable Metrics: Practicing the Rubrics‑Based Expert Knowledge Extraction and Annotation System CRAFT

The article analyzes the growing difficulty of evaluating large AI models, critiques traditional RLVR and RLHF approaches, introduces a Rubrics‑based evaluation paradigm, describes the design and three‑stage workflow of the CRAFT system, reports math‑domain experiments showing up to 6.2 percentage‑point gains, and outlines future extensions to other domains.

AI evaluationCRAFTRubrics
0 likes · 14 min read
From Gut Feelings to Measurable Metrics: Practicing the Rubrics‑Based Expert Knowledge Extraction and Annotation System CRAFT
PaperAgent
PaperAgent
Jun 9, 2026 · Artificial Intelligence

Why Small Models Can Never Match Large Models, Even with Unlimited Data

The article analyzes scaling laws and synthetic experiments to show that, due to power‑law data distributions and interference, some tasks remain unreachable for small models even with infinite data, a finding confirmed on real LLMs such as OLMo.

interferencelarge language modelsmodel capacity
0 likes · 10 min read
Why Small Models Can Never Match Large Models, Even with Unlimited Data
Data Party THU
Data Party THU
Jun 8, 2026 · Artificial Intelligence

Can Large Language Models Design Chemical Synthesis? ChemReason‑Bench Exposes AI’s Logic Gaps

The ChemReason‑Bench benchmark, introduced by Shanghai Jiao Tong University, evaluates large language models on six program‑reasoning tasks for chemical synthesis, revealing that while top general models show modest reasoning ability, step‑completion remains difficult and domain‑specific models lag behind, prompting new training datasets for improvement.

AI chemistryChemReason-Benchbenchmark
0 likes · 8 min read
Can Large Language Models Design Chemical Synthesis? ChemReason‑Bench Exposes AI’s Logic Gaps
Data Party THU
Data Party THU
Jun 7, 2026 · Artificial Intelligence

When Long Prompts Cause Forgetting: Understanding Generalization in In‑Context Continual Learning

The paper introduces a theoretical framework for In‑Context Continual Learning, showing how shared attention in large language models creates bias, variance, and a novel interference term that explains why longer prompts can lead to forgetting, and provides concrete guidelines for prompt design based on task similarity, context length, and order.

Attention MechanismContinual Learningbias-variance-interference
0 likes · 25 min read
When Long Prompts Cause Forgetting: Understanding Generalization in In‑Context Continual Learning
Machine Heart
Machine Heart
Jun 7, 2026 · Artificial Intelligence

How GoS Gives Agents a Shared Belief State for True Multi-Agent Collaboration

The paper introduces Graph of States (GoS), a neural‑symbolic framework that equips multi‑agent systems with an explicit, maintainable belief state, enabling backtracking and drill‑down during long‑horizon abductive tasks such as medical diagnosis and distributed‑system fault analysis, and demonstrates superior Match and Relevant scores over existing baselines.

AIOpsAbductive Reasoningcausal graph
0 likes · 11 min read
How GoS Gives Agents a Shared Belief State for True Multi-Agent Collaboration
DataFunTalk
DataFunTalk
Jun 7, 2026 · Artificial Intelligence

Exploring Multimodal GraphRAG: Combining Document Intelligence, Knowledge Graphs, and Large Models

This article presents a comprehensive technical analysis of multimodal GraphRAG, covering document‑intelligence parsing pipelines, multimodal graph indexing, retrieval‑generation workflows, knowledge‑graph enhancements for chunk relations, and a detailed comparison of RAG, GraphRAG, and KG‑QA approaches.

GraphRAGKnowledge GraphMultimodal
0 likes · 26 min read
Exploring Multimodal GraphRAG: Combining Document Intelligence, Knowledge Graphs, and Large Models
Machine Heart
Machine Heart
Jun 7, 2026 · Artificial Intelligence

FusionRoute: Token-Level Expert Routing and Self-Correction for Multi-LLM Collaboration

FusionRoute introduces a token‑level routing framework that dynamically selects the most suitable expert LLM for each token and adds a complementary generation step, enabling fine‑grained, stable multi‑model collaboration that outperforms existing sequence‑level and expert‑selection methods across diverse benchmarks.

AI researchModel MergingSelf-Correction
0 likes · 11 min read
FusionRoute: Token-Level Expert Routing and Self-Correction for Multi-LLM Collaboration
Model Perspective
Model Perspective
Jun 6, 2026 · Artificial Intelligence

Why Exam Proctors Are Targeting Smart Glasses for Cheating Prevention

The article analyzes how rapidly advancing smart‑glass technology, combined with large AI models, enables sophisticated cheating in the Chinese college entrance exam, examines market growth, outlines the evolution of cheating methods, and evaluates both exam‑room defenses and AI platform countermeasures.

AI cheatingEducation TechnologyGPT-5.2
0 likes · 9 min read
Why Exam Proctors Are Targeting Smart Glasses for Cheating Prevention
DataFunTalk
DataFunTalk
Jun 6, 2026 · Artificial Intelligence

How Knora Uses Ontology + Large Models to Overcome Enterprise AI Hallucinations and Execution Gaps

The article explains how Knora 4.0 combines ontology with large‑model AI to address six core challenges of enterprise AI—hallucinations, unstable output, weak planning, poor responsiveness, data integration, and long cold‑start—by structuring business knowledge, defining executable actions, and deploying autonomous agents that close the analysis‑decision‑execution loop.

AI platformEnterprise AIKnowledge Graph
0 likes · 16 min read
How Knora Uses Ontology + Large Models to Overcome Enterprise AI Hallucinations and Execution Gaps
Alimama Tech
Alimama Tech
Jun 4, 2026 · Artificial Intelligence

ICML 2026 Highlights: Five Taotian Group Papers Pushing Multimodal AI Boundaries

The article showcases five ICML 2026 papers from the Taotian Group that tackle core multimodal AI challenges—interactive video try‑on, high‑resolution vision, e‑commerce video reasoning, sparse‑reward reinforcement learning, and curriculum learning for large language models—detailing their problem statements, novel solutions, and strong experimental results.

ICML 2026benchmarkcurriculum learning
0 likes · 15 min read
ICML 2026 Highlights: Five Taotian Group Papers Pushing Multimodal AI Boundaries
Alibaba Cloud Native
Alibaba Cloud Native
Jun 3, 2026 · Operations

How Ontology Can Help Enterprises Overcome Token‑Maxxing Costs

This article analyses why AI agents consume massive token budgets—showing that input tokens dominate costs, presenting data from academic papers, industry benchmarks, and Reddit traces, and demonstrating how ontology‑driven solutions like UModel and STAROps can dramatically reduce token usage in real‑world operations.

AIOpsDependency ExplorationSTAROps
0 likes · 15 min read
How Ontology Can Help Enterprises Overcome Token‑Maxxing Costs
Data Party THU
Data Party THU
Jun 2, 2026 · Artificial Intelligence

When AI Starts Evolving Itself: Recursive Self‑Improvement Is Emerging Far Faster Than the Singularity

The article examines how recent advances in large language models, AutoML, and evolutionary algorithms are pushing AI toward recursive self‑improvement, outlines current capabilities and limitations, and discusses the technical, economic, and safety challenges that still prevent a fully autonomous intelligence explosion.

AI safetyArtificial IntelligenceAutoML
0 likes · 10 min read
When AI Starts Evolving Itself: Recursive Self‑Improvement Is Emerging Far Faster Than the Singularity
Machine Heart
Machine Heart
Jun 2, 2026 · Artificial Intelligence

Training Transformers to Be Compression‑Friendly: A New Memory‑Discard Paradigm

The article analyzes the KV‑Cache memory bottleneck of long‑context Transformers, introduces the KV‑CAT (KV‑Compression Aware Training) approach that simulates cache compression during pre‑training, and presents experiments showing unchanged base abilities while dramatically improving post‑training compression, retrieval and long‑text QA performance.

KV CacheKV-CATMemory Efficiency
0 likes · 10 min read
Training Transformers to Be Compression‑Friendly: A New Memory‑Discard Paradigm
Machine Learning Algorithms & Natural Language Processing
Machine Learning Algorithms & Natural Language Processing
Jun 1, 2026 · Artificial Intelligence

MetaAgent-X Enables Agents to Self‑Evolve: A New Paradigm for Native Collaboration

MetaAgent‑X integrates system design and execution within a single base model, using hierarchical rollout and stagewise co‑evolution to jointly train Designer and Executor roles, and achieves significant gains over single‑agent and prior multi‑agent baselines on math and code benchmarks.

AI collaborationMetaAgent-Xhierarchical rollout
0 likes · 13 min read
MetaAgent-X Enables Agents to Self‑Evolve: A New Paradigm for Native Collaboration
DeepHub IMBA
DeepHub IMBA
Jun 1, 2026 · Artificial Intelligence

The Essence of Prompt Engineering: Roles, Tasks, Context, Format, and Constraints

Prompt engineering designs inputs for large language models by combining clear intent, relevant context, explicit format, and constraints, turning ambiguous queries into reliable, high‑quality outputs through a structured, iterative process illustrated with concrete examples and advanced techniques.

AI communicationLLM reliabilityPrompt design
0 likes · 23 min read
The Essence of Prompt Engineering: Roles, Tasks, Context, Format, and Constraints
Data Party THU
Data Party THU
Jun 1, 2026 · Artificial Intelligence

How Steering Unlocks Controllable Large Models: Mechanisms, Evaluation, and Open‑Source Tools

This article reviews two ACL 2026 papers that explain why steering works for large language models, introduce a three‑stage behavior model and activation‑manifold hypothesis, propose the SPLIT method, present the SteerEval evaluation framework, and describe the EasyEdit2 open‑source toolkit.

Activation ManifoldEasyEdit2Model Control
0 likes · 13 min read
How Steering Unlocks Controllable Large Models: Mechanisms, Evaluation, and Open‑Source Tools
AI Large-Model Wave and Transformation Guide
AI Large-Model Wave and Transformation Guide
Jun 1, 2026 · Industry Insights

AI Weekly Roundup (May 25‑31 2026): OpenAI Solves 80‑Year Math Problem, Anthropic Hits $109B Revenue and $300B Funding, Plus Major Industry Moves

From OpenAI's autonomous proof of the 80‑year‑old Erdős unit‑distance conjecture to Anthropic's $109 billion Q2 revenue, a $300 billion financing round surpassing OpenAI, OpenAI's confidential IPO filing, the Pope's first AI encyclical, Karpathy joining Anthropic, Google Gemini's creative‑tool integration, a revamped Google Search bar, Intuit's 17% layoff, and new AI safety guidelines, this week’s AI landscape is reshaped by breakthroughs, funding milestones, and policy shifts.

AI PolicyAI industryAnthropic
0 likes · 11 min read
AI Weekly Roundup (May 25‑31 2026): OpenAI Solves 80‑Year Math Problem, Anthropic Hits $109B Revenue and $300B Funding, Plus Major Industry Moves
AI Engineer Programming
AI Engineer Programming
Jun 1, 2026 · Artificial Intelligence

Why AI Forgets Your Input and How to Fix It

The article explains that large language models have a limited context window, causing the “lost in the middle” effect where information in the middle of long inputs is ignored, and offers practical strategies such as using larger windows, chunking, summarizing, positioning key data, and caching to mitigate forgetting.

RAGcontext windowlarge language models
0 likes · 12 min read
Why AI Forgets Your Input and How to Fix It
Machine Heart
Machine Heart
May 31, 2026 · Artificial Intelligence

Defining a Good Answer in the Agent Era: A Rubrics Survey

This survey examines how rubrics can decompose the vague notion of a "good answer" for large language models into concrete, multi‑dimensional evaluation criteria, detailing their definition, construction methods, applications in training and evaluation, and the open challenges they present.

AI AlignmentAgentic AIRubrics
0 likes · 13 min read
Defining a Good Answer in the Agent Era: A Rubrics Survey
Architect's Guide
Architect's Guide
May 31, 2026 · Artificial Intelligence

10 Hot Open‑Source AI Projects on GitHub This Week (Last One Praised by Jensen Huang)

This article reviews the ten fastest‑growing open‑source AI projects on GitHub over the past week, detailing each project's core capabilities, architecture, and impact while highlighting three emerging trends: AI agents becoming production tools, the rise of edge and lightweight deployments, and accelerated open‑source contributions from major tech firms.

AI AgentsEdge AIMultimodal
0 likes · 22 min read
10 Hot Open‑Source AI Projects on GitHub This Week (Last One Praised by Jensen Huang)
Machine Heart
Machine Heart
May 30, 2026 · Artificial Intelligence

From 6 to 8: DeliAutoResearch SKILL’s Leap in Continual Learning and Self‑Iteration

The paper presents a unified three‑axis framework for continual learning and self‑iteration, classifies over a hundred prior works into five method categories, formalizes convergence conditions, highlights a jump from a 6‑point to an 8‑point peer‑review score, and outlines six open research challenges for autonomous LLMs.

AI autonomyContinual Learninglarge language models
0 likes · 11 min read
From 6 to 8: DeliAutoResearch SKILL’s Leap in Continual Learning and Self‑Iteration
Machine Heart
Machine Heart
May 30, 2026 · Artificial Intelligence

How Abstract Symbols Cut AI Inference Cost by 11×

The article examines IBM Research's Abstract‑CoT approach, which replaces verbose natural‑language chain‑of‑thought reasoning with a compact abstract token vocabulary, achieving up to an 11‑fold reduction in inference tokens while maintaining comparable accuracy across math, instruction‑following, and multi‑hop QA benchmarks.

AI inferenceAbstract-CoTchain-of-thought
0 likes · 11 min read
How Abstract Symbols Cut AI Inference Cost by 11×
Data Party THU
Data Party THU
May 30, 2026 · Artificial Intelligence

How USTC’s Tiny LCPO Training Cuts Large Model Overthinking in Half

The paper introduces LCPO, a lightweight preference‑optimization technique that uses only 800 training examples and 50 steps to teach large language models to produce concise, accurate answers, halving inference length while often improving accuracy and reducing training cost by up to two orders of magnitude.

LCPOLow-Resource Trainingefficient inference
0 likes · 8 min read
How USTC’s Tiny LCPO Training Cuts Large Model Overthinking in Half
Machine Heart
Machine Heart
May 30, 2026 · Artificial Intelligence

Solving AdamW & Muon Instability: Pion Optimizer Updates Large Models on an Iso‑Spectral Manifold

The Pion optimizer leverages iso‑spectral manifold updates to preserve the spectral norm of weight matrices, eliminating additive‑update instability and enabling stable, efficient training of billion‑parameter LLMs across pre‑training, fine‑tuning, and reinforcement‑learning stages, outperforming AdamW and Muon.

AdamWMuonPion optimizer
0 likes · 14 min read
Solving AdamW & Muon Instability: Pion Optimizer Updates Large Models on an Iso‑Spectral Manifold
Machine Heart
Machine Heart
May 29, 2026 · Artificial Intelligence

How Meta’s AI Consumed 183 Billion Tokens to Build a Massive Lean Math Library

Meta’s ATLAS project uses the AutoformBot pipeline to automatically translate 26 undergraduate and graduate math textbooks into a Lean codebase of over 630,000 lines, consuming more than 183 billion tokens, while exposing coverage statistics, adversarial dynamics, and model‑level performance trade‑offs.

AtlasAutoformBotFormal Verification
0 likes · 11 min read
How Meta’s AI Consumed 183 Billion Tokens to Build a Massive Lean Math Library